1 00:00:00,280 --> 00:00:02,400 Speaker 1: Welcome to another edition of the Giants Little Podcast, brought 2 00:00:02,440 --> 00:00:04,560 Speaker 1: to you by Citizens, the official bank of the Giants. 3 00:00:04,600 --> 00:00:07,000 Speaker 1: I am John Schmelke, coming to you from the Giants 4 00:00:07,000 --> 00:00:10,560 Speaker 1: Podcast Studio, presented by Hackensack and maridein Health Keep getting 5 00:00:10,680 --> 00:00:13,600 Speaker 1: better fun guests today, And I mentioned this platform a 6 00:00:13,640 --> 00:00:16,720 Speaker 1: little bit earlier in the draft process. NFL launched a 7 00:00:16,760 --> 00:00:19,320 Speaker 1: new platform this year, nfliq. You can find the NFL 8 00:00:19,360 --> 00:00:22,520 Speaker 1: dot com slash iq. One of the architects behind it 9 00:00:22,800 --> 00:00:25,120 Speaker 1: is the senior manager of Research and Analytics for Next 10 00:00:25,120 --> 00:00:28,000 Speaker 1: Gen Stats, Keegan abdu Keegan, what's going on? 11 00:00:28,080 --> 00:00:28,720 Speaker 2: Man? How are you? 12 00:00:29,120 --> 00:00:31,560 Speaker 3: Thanks for having me on, John, Yeah, So. 13 00:00:31,840 --> 00:00:33,400 Speaker 1: We try to get nitty gritty on some of the 14 00:00:33,400 --> 00:00:36,440 Speaker 1: stuff so fans understand, you know, people like throwing these 15 00:00:36,479 --> 00:00:38,320 Speaker 1: numbers out there and fans will understand what they mean 16 00:00:38,400 --> 00:00:40,320 Speaker 1: and how you get them. So I figured this would 17 00:00:40,320 --> 00:00:42,280 Speaker 1: be a really good good way to do that. We're 18 00:00:42,320 --> 00:00:44,840 Speaker 1: talking about nfl Q some next Gen stats as well. 19 00:00:45,080 --> 00:00:47,120 Speaker 1: So Ke, let's start here. When you guys were putting 20 00:00:47,200 --> 00:00:51,479 Speaker 1: nfl IQ together, why and what were the goals you 21 00:00:51,520 --> 00:00:53,120 Speaker 1: had in mind when you were trying to build it out. 22 00:00:53,880 --> 00:00:56,320 Speaker 3: So the why is a great question to start off, 23 00:00:56,360 --> 00:00:59,320 Speaker 3: so next gen stats. You know, we collect data during 24 00:00:59,400 --> 00:01:01,600 Speaker 3: real game, but what do we do in the off season. 25 00:01:01,680 --> 00:01:04,520 Speaker 3: So a few years back we started by building out 26 00:01:04,560 --> 00:01:08,120 Speaker 3: our draft scores, and then last year we launched Combine 27 00:01:08,120 --> 00:01:12,320 Speaker 3: iq and draft iq that really gave fans more of 28 00:01:12,680 --> 00:01:15,000 Speaker 3: the data that we're collecting at the Combine Live and 29 00:01:15,040 --> 00:01:17,679 Speaker 3: then heading into the draft, just kind of trying to 30 00:01:17,680 --> 00:01:21,679 Speaker 3: serve fans in the off season by giving them different 31 00:01:21,760 --> 00:01:24,759 Speaker 3: GM tendencies, kind of like an overall look of what 32 00:01:24,800 --> 00:01:28,200 Speaker 3: your team might do in the draft. And heading into 33 00:01:28,200 --> 00:01:31,360 Speaker 3: this offseason, we said, hey, let's might as well do 34 00:01:31,480 --> 00:01:36,039 Speaker 3: the entire timeline of an offseason from free agency all 35 00:01:36,080 --> 00:01:39,319 Speaker 3: the way through draft to see how each team is 36 00:01:39,319 --> 00:01:43,720 Speaker 3: approaching roster management and roster turnover as they build towards 37 00:01:43,760 --> 00:01:45,240 Speaker 3: the twenty twenty sixth season. 38 00:01:46,040 --> 00:01:48,240 Speaker 1: Yeah, and look, if you go to NFL dot com 39 00:01:48,240 --> 00:01:50,240 Speaker 1: slash iq, there's different tabs at the top. You can 40 00:01:50,320 --> 00:01:53,800 Speaker 1: go by teams, there's free agency, there's the draft. So 41 00:01:54,600 --> 00:01:56,520 Speaker 1: let's start with the team stuff first, then we'll kind 42 00:01:56,520 --> 00:01:59,360 Speaker 1: of slide into draft lend here key. Yeah, when you 43 00:01:59,400 --> 00:02:01,240 Speaker 1: look at the team pages, you have the roster or 44 00:02:01,240 --> 00:02:03,960 Speaker 1: the free agent board, you have a draft preview talk 45 00:02:04,000 --> 00:02:06,440 Speaker 1: about what some of the fans should should make sure 46 00:02:06,480 --> 00:02:09,000 Speaker 1: that they cycle to and look at when they go 47 00:02:09,160 --> 00:02:12,120 Speaker 1: to their favorite team's page Giant Fans in this case 48 00:02:12,400 --> 00:02:13,920 Speaker 1: on NFL dot Com slash iq. 49 00:02:14,800 --> 00:02:16,799 Speaker 3: Well, I think you start out with what the roster 50 00:02:16,919 --> 00:02:20,280 Speaker 3: is right now, you know, maybe a quick little refresh 51 00:02:20,400 --> 00:02:22,880 Speaker 3: who might have left and who's still there, who are 52 00:02:22,880 --> 00:02:25,360 Speaker 3: the projected starters, and then you can kind of start 53 00:02:25,360 --> 00:02:28,280 Speaker 3: to piece together, you know, what might the areas of 54 00:02:28,400 --> 00:02:31,120 Speaker 3: needs be as we had in the draft. So on 55 00:02:31,160 --> 00:02:33,160 Speaker 3: the bottom of the page was a team needs section 56 00:02:33,600 --> 00:02:36,200 Speaker 3: and that'll go through as identified by NFL dot Com 57 00:02:36,200 --> 00:02:38,919 Speaker 3: the top five needs that we're heading into free agency 58 00:02:38,960 --> 00:02:42,920 Speaker 3: and how each team might have addressed those already, and 59 00:02:43,000 --> 00:02:46,399 Speaker 3: kind of now shifting focus towards the draft, what might be, 60 00:02:46,560 --> 00:02:49,079 Speaker 3: you know, the top areas that a team wants to 61 00:02:49,120 --> 00:02:52,360 Speaker 3: improve on if they are drafting from needing to fill 62 00:02:52,400 --> 00:02:55,200 Speaker 3: holes rather than you know, best player available, which many 63 00:02:55,280 --> 00:02:59,040 Speaker 3: gms do as well. We also have you know, contract value, 64 00:02:59,440 --> 00:03:03,480 Speaker 3: cap spending, et cetera, as well as your draft capital 65 00:03:03,520 --> 00:03:05,680 Speaker 3: over the next couple of drafts to really see what 66 00:03:05,800 --> 00:03:10,440 Speaker 3: sort of resources you're working with. As you look towards 67 00:03:10,520 --> 00:03:12,000 Speaker 3: you know what your team is going to do in 68 00:03:12,040 --> 00:03:15,480 Speaker 3: the draft. Another cool part of this is the mock 69 00:03:15,560 --> 00:03:19,760 Speaker 3: Draft tracker. So we have partnered with Grinding the Mox, 70 00:03:19,800 --> 00:03:24,040 Speaker 3: who has scraped every single mock draft that you could 71 00:03:24,200 --> 00:03:26,799 Speaker 3: ever think as possible on the internet. I believe there's 72 00:03:26,840 --> 00:03:31,680 Speaker 3: like over one thousand per prospect essentially, and you can see, 73 00:03:31,720 --> 00:03:34,400 Speaker 3: you know, for the Giants specifically, who the top three 74 00:03:34,440 --> 00:03:37,320 Speaker 3: players are that are being mocked to the Giants right now? 75 00:03:37,480 --> 00:03:40,240 Speaker 3: Three Ohio State guys. I'm sure it's no surprise to you, 76 00:03:40,280 --> 00:03:42,760 Speaker 3: but it's Carnel Tate Kale downs in Sunny Styles. 77 00:03:43,480 --> 00:03:44,720 Speaker 2: Yeah, no doubt about it. 78 00:03:44,760 --> 00:03:47,040 Speaker 1: You mentioned earlier that you guys in the past, and 79 00:03:47,080 --> 00:03:49,640 Speaker 1: maybe it's it's part of some separate maybe it's hidden 80 00:03:49,680 --> 00:03:54,360 Speaker 1: here somewhere. Trends and draft history for gms and certain organizations. 81 00:03:54,360 --> 00:03:56,040 Speaker 1: Do you guys have that included in here as well. 82 00:03:56,720 --> 00:03:58,920 Speaker 3: Yeah, so we're still gathering that. You know, there's new 83 00:03:58,960 --> 00:04:01,800 Speaker 3: gms every year. Yeah, we got to update that information, 84 00:04:02,000 --> 00:04:05,080 Speaker 3: but that will be launched ahead of the draft, and 85 00:04:05,160 --> 00:04:07,920 Speaker 3: we're trying to get even further down the list of 86 00:04:07,960 --> 00:04:10,640 Speaker 3: like what their draft history is, like what sort of 87 00:04:10,680 --> 00:04:14,640 Speaker 3: thresholds these gms and coaches might look for in certain positions, 88 00:04:14,640 --> 00:04:18,159 Speaker 3: and players are looking for longer cornerbacks. Do they have 89 00:04:18,240 --> 00:04:21,920 Speaker 3: a minimum arm length for their offensive tackles that they're drafting? 90 00:04:22,760 --> 00:04:26,000 Speaker 3: How does you know body size and wait for edge players, 91 00:04:26,000 --> 00:04:29,200 Speaker 3: et cetera. Like that, we'll also have tendencies of you know, 92 00:04:29,240 --> 00:04:32,599 Speaker 3: their total draft history in terms of how you know 93 00:04:32,640 --> 00:04:36,600 Speaker 3: what positions they have valued the most, and you know, 94 00:04:36,640 --> 00:04:39,359 Speaker 3: are they a team that likes to trade up or 95 00:04:39,400 --> 00:04:42,640 Speaker 3: trade down or has that changed recently in the past 96 00:04:42,680 --> 00:04:43,120 Speaker 3: as well. 97 00:04:43,720 --> 00:04:46,040 Speaker 1: I love that because, honestly, I was gonna have one 98 00:04:46,040 --> 00:04:47,920 Speaker 1: of my guys here try to go back the last 99 00:04:47,920 --> 00:04:49,880 Speaker 1: few years and track that manually. And now you just 100 00:04:49,880 --> 00:04:51,760 Speaker 1: saved them awful lot of work because I know that's 101 00:04:51,760 --> 00:04:55,080 Speaker 1: going to be popping up on NFL dot cop slash iq, 102 00:04:55,160 --> 00:04:55,799 Speaker 1: which is great. 103 00:04:55,960 --> 00:04:58,760 Speaker 2: All right, let's footboard of the draft tab. Now for 104 00:04:58,839 --> 00:04:59,240 Speaker 2: people that. 105 00:04:59,240 --> 00:05:00,920 Speaker 1: Maybe you're listen to the and they're on their computer, 106 00:05:01,920 --> 00:05:04,400 Speaker 1: it's fantastic. So the first thing that you notice for 107 00:05:04,440 --> 00:05:06,520 Speaker 1: people that want to know what happened in the NFL combine, 108 00:05:06,520 --> 00:05:09,440 Speaker 1: you guys have every single result of every single drill, 109 00:05:09,600 --> 00:05:14,159 Speaker 1: every single measurement, all in one place, and then you 110 00:05:14,200 --> 00:05:17,719 Speaker 1: have something called a raw athletic score. Talk about how 111 00:05:17,720 --> 00:05:20,320 Speaker 1: you kind of put that together? You have an athleticism rank. 112 00:05:20,400 --> 00:05:22,479 Speaker 1: Then I think, as well, how do you kind of 113 00:05:22,480 --> 00:05:24,839 Speaker 1: put all those numbers together to try to come up 114 00:05:24,880 --> 00:05:28,080 Speaker 1: with those composite athleticism scores that you guys have here 115 00:05:28,080 --> 00:05:28,800 Speaker 1: and what do they mean? 116 00:05:29,640 --> 00:05:32,839 Speaker 3: Yeah, So for the athleticism and the production scores, those 117 00:05:32,880 --> 00:05:38,880 Speaker 3: are our draft models, and those are essentially a regression tree. 118 00:05:39,000 --> 00:05:43,839 Speaker 3: So basically it finds Keith thresholds for each position that 119 00:05:44,200 --> 00:05:48,000 Speaker 3: you translate to NFL success. And NFL success as we've 120 00:05:48,000 --> 00:05:50,880 Speaker 3: defined it is where you a starter and were you 121 00:05:50,960 --> 00:05:53,599 Speaker 3: a pro bowler in your first three years in the NFL. 122 00:05:53,960 --> 00:05:55,960 Speaker 3: So we go back to two thousand and three and 123 00:05:56,000 --> 00:05:58,359 Speaker 3: we take all this data and we build a model 124 00:05:58,400 --> 00:06:02,200 Speaker 3: for every single position and see, basically what are the 125 00:06:02,240 --> 00:06:04,880 Speaker 3: key thresholds. And I think this is a good way 126 00:06:05,000 --> 00:06:08,080 Speaker 3: of looking at prospects because it doesn't really especially when 127 00:06:08,120 --> 00:06:10,599 Speaker 3: it comes to athleticism, it doesn't really matter in terms 128 00:06:10,640 --> 00:06:13,760 Speaker 3: of translating to directly to onfield success. If you are 129 00:06:13,800 --> 00:06:16,000 Speaker 3: the fastest player at your position, you just need to 130 00:06:16,000 --> 00:06:19,039 Speaker 3: be fast enough so you know every year we have 131 00:06:19,120 --> 00:06:21,560 Speaker 3: these combine workout warriors that come in and you know, 132 00:06:21,960 --> 00:06:25,320 Speaker 3: hit new records, but like, how often are those combine 133 00:06:25,320 --> 00:06:28,240 Speaker 3: workout warriors actually the best players in the NFL? You know, 134 00:06:28,640 --> 00:06:31,120 Speaker 3: it's a it's a proxy for athleticism, but it's not 135 00:06:31,800 --> 00:06:35,960 Speaker 3: the full picture. So by kind of working how all 136 00:06:36,000 --> 00:06:39,480 Speaker 3: of these different features work together. You know, what's your 137 00:06:39,520 --> 00:06:44,040 Speaker 3: weight compared to your speed, what's your you know, bursts 138 00:06:44,080 --> 00:06:47,080 Speaker 3: as measured by a broad jump compared to how you 139 00:06:47,080 --> 00:06:50,160 Speaker 3: know straight line speed or your agility and short shuttles 140 00:06:50,160 --> 00:06:52,720 Speaker 3: and everything, and how does that relate to all prospects 141 00:06:52,720 --> 00:06:55,920 Speaker 3: over the last twenty three years and how they've actually 142 00:06:55,920 --> 00:06:59,920 Speaker 3: had success in the NFL. So that's kind of how 143 00:07:00,200 --> 00:07:04,279 Speaker 3: what's working under the lens of each of these models. 144 00:07:04,320 --> 00:07:07,960 Speaker 3: And then we translate those probabilities of becoming a starter 145 00:07:08,160 --> 00:07:11,480 Speaker 3: or a pro bowler into a more digestible fifty to 146 00:07:11,600 --> 00:07:14,880 Speaker 3: ninety nine scale allah like Madden ratings or something to 147 00:07:14,880 --> 00:07:17,280 Speaker 3: make it a little bit more fan friendly in that way. 148 00:07:17,400 --> 00:07:19,840 Speaker 3: And then those two models are also kind of fed 149 00:07:20,120 --> 00:07:23,000 Speaker 3: in along with a bunch of other features including size 150 00:07:23,320 --> 00:07:28,120 Speaker 3: and metrics like that, into an overall draft score as 151 00:07:28,120 --> 00:07:31,040 Speaker 3: well as you know, having a composite scouts grade which 152 00:07:31,080 --> 00:07:34,640 Speaker 3: we take as a proxy just overall, combine big board 153 00:07:34,920 --> 00:07:38,120 Speaker 3: rankings from a consensus standpoint, and that's how we get 154 00:07:38,160 --> 00:07:40,160 Speaker 3: to our final overall draft score. 155 00:07:40,400 --> 00:07:42,400 Speaker 1: Okay, interesting, I want to break that down to a 156 00:07:42,400 --> 00:07:44,960 Speaker 1: couple different questions after For I think that that that's 157 00:07:45,080 --> 00:07:46,640 Speaker 1: very complete and is more in there than I thought. 158 00:07:46,880 --> 00:07:50,960 Speaker 1: So athleticism, does that only count testing or does that 159 00:07:51,080 --> 00:07:52,640 Speaker 1: also count measurements? 160 00:07:54,280 --> 00:07:58,240 Speaker 3: It does? It does count measurements for accounting for height 161 00:07:58,240 --> 00:08:00,560 Speaker 3: and weight in there. We're not necessarily counting for you know, 162 00:08:00,640 --> 00:08:02,760 Speaker 3: hand size or arm length in there. That's not really 163 00:08:02,800 --> 00:08:04,960 Speaker 3: what we think of as leticism. That's more in the 164 00:08:05,000 --> 00:08:07,920 Speaker 3: overall model. But you know, I think if you're looking 165 00:08:07,960 --> 00:08:11,480 Speaker 3: at athleticism and you're not accounting for weight, then you 166 00:08:11,520 --> 00:08:16,200 Speaker 3: know it's so important in terms of you know, Xavier Worthy, 167 00:08:16,200 --> 00:08:18,040 Speaker 3: we might have ran the fastest forty, but he was 168 00:08:18,080 --> 00:08:21,200 Speaker 3: one hundred and you know, seventy eight pounds tough to 169 00:08:21,200 --> 00:08:24,080 Speaker 3: beat press coverage there. So weight and how it relates 170 00:08:24,120 --> 00:08:27,080 Speaker 3: to speed is I think an essential part of any 171 00:08:27,400 --> 00:08:29,080 Speaker 3: draft prospect. There. 172 00:08:29,440 --> 00:08:31,320 Speaker 1: No, I think that makes total sense, Like there might 173 00:08:31,320 --> 00:08:33,960 Speaker 1: be a more impressive forty score for a two orches 174 00:08:34,000 --> 00:08:38,200 Speaker 1: sixty pounder, even though his overall time is obviously much, 175 00:08:38,520 --> 00:08:41,000 Speaker 1: you know, slower than say, is Xavier Worthies. 176 00:08:41,080 --> 00:08:43,760 Speaker 3: Right, And it honestly helps a lot with the historical 177 00:08:44,520 --> 00:08:47,960 Speaker 3: kind of comparisons because players have gotten lighter and faster 178 00:08:48,080 --> 00:08:51,280 Speaker 3: every year pretty much like across the board. This was 179 00:08:51,400 --> 00:08:55,199 Speaker 3: the lightest pretty much at every position, the lightest draft class, 180 00:08:55,559 --> 00:08:58,720 Speaker 3: and the fastest except for I believe, like defensive tackle 181 00:08:58,800 --> 00:09:01,800 Speaker 3: an offensive linemen, they're actually and bigger in terms of weight. 182 00:09:01,840 --> 00:09:04,240 Speaker 3: But like it does put into context, hey, all these 183 00:09:04,280 --> 00:09:06,920 Speaker 3: guys are running these blazing fast speeds, but they're also 184 00:09:07,000 --> 00:09:09,520 Speaker 3: lighter than they have been in the past, and that's 185 00:09:09,559 --> 00:09:12,920 Speaker 3: a you know, important impact into how the league is 186 00:09:13,000 --> 00:09:13,680 Speaker 3: changing as well. 187 00:09:14,000 --> 00:09:17,080 Speaker 1: No, I think that makes total sense. And then I'm 188 00:09:17,120 --> 00:09:23,440 Speaker 1: assuming by position, certain metrics are more important than others. 189 00:09:23,520 --> 00:09:25,679 Speaker 1: And you managed to kind of figure that out through 190 00:09:25,679 --> 00:09:29,360 Speaker 1: your aggression formula, right, based on what position, what traits, 191 00:09:29,800 --> 00:09:33,280 Speaker 1: and what athletics scores are most important for each individual position. 192 00:09:33,400 --> 00:09:35,959 Speaker 1: So probably the model for each position has got to 193 00:09:35,960 --> 00:09:36,920 Speaker 1: be a little bit different then. 194 00:09:36,880 --> 00:09:39,400 Speaker 3: Right, very different. Yeah, and it's only trained on that 195 00:09:39,480 --> 00:09:42,200 Speaker 3: position's data, So I mean, you could imagine that a 196 00:09:42,320 --> 00:09:44,120 Speaker 3: forty yard dash is going to be a lot more 197 00:09:44,120 --> 00:09:48,400 Speaker 3: important for a wide receiver than an interior offensive lineman, 198 00:09:48,679 --> 00:09:52,040 Speaker 3: or you know, a short shuttle speed is going to 199 00:09:52,080 --> 00:09:54,520 Speaker 3: be a lot more important for an edge rusher than 200 00:09:54,840 --> 00:09:57,280 Speaker 3: you know, maybe a tight end or something along those lines. 201 00:09:57,320 --> 00:10:00,000 Speaker 3: And we can see those in We call it shat 202 00:10:00,120 --> 00:10:03,760 Speaker 3: plots in terms of the variable importance for each position, 203 00:10:04,200 --> 00:10:07,480 Speaker 3: in terms of what are the most important drills that 204 00:10:07,520 --> 00:10:11,400 Speaker 3: predict future success at those positions. And again we're taking 205 00:10:11,440 --> 00:10:15,319 Speaker 3: it from a threshold perspective, not from a pure one 206 00:10:15,360 --> 00:10:19,720 Speaker 3: to one, so the outliers are not you know, dominating 207 00:10:19,800 --> 00:10:23,240 Speaker 3: as much in terms of how it translates you just 208 00:10:23,440 --> 00:10:26,600 Speaker 3: how important is hitting a certain threshold in this measurable 209 00:10:26,800 --> 00:10:29,640 Speaker 3: to your NFL success rather than how important is being 210 00:10:29,679 --> 00:10:32,600 Speaker 3: the best at this specific drill to your NFL success. 211 00:10:33,240 --> 00:10:36,720 Speaker 1: Yeah, you're basically trying to take exceptions out of the way, right. 212 00:10:36,760 --> 00:10:39,800 Speaker 1: I mean, I'm gonna caute Bill Parcels back in the day. 213 00:10:39,800 --> 00:10:42,240 Speaker 1: He goes, Look, once you start drafting exceptions, you're gonna 214 00:10:42,240 --> 00:10:43,400 Speaker 1: have a team full of exceptions. 215 00:10:43,440 --> 00:10:46,719 Speaker 2: And might it work now or then, sure. 216 00:10:46,920 --> 00:10:50,040 Speaker 1: But I think the probability to your point of finding 217 00:10:50,120 --> 00:10:55,680 Speaker 1: exceptions at multiple positions every year and having success there. Look, again, 218 00:10:55,720 --> 00:10:58,160 Speaker 1: it's a possible, but it's probably not how you're maximizing 219 00:10:58,160 --> 00:10:59,400 Speaker 1: your probability of success. 220 00:11:00,200 --> 00:11:03,679 Speaker 3: Yeah, and I think in general, any of these analytical 221 00:11:03,720 --> 00:11:07,680 Speaker 3: measures and like overall profiles of an athlete are meant 222 00:11:07,679 --> 00:11:09,680 Speaker 3: to be just a piece of the puzzle. You know, 223 00:11:09,920 --> 00:11:12,319 Speaker 3: it might be a tie breaker, but in the end, 224 00:11:12,400 --> 00:11:15,200 Speaker 3: you really need to rely on your scouts and trust 225 00:11:15,200 --> 00:11:17,680 Speaker 3: your scouts that they're you know, they're the professionals. They're 226 00:11:17,760 --> 00:11:20,440 Speaker 3: catching things that aren't gonna be caught in just these 227 00:11:20,440 --> 00:11:24,800 Speaker 3: peer stats and measurables. But you know, if it comes 228 00:11:24,800 --> 00:11:27,840 Speaker 3: down to half the scouting staff really likes this guy 229 00:11:27,880 --> 00:11:29,600 Speaker 3: and the other really likes this guy at the position, 230 00:11:29,720 --> 00:11:32,320 Speaker 3: but we think that he's a much better athlete as 231 00:11:32,320 --> 00:11:34,120 Speaker 3: by his measurables, maybe you take a chance on the 232 00:11:34,120 --> 00:11:34,920 Speaker 3: better athlete. 233 00:11:35,679 --> 00:11:37,320 Speaker 1: I don't want to make you reveal some of the 234 00:11:37,360 --> 00:11:40,040 Speaker 1: magic behind the math, but what are some of the 235 00:11:40,040 --> 00:11:43,840 Speaker 1: positions maybe where you find either one drill or one 236 00:11:43,880 --> 00:11:48,000 Speaker 1: set of athletic traits are really important for each position 237 00:11:48,120 --> 00:11:50,200 Speaker 1: that that really, as you've gone along and done this, 238 00:11:50,280 --> 00:11:52,920 Speaker 1: that maybe even surprised you where you're like, wow, there's 239 00:11:52,960 --> 00:11:56,800 Speaker 1: a really heavy correlation here between this position and this 240 00:11:56,840 --> 00:11:59,720 Speaker 1: specific you know, athletic number or trait. 241 00:12:00,640 --> 00:12:04,600 Speaker 3: So I do think that for edge rushers, probably the 242 00:12:04,640 --> 00:12:08,719 Speaker 3: most important drill is unfortunately one that people are doing 243 00:12:08,800 --> 00:12:10,760 Speaker 3: less and less of the three cone, right, and that's 244 00:12:10,760 --> 00:12:14,760 Speaker 3: the three cone. So like one of the I think 245 00:12:14,800 --> 00:12:18,640 Speaker 3: really interesting is looking forward, is how do we impute 246 00:12:18,720 --> 00:12:22,400 Speaker 3: that data when less and less players are competing in 247 00:12:22,440 --> 00:12:24,800 Speaker 3: these drills are only doing one So how do we 248 00:12:24,840 --> 00:12:27,680 Speaker 3: get a good estimate for what a player's short shuttle 249 00:12:27,720 --> 00:12:29,400 Speaker 3: or three cone would be. You know, if we had 250 00:12:29,600 --> 00:12:32,080 Speaker 3: player tracking data like we collected at the NFL level 251 00:12:32,840 --> 00:12:35,840 Speaker 3: for college football available, that would probably be a very 252 00:12:35,880 --> 00:12:38,400 Speaker 3: easy way to get those proxies, but we don't right now. 253 00:12:38,520 --> 00:12:40,920 Speaker 3: There are some companies that are collecting that based on 254 00:12:40,960 --> 00:12:43,320 Speaker 3: computer vision, but you know, we don't have at the 255 00:12:43,400 --> 00:12:47,120 Speaker 3: NFL don't have access to that right now. So you know, 256 00:12:47,200 --> 00:12:50,320 Speaker 3: there's there's a few other interesting companies out there that 257 00:12:50,520 --> 00:12:52,600 Speaker 3: you know, force plate data. They might be able to 258 00:12:52,920 --> 00:12:57,360 Speaker 3: meaningly project these drills. And again, the value of combine 259 00:12:57,400 --> 00:13:00,840 Speaker 3: measurables are not that like they're one to one what 260 00:13:00,880 --> 00:13:04,320 Speaker 3: you're doing on a field without any pads on is 261 00:13:04,320 --> 00:13:07,280 Speaker 3: going to directly translate to the NFL. It's the history 262 00:13:07,280 --> 00:13:09,880 Speaker 3: of it. It's having twenty years of data to go 263 00:13:09,960 --> 00:13:13,640 Speaker 3: back and compare to. And that's why I do think 264 00:13:13,679 --> 00:13:16,040 Speaker 3: the combine will not become obsolete even with all these 265 00:13:16,080 --> 00:13:19,720 Speaker 3: other things, is because you know that that historical context 266 00:13:19,760 --> 00:13:21,920 Speaker 3: is just too too valuable to have right now. 267 00:13:22,600 --> 00:13:24,719 Speaker 1: Yeah, every entry you give you there, there are more 268 00:13:24,760 --> 00:13:27,280 Speaker 1: pass me to go down here. So would you think 269 00:13:27,320 --> 00:13:30,320 Speaker 1: at some point we're going to get where some of 270 00:13:30,360 --> 00:13:34,760 Speaker 1: that GPS data from college is either available to NFL, 271 00:13:34,800 --> 00:13:36,040 Speaker 1: because I know I think right now it's by a 272 00:13:36,040 --> 00:13:38,240 Speaker 1: conference by conference bases in college football, which is how 273 00:13:38,240 --> 00:13:39,679 Speaker 1: a lot of the stuff has done now, which makes 274 00:13:39,679 --> 00:13:42,200 Speaker 1: it very difficult as someone that once we draft players 275 00:13:42,200 --> 00:13:44,600 Speaker 1: has to track down college film from these guys. Trust me, Yeah, 276 00:13:44,600 --> 00:13:46,680 Speaker 1: I understand how difficult this time can be. 277 00:13:47,080 --> 00:13:49,760 Speaker 3: It's very easy, very buzantine system. 278 00:13:50,320 --> 00:13:51,400 Speaker 2: No exactly right. 279 00:13:51,679 --> 00:13:54,400 Speaker 1: And the other problem is that each conference is different systems, right, 280 00:13:54,440 --> 00:13:56,840 Speaker 1: so the data sometimes might not be comparable from conference 281 00:13:56,840 --> 00:13:58,760 Speaker 1: to conference. Are we going to get to a point 282 00:13:58,800 --> 00:14:01,200 Speaker 1: at some point where there's a world where even if 283 00:14:01,240 --> 00:14:03,440 Speaker 1: it's not publicly available, there's going to be some type 284 00:14:03,440 --> 00:14:07,280 Speaker 1: of uniform tracking in college where teams can get comparable 285 00:14:07,400 --> 00:14:10,079 Speaker 1: GPS data from player to player, team to team, conference 286 00:14:10,080 --> 00:14:10,600 Speaker 1: to conference. 287 00:14:11,600 --> 00:14:16,640 Speaker 3: Yeah, the the hardware like actual in chips shoulder pads 288 00:14:16,679 --> 00:14:19,000 Speaker 3: like we collect in the NFL with next gen stats, 289 00:14:19,680 --> 00:14:21,920 Speaker 3: that that one might be harder. And with all the 290 00:14:21,960 --> 00:14:24,640 Speaker 3: advances in computer vision, I think that might be you know, 291 00:14:24,680 --> 00:14:27,920 Speaker 3: it might not be as accurate data, but it might 292 00:14:27,960 --> 00:14:30,760 Speaker 3: be good enough. And there are already a few companies 293 00:14:30,760 --> 00:14:33,040 Speaker 3: out there that have been collecting that for the past 294 00:14:33,040 --> 00:14:37,040 Speaker 3: few years and are able to get pretty reasonable predictions 295 00:14:37,080 --> 00:14:40,480 Speaker 3: of athletic profiles based on that data. That's really kind 296 00:14:40,480 --> 00:14:43,680 Speaker 3: of the next frontier. And there's an arms race there, 297 00:14:43,720 --> 00:14:46,480 Speaker 3: you know, with TeamWorks buying PFF and all that kind 298 00:14:46,480 --> 00:14:49,040 Speaker 3: of stuff of how they can use computer vision which 299 00:14:49,080 --> 00:14:53,920 Speaker 3: is sports right or yeah, in order to get that 300 00:14:54,080 --> 00:14:56,760 Speaker 3: college data that they can then sell the teams and 301 00:14:56,920 --> 00:14:59,880 Speaker 3: you know, build these better athletic profiles and just what 302 00:15:00,080 --> 00:15:01,520 Speaker 3: available at the combine. 303 00:15:01,160 --> 00:15:03,640 Speaker 1: Puddle up, get in here, If you're lined up here, 304 00:15:03,760 --> 00:15:06,440 Speaker 1: you gotta go over the middle with it the score great. 305 00:15:06,960 --> 00:15:08,000 Speaker 3: How do we make that happen? 306 00:15:08,840 --> 00:15:12,280 Speaker 1: I don't know, but citizens does makes sense of your 307 00:15:12,320 --> 00:15:16,680 Speaker 1: money with Citizens Official Bank of ELI Manning. All right, 308 00:15:16,720 --> 00:15:21,640 Speaker 1: so just computer vision that's basically judging speed and athleticism 309 00:15:21,880 --> 00:15:24,480 Speaker 1: off of tape and basically kind of just the computer 310 00:15:24,520 --> 00:15:26,480 Speaker 1: figures it out based on what it sees on video. 311 00:15:27,000 --> 00:15:30,400 Speaker 3: Yeah, it's the same like image processing that you know 312 00:15:30,560 --> 00:15:35,680 Speaker 3: maybe a WAYMO uses to aunonymously drive. So like that's 313 00:15:35,720 --> 00:15:38,600 Speaker 3: the kind of stuff that we're looking at, adding computer vision. 314 00:15:38,640 --> 00:15:41,960 Speaker 3: And for the NFL, we actually just started. We put 315 00:15:42,240 --> 00:15:45,000 Speaker 3: hawkeye cameras in the stadium, so rather than basing it 316 00:15:45,040 --> 00:15:49,560 Speaker 3: off film, we're collecting it in stadium using very high 317 00:15:49,560 --> 00:15:54,520 Speaker 3: fidelity cameras. And what that evolves from two dimensional tracking 318 00:15:54,560 --> 00:15:56,160 Speaker 3: data where you just know where they are on the 319 00:15:56,200 --> 00:16:00,760 Speaker 3: field to three dimensional skeletal tracking data where you know 320 00:16:00,800 --> 00:16:04,560 Speaker 3: you can see the actual throwing motion of a quarterback 321 00:16:04,760 --> 00:16:07,080 Speaker 3: or a catch radius, or you know whether a guy 322 00:16:07,400 --> 00:16:11,120 Speaker 3: is jump hurdling someone. Like it's just a it's a 323 00:16:11,120 --> 00:16:13,840 Speaker 3: whole new frontier and data and uh, we're really just 324 00:16:13,840 --> 00:16:16,520 Speaker 3: scratching the surface making sure it's clean enough to use. 325 00:16:16,600 --> 00:16:19,320 Speaker 3: But that's where everything's kind of going in in this space. 326 00:16:20,000 --> 00:16:22,520 Speaker 1: It's funny and you know, we go to the commune, 327 00:16:22,520 --> 00:16:23,840 Speaker 1: we watch the comm at everyone makes a big go 328 00:16:23,880 --> 00:16:25,480 Speaker 1: about the forty yard dash, and look, the forty yard 329 00:16:25,520 --> 00:16:26,200 Speaker 1: dash is important. 330 00:16:26,360 --> 00:16:27,160 Speaker 2: I'm not saying it's not. 331 00:16:27,360 --> 00:16:29,560 Speaker 1: Yeah, but to me, as we watch the league more 332 00:16:29,560 --> 00:16:32,760 Speaker 1: and more, this leads about changing direction and start stop right, 333 00:16:33,000 --> 00:16:35,480 Speaker 1: and that's how you can create separation and get by 334 00:16:35,560 --> 00:16:39,080 Speaker 1: people and create differences. And I do feel like, unfortunately, 335 00:16:39,080 --> 00:16:40,440 Speaker 1: to your point, a lot of the change of direction 336 00:16:40,520 --> 00:16:42,800 Speaker 1: drills guys aren't doing anymore. I think the jumps measure 337 00:16:42,840 --> 00:16:46,320 Speaker 1: some of the start stop explosion ability. Yeah, but as 338 00:16:46,320 --> 00:16:48,640 Speaker 1: you guys try to measure some of this stuff that 339 00:16:49,120 --> 00:16:52,520 Speaker 1: isn't traditional, what are some of the things you're trying 340 00:16:52,560 --> 00:16:54,280 Speaker 1: to do in order to do that, and maybe some 341 00:16:54,320 --> 00:16:56,080 Speaker 1: of the findings you've had as you've tried to kind 342 00:16:56,080 --> 00:16:57,560 Speaker 1: of dig into that stuff a little bit more. 343 00:16:58,240 --> 00:17:00,720 Speaker 3: Yeah, I would say that one of the areas we're 344 00:17:00,720 --> 00:17:04,480 Speaker 3: starting to look into for trying to better impute the 345 00:17:04,560 --> 00:17:08,120 Speaker 3: changing direction skills is you know, all these players, maybe 346 00:17:08,119 --> 00:17:12,040 Speaker 3: they're not performing in the actual shuttles at the end 347 00:17:12,040 --> 00:17:14,000 Speaker 3: of this workout because they're tired and they don't want 348 00:17:14,000 --> 00:17:17,080 Speaker 3: to get a bad time, but you know they're doing 349 00:17:17,080 --> 00:17:20,080 Speaker 3: a lot of on field drills that are also testing 350 00:17:20,600 --> 00:17:23,520 Speaker 3: that change of direction ability, so like a pass pro 351 00:17:23,600 --> 00:17:26,159 Speaker 3: mirror drill you'll see during the combine. Like, so we 352 00:17:26,400 --> 00:17:30,520 Speaker 3: are collecting tracking data during all those drills. So you know, 353 00:17:30,680 --> 00:17:33,040 Speaker 3: it hasn't been like the highest priority, but I'm sure 354 00:17:33,080 --> 00:17:36,240 Speaker 3: teams are using that tracking data to try and get 355 00:17:36,280 --> 00:17:40,280 Speaker 3: better athletic measurables. For our perspective, you know, we we 356 00:17:40,359 --> 00:17:44,440 Speaker 3: want to try and capture those to get those better 357 00:17:44,560 --> 00:17:49,360 Speaker 3: estimates of the stop start, the changing direction skills. It's 358 00:17:49,400 --> 00:17:51,680 Speaker 3: just a matter of resources and where we put our 359 00:17:52,440 --> 00:17:55,160 Speaker 3: you know, own efforts any given season. 360 00:17:55,480 --> 00:17:57,399 Speaker 1: Yeah, and look, and there was one piece of this 361 00:17:57,680 --> 00:18:01,160 Speaker 1: that type of data that's up on com slash iq 362 00:18:01,359 --> 00:18:05,440 Speaker 1: right now, and you give the highest miles per hour 363 00:18:05,680 --> 00:18:10,560 Speaker 1: player reaches on their forty Why to you is that important? 364 00:18:10,680 --> 00:18:14,080 Speaker 1: And how what can fans kind of gleam from that data? 365 00:18:14,160 --> 00:18:16,840 Speaker 1: Maybe in combination with the ten yards split, because I 366 00:18:16,840 --> 00:18:18,679 Speaker 1: think you're kind of seeing two different things with that 367 00:18:18,800 --> 00:18:21,159 Speaker 1: together where you can kind of give a little bit 368 00:18:21,200 --> 00:18:23,520 Speaker 1: more nuance in detail to a forty yard dash time. 369 00:18:24,200 --> 00:18:27,960 Speaker 3: Yeah, I think, as we said before, like the value 370 00:18:28,000 --> 00:18:30,040 Speaker 3: of the timing of the forty is just it's been 371 00:18:30,080 --> 00:18:33,520 Speaker 3: a standard measurement for twenty three years, but doesn't tell 372 00:18:33,520 --> 00:18:35,560 Speaker 3: you how they reach that speed, doesn't tell you anything 373 00:18:35,560 --> 00:18:39,040 Speaker 3: about their acceleration. So we actually have speed curves and 374 00:18:39,080 --> 00:18:41,320 Speaker 3: they'll be launching on our sites sometimes soon in terms 375 00:18:41,320 --> 00:18:44,760 Speaker 3: of like what your speed is at every ten yards, 376 00:18:44,960 --> 00:18:46,920 Speaker 3: So how fast did you reach your top speed in 377 00:18:46,960 --> 00:18:50,000 Speaker 3: the forty? The better measure acceleration and like time to 378 00:18:50,000 --> 00:18:52,879 Speaker 3: top speed rather than you know, you just run in 379 00:18:52,920 --> 00:18:54,639 Speaker 3: a straight line and how fast do you can you 380 00:18:54,720 --> 00:18:57,400 Speaker 3: run in that straight line? It's a little bit more translatable. 381 00:18:58,160 --> 00:19:00,240 Speaker 3: One fun finding that we have over the last four 382 00:19:00,359 --> 00:19:03,360 Speaker 3: years of this data is that you know, Xavier Worthy 383 00:19:03,480 --> 00:19:07,399 Speaker 3: might have ran the fastest forty by a timing perspective, 384 00:19:07,400 --> 00:19:11,040 Speaker 3: but Devon Anchen reached a faster top speed on his 385 00:19:11,160 --> 00:19:15,320 Speaker 3: forty than Xavier Worthy. And you know, I might be 386 00:19:15,400 --> 00:19:17,560 Speaker 3: cherry picking a little bit there for two names that 387 00:19:17,600 --> 00:19:21,000 Speaker 3: people know, but you know, I think there's some signal 388 00:19:21,000 --> 00:19:23,000 Speaker 3: in there. We got to figure out what that might be. 389 00:19:23,520 --> 00:19:26,639 Speaker 3: And also I think one fun thing to do I 390 00:19:26,640 --> 00:19:30,399 Speaker 3: know SUMER does, but basically use the tracking data at 391 00:19:30,400 --> 00:19:33,679 Speaker 3: the NFL level to see how players maybe develop in 392 00:19:33,760 --> 00:19:36,119 Speaker 3: the league. Like they One of their big things is 393 00:19:36,119 --> 00:19:40,440 Speaker 3: that JSN has gotten actually faster in the NFL than 394 00:19:40,480 --> 00:19:43,760 Speaker 3: he was at the Combine, which I think is a 395 00:19:43,800 --> 00:19:46,159 Speaker 3: super interesting thing. And you know, these players are not 396 00:19:46,200 --> 00:19:49,399 Speaker 3: done developing obviously when they enter the league, and you know, 397 00:19:49,640 --> 00:19:53,080 Speaker 3: whether that's their personal training or you know, the sports 398 00:19:53,119 --> 00:19:55,240 Speaker 3: science staff at a team, or just you know, the 399 00:19:55,280 --> 00:19:59,160 Speaker 3: coaching in general. Like, I think that's a very underappreciated 400 00:19:59,240 --> 00:20:02,760 Speaker 3: part of what actually makes prospects successful or not is 401 00:20:02,800 --> 00:20:06,560 Speaker 3: the environment and how they're continued developed at the league level. 402 00:20:07,240 --> 00:20:09,200 Speaker 1: You guys mentioned as you build this out, more feature 403 00:20:09,200 --> 00:20:11,120 Speaker 1: will be available depending on what you want to focus on. 404 00:20:12,000 --> 00:20:13,840 Speaker 1: Of the drills in the positions, What are some of 405 00:20:13,880 --> 00:20:16,080 Speaker 1: the ones you're excited to maybe get a little bit 406 00:20:16,080 --> 00:20:18,240 Speaker 1: deeper into this with with some of the player tracking 407 00:20:18,280 --> 00:20:21,400 Speaker 1: the mphs, the miles per hour, how fast they're going 408 00:20:21,400 --> 00:20:22,879 Speaker 1: that sort of change the direction. What are some of 409 00:20:22,920 --> 00:20:24,800 Speaker 1: the drills and some of the things you guys are 410 00:20:24,800 --> 00:20:26,440 Speaker 1: really kind of even if you're not going to get 411 00:20:26,440 --> 00:20:27,720 Speaker 1: to it this year that you're like, you know what, 412 00:20:28,000 --> 00:20:30,560 Speaker 1: that actually could give some pretty good detail of nuance 413 00:20:30,560 --> 00:20:32,359 Speaker 1: and I think it could be helpful to teams. 414 00:20:32,840 --> 00:20:35,440 Speaker 3: Yeah, I'll start with what we have found so far 415 00:20:35,520 --> 00:20:37,800 Speaker 3: that I think was fun and again like we haven't 416 00:20:37,960 --> 00:20:40,560 Speaker 3: done a comprehensive analysis where I can tell you that 417 00:20:40,600 --> 00:20:43,880 Speaker 3: the gauntlet drill really matters to NFL success. But there 418 00:20:43,880 --> 00:20:46,399 Speaker 3: are some fun players that have done well at the 419 00:20:46,800 --> 00:20:50,040 Speaker 3: gauntlet drill if you compare it to their just regular 420 00:20:50,080 --> 00:20:52,199 Speaker 3: forty yard dash, because that's game speed. You have to 421 00:20:52,200 --> 00:20:56,080 Speaker 3: be focusing on catching balls as you're running. And you know, 422 00:20:56,600 --> 00:20:58,760 Speaker 3: four years ago we saw Puka and Akua have run 423 00:20:58,800 --> 00:21:03,840 Speaker 3: the slowest forty but reach the fastest speed in his gauntlet, 424 00:21:03,960 --> 00:21:05,880 Speaker 3: And we've always kind of been like, huh, there might 425 00:21:05,880 --> 00:21:08,480 Speaker 3: be something here in terms of like we might be better 426 00:21:08,440 --> 00:21:12,320 Speaker 3: in measuring game speed better during the gauntlet than we 427 00:21:12,400 --> 00:21:14,879 Speaker 3: are actually during the forty yard dash, which is more 428 00:21:14,880 --> 00:21:17,040 Speaker 3: of a track runner. You can train for it, and 429 00:21:17,040 --> 00:21:20,960 Speaker 3: everything you're reacting live to how the balls are. I 430 00:21:20,960 --> 00:21:23,480 Speaker 3: think a lot of the defensive back drills too, just 431 00:21:23,520 --> 00:21:26,600 Speaker 3: in terms of determining when players are stopping and starting 432 00:21:26,600 --> 00:21:28,200 Speaker 3: and how fast they're breaking on the ball. 433 00:21:28,960 --> 00:21:30,840 Speaker 1: They're flipping their hips too, right, like how fast they 434 00:21:30,880 --> 00:21:32,080 Speaker 1: can flip their hips. 435 00:21:32,359 --> 00:21:36,000 Speaker 3: Yeah, Unfortunately, we're not going to capture the flipping of 436 00:21:36,040 --> 00:21:38,359 Speaker 3: the hips just through you know, chips in the shoulder 437 00:21:38,400 --> 00:21:42,200 Speaker 3: pads at least maintaining that speed. If we can identify 438 00:21:42,280 --> 00:21:44,399 Speaker 3: when they're flipping their hips, you know, that could be 439 00:21:44,400 --> 00:21:45,520 Speaker 3: a good proxy as well. 440 00:21:46,240 --> 00:21:47,919 Speaker 1: Yeah, and I got to imagine too, you know, you 441 00:21:47,960 --> 00:21:50,200 Speaker 1: talk about the Hawkey stuff in game. Maybe eventually you 442 00:21:50,200 --> 00:21:52,000 Speaker 1: guys will put that into the combines. You can do 443 00:21:52,040 --> 00:21:55,440 Speaker 1: some of the three D stuff. I imagine at some point 444 00:21:55,520 --> 00:21:58,040 Speaker 1: you might even be able to measure like bend for 445 00:21:58,160 --> 00:22:00,800 Speaker 1: edge rushers, right, like ankle, fletch and stuff like that. 446 00:22:01,520 --> 00:22:04,200 Speaker 3: Yeah, that's that's where we're trying to go. We've done 447 00:22:04,200 --> 00:22:09,600 Speaker 3: some initial studies with maybe some QB just overall throwing mechanics. 448 00:22:09,920 --> 00:22:14,560 Speaker 3: That has been very interesting. But in general, like I 449 00:22:14,600 --> 00:22:17,800 Speaker 3: think the worlds really are oyster in terms of where 450 00:22:17,840 --> 00:22:21,640 Speaker 3: we go with this data. Whether it's you know, actually 451 00:22:21,800 --> 00:22:24,040 Speaker 3: being able to tell whether players are engaged in the 452 00:22:24,080 --> 00:22:26,320 Speaker 3: block and not just guess it based on their proximity 453 00:22:26,359 --> 00:22:29,359 Speaker 3: to one another, to getting a catch radius. You know, 454 00:22:29,400 --> 00:22:32,080 Speaker 3: people always ask us when it comes to completion probability, 455 00:22:32,160 --> 00:22:35,600 Speaker 3: like what was the completion probability on Odell's famous catch one? 456 00:22:35,640 --> 00:22:37,600 Speaker 3: We didn't have tracking data back then, but too, even 457 00:22:37,640 --> 00:22:40,159 Speaker 3: if we did, it wouldn't be an appropriate use of 458 00:22:40,200 --> 00:22:44,080 Speaker 3: our model because we didn't know that he's horizontal fully 459 00:22:44,200 --> 00:22:46,679 Speaker 3: arm stretched out, like, we just know where he is 460 00:22:46,720 --> 00:22:50,480 Speaker 3: on the field. So like getting from completion probability to 461 00:22:50,520 --> 00:22:54,800 Speaker 3: catch probability, I think we'll clarify a lot of you know, 462 00:22:55,000 --> 00:22:58,240 Speaker 3: where our model fails right now, and maybe for some 463 00:22:58,280 --> 00:23:01,159 Speaker 3: of the TV execs that are that a one handed 464 00:23:01,160 --> 00:23:03,359 Speaker 3: catch isn't as low as they want it to be. 465 00:23:04,520 --> 00:23:06,520 Speaker 3: You know, it'll help us out there as well. 466 00:23:08,000 --> 00:23:10,439 Speaker 2: There's only about TV man, I get it, I get it. 467 00:23:11,720 --> 00:23:12,440 Speaker 2: Production score. 468 00:23:12,480 --> 00:23:14,280 Speaker 1: You guys talked about how you kind of do that 469 00:23:14,320 --> 00:23:16,520 Speaker 1: regression to come up with the production score. Do you 470 00:23:16,600 --> 00:23:19,960 Speaker 1: find that there's a better correlation for success for the 471 00:23:20,000 --> 00:23:24,480 Speaker 1: athleticism score or the production score or is that why 472 00:23:24,520 --> 00:23:26,840 Speaker 1: you put them together to kind of create that overall metric. 473 00:23:27,359 --> 00:23:30,560 Speaker 3: Yeah, we haven't done too many you know, back testing 474 00:23:30,600 --> 00:23:33,600 Speaker 3: in terms of which of these components is more important 475 00:23:33,640 --> 00:23:36,960 Speaker 3: towards NFL success. I think just being able to view 476 00:23:37,000 --> 00:23:40,679 Speaker 3: both can really give you just an overall profile. Like 477 00:23:40,720 --> 00:23:44,120 Speaker 3: there are the guys that are just great college players 478 00:23:44,160 --> 00:23:47,560 Speaker 3: that might not be the best athletes, and you can 479 00:23:47,640 --> 00:23:50,359 Speaker 3: see that in their profile. Whereas there's the combine workout 480 00:23:50,359 --> 00:23:53,239 Speaker 3: Warriors that didn't do anything in college, but you know, 481 00:23:54,320 --> 00:23:58,679 Speaker 3: you're taking a bet on that player based on their traits, 482 00:23:59,080 --> 00:24:01,080 Speaker 3: and I think just being able to see that is 483 00:24:01,080 --> 00:24:05,000 Speaker 3: a good, good overall archetype for fans to be able 484 00:24:05,040 --> 00:24:07,520 Speaker 3: to kind of evaluate players in their development curves and 485 00:24:07,520 --> 00:24:10,800 Speaker 3: what you should expect out of them coming into the league. Hypothetically, 486 00:24:10,800 --> 00:24:14,280 Speaker 3: if someone's maxed out from athleticism standpoint and has a 487 00:24:14,280 --> 00:24:17,040 Speaker 3: low athleticism score but their production was super high, they're 488 00:24:17,080 --> 00:24:19,080 Speaker 3: coming into the league a lot more polished and you 489 00:24:19,080 --> 00:24:22,320 Speaker 3: would expect them to contribute more immediately. Whereas a guy 490 00:24:22,359 --> 00:24:24,959 Speaker 3: with a higher athleticism score, lower production score, it might 491 00:24:24,960 --> 00:24:26,439 Speaker 3: take him a few years to get there, but his 492 00:24:26,520 --> 00:24:31,240 Speaker 3: potential is kind of what we're inferring. His potential with 493 00:24:31,320 --> 00:24:34,040 Speaker 3: the right development could be very high. 494 00:24:34,440 --> 00:24:37,399 Speaker 2: How do you do production scores for offensive linemen? 495 00:24:38,920 --> 00:24:42,959 Speaker 3: So production scores for offensive linemen, we do have some 496 00:24:43,240 --> 00:24:48,439 Speaker 3: charting data from college that goes back, you know, about 497 00:24:48,440 --> 00:24:52,359 Speaker 3: a decade before then. It's really just kind of a 498 00:24:52,400 --> 00:24:55,959 Speaker 3: consensus grade. There's nothing really too much to it because 499 00:24:56,080 --> 00:24:59,160 Speaker 3: we don't have any stats for offensive lineman other than 500 00:24:59,240 --> 00:25:03,400 Speaker 3: games started and games played, you know, back in twenty 501 00:25:03,480 --> 00:25:08,880 Speaker 3: ten or whatever. But yeah, you know, offensive line in general, 502 00:25:09,000 --> 00:25:11,720 Speaker 3: I think is somewhere that we're still continuing, even at 503 00:25:11,720 --> 00:25:15,439 Speaker 3: the league level, to try and build out more stats 504 00:25:15,440 --> 00:25:18,119 Speaker 3: to highlight the big guys. You know, this off season 505 00:25:18,640 --> 00:25:21,760 Speaker 3: we are focusing a little bit on We're focusing on 506 00:25:21,880 --> 00:25:24,720 Speaker 3: classifying run blocking matchups so we can start to get, 507 00:25:24,720 --> 00:25:26,760 Speaker 3: you know, who are the best run blockers in the league. 508 00:25:26,760 --> 00:25:30,439 Speaker 3: From a tracking perspective, how much displacement do you for us, Like, 509 00:25:30,560 --> 00:25:36,159 Speaker 3: are you the intended like lead blocker and a specific gap? 510 00:25:36,600 --> 00:25:38,240 Speaker 3: How much does your team trust you to do that? 511 00:25:38,960 --> 00:25:41,680 Speaker 3: Similar to how you know, from a pass blocking perspective, 512 00:25:41,680 --> 00:25:43,960 Speaker 3: we could tell you how often you get ship block help, 513 00:25:43,960 --> 00:25:46,680 Speaker 3: how often you're left one on one, and then what 514 00:25:46,720 --> 00:25:50,480 Speaker 3: pressure rates do you give up there? So that's kind 515 00:25:50,520 --> 00:25:53,960 Speaker 3: of one of the last frontiers is run blocking matchups 516 00:25:54,000 --> 00:25:56,400 Speaker 3: and run schemes, and that's kind of what we're focusing 517 00:25:56,480 --> 00:25:59,520 Speaker 3: this offseason from a stat development perspective to kind of 518 00:25:59,600 --> 00:26:00,639 Speaker 3: round out our toolbox. 519 00:26:00,760 --> 00:26:03,240 Speaker 2: Yeah, I remember we did two years ago. 520 00:26:03,280 --> 00:26:05,240 Speaker 1: Now, we did like a forty five minute podcast with 521 00:26:05,440 --> 00:26:09,040 Speaker 1: Cynthia Freeland or did her who did her doctorate on 522 00:26:09,640 --> 00:26:13,080 Speaker 1: trying to predict NFL offensive Lineman success from the combine, 523 00:26:13,080 --> 00:26:15,520 Speaker 1: and she told us about how, yeah, it's like this 524 00:26:15,600 --> 00:26:18,119 Speaker 1: little area and you're just trying to measure how in 525 00:26:18,160 --> 00:26:21,520 Speaker 1: the first ten yards, how quickly like the offensive Lieman 526 00:26:21,560 --> 00:26:24,040 Speaker 1: becomes like more upright and vertical and you're just trying 527 00:26:24,040 --> 00:26:27,520 Speaker 1: to with lasers measure the amount at distance. And I'm like, man, 528 00:26:27,600 --> 00:26:32,040 Speaker 1: like this is this is some crazy stuff. Yeah, and 529 00:26:32,080 --> 00:26:34,680 Speaker 1: it's just it's again with that without the production just 530 00:26:34,680 --> 00:26:35,280 Speaker 1: it gets. 531 00:26:35,119 --> 00:26:37,160 Speaker 2: Very very very very very tough. 532 00:26:37,960 --> 00:26:41,600 Speaker 1: Yeah, So any surprises that when you went through all 533 00:26:41,600 --> 00:26:43,879 Speaker 1: the data at the combine, you're not a scout just 534 00:26:43,920 --> 00:26:46,399 Speaker 1: from the data where maybe guys that based on the 535 00:26:46,400 --> 00:26:49,199 Speaker 1: consensus bores and the grind and the mockx guys that 536 00:26:49,240 --> 00:26:52,679 Speaker 1: you saw based on your consensus score overall score with 537 00:26:52,720 --> 00:26:56,040 Speaker 1: the athleticism and the production together maybe would surprise some 538 00:26:56,080 --> 00:26:58,520 Speaker 1: people based on how they're talked about in the draft community. 539 00:26:59,359 --> 00:27:04,720 Speaker 3: Yeah, surprises wise, like Cornel Tait of like the top 540 00:27:04,800 --> 00:27:09,560 Speaker 3: ten isn't necessarily in that eighty above level right now. Now, 541 00:27:09,600 --> 00:27:12,560 Speaker 3: he didn't do that much at the combines, so it'll 542 00:27:12,600 --> 00:27:14,879 Speaker 3: be interesting to see if he, you know, at his 543 00:27:14,960 --> 00:27:16,760 Speaker 3: pro day kind of fills in a few of those 544 00:27:17,119 --> 00:27:20,760 Speaker 3: other drills, because it is the athleticism score that's dragging 545 00:27:20,880 --> 00:27:23,440 Speaker 3: him down, but it's really just based on the four 546 00:27:23,520 --> 00:27:27,280 Speaker 3: or five three and like maybe he's winning in short 547 00:27:27,320 --> 00:27:30,320 Speaker 3: area agility or the jumps are better than you know, 548 00:27:30,359 --> 00:27:32,919 Speaker 3: we're inferring from that. So I hate to put too 549 00:27:33,040 --> 00:27:36,520 Speaker 3: much into that right now with the incomplete information. Two 550 00:27:36,560 --> 00:27:39,879 Speaker 3: guys that'll highlight just in terms of like being historically 551 00:27:40,400 --> 00:27:44,800 Speaker 3: high players for their position, especially one of them is 552 00:27:44,880 --> 00:27:48,159 Speaker 3: Sunny Styles, who you know, Giants are very heavily linked to. 553 00:27:48,320 --> 00:27:54,520 Speaker 3: Right now, he's sandwiched in between Luke Keickley and Patrick Willis. 554 00:27:54,560 --> 00:27:57,280 Speaker 3: In our overall draft score going back to two thousand 555 00:27:57,280 --> 00:28:00,000 Speaker 3: and three, ranks the top five of off ball lineback 556 00:28:00,359 --> 00:28:02,840 Speaker 3: in that period. And I think that it's fun to 557 00:28:02,880 --> 00:28:05,960 Speaker 3: see an off ball linebacker talked about as a top 558 00:28:06,000 --> 00:28:11,000 Speaker 3: ten lottery pick because you know, Gonner the days, you know, 559 00:28:11,359 --> 00:28:14,480 Speaker 3: back when they were going top ten consistently. I believe 560 00:28:14,480 --> 00:28:17,160 Speaker 3: there's only been three top ten picks since twenty fifteen 561 00:28:17,800 --> 00:28:20,680 Speaker 3: at off ball linebacker, but it seems like he might 562 00:28:20,680 --> 00:28:23,280 Speaker 3: be a lock this year. And you know, he's in 563 00:28:23,320 --> 00:28:27,360 Speaker 3: some very high company in terms of his overall athletic 564 00:28:27,600 --> 00:28:32,480 Speaker 3: and production profile with a ninety in both. Just a 565 00:28:32,480 --> 00:28:37,160 Speaker 3: well rounded player converted safety. And then the other guy 566 00:28:37,160 --> 00:28:39,680 Speaker 3: who stands out is kind of historically good in this 567 00:28:39,800 --> 00:28:45,120 Speaker 3: year's draft class other than Jeremiah Love is Kenyon Sadik, 568 00:28:45,320 --> 00:28:48,640 Speaker 3: who's been rising up and down draft board, rising up 569 00:28:48,720 --> 00:28:52,280 Speaker 3: draft boards since his you know, very very strong Combine performance. 570 00:28:53,160 --> 00:28:56,280 Speaker 3: We have him as a top five tight end prospect 571 00:28:56,320 --> 00:29:00,440 Speaker 3: of the last twenty years as well, and it's he's 572 00:29:00,480 --> 00:29:03,920 Speaker 3: one of the guys who has seen a sharp rise 573 00:29:04,000 --> 00:29:06,840 Speaker 3: in his grinding the MOCKX just overall gone from like 574 00:29:07,040 --> 00:29:11,040 Speaker 3: mid twenties to about twelve is his range now in 575 00:29:11,080 --> 00:29:14,760 Speaker 3: the last three weeks since the conclusion of the combine. 576 00:29:14,960 --> 00:29:16,600 Speaker 2: If you want to know how to manage two minutes 577 00:29:16,640 --> 00:29:18,840 Speaker 2: of crunch time football, I'm your man. But if you're 578 00:29:18,840 --> 00:29:21,640 Speaker 2: wondering about a long term financial plan, you should talk 579 00:29:21,680 --> 00:29:24,320 Speaker 2: to Citizens. Hey, I can also talk long care I'd 580 00:29:24,360 --> 00:29:26,680 Speaker 2: like to learn about a money routine. Yes, I knew 581 00:29:26,720 --> 00:29:29,880 Speaker 2: I could help makes sense of your money with Citizens. 582 00:29:30,320 --> 00:29:33,600 Speaker 1: And then fans are familiar with with Trey Lance obviously, 583 00:29:33,600 --> 00:29:36,760 Speaker 1: but his brother also blew things up at the Combine. 584 00:29:36,800 --> 00:29:38,240 Speaker 2: His production score is pretty good too. 585 00:29:38,600 --> 00:29:42,840 Speaker 1: Yeah, yeah, I think he's somebody that you know, fans 586 00:29:42,840 --> 00:29:44,600 Speaker 1: probably haven't paid much attention to. You got to go 587 00:29:44,640 --> 00:29:47,320 Speaker 1: find that tape and give him a look and see 588 00:29:47,360 --> 00:29:49,760 Speaker 1: what you think of him on tape. All right, real quick, 589 00:29:49,800 --> 00:29:51,959 Speaker 1: before we get to next Gen status, to remind our 590 00:29:52,000 --> 00:29:53,840 Speaker 1: fans that GIHNS Little Podcast is brought to you by 591 00:29:53,880 --> 00:29:56,680 Speaker 1: Citizens Official Bank of the Giants from Game Day celebrations, 592 00:29:56,680 --> 00:29:58,680 Speaker 1: the everyday financial needs Big WU fans get the most 593 00:29:58,720 --> 00:30:00,960 Speaker 1: out of every moment with Citizens or more Citizens Bank 594 00:30:01,000 --> 00:30:03,880 Speaker 1: dot Com slash Giants. All right, I want to jump 595 00:30:03,920 --> 00:30:06,360 Speaker 1: to the next gen stats here, and can you just 596 00:30:06,400 --> 00:30:08,480 Speaker 1: explain the fans before we kind of get into. 597 00:30:08,360 --> 00:30:09,680 Speaker 2: A couple of the detailed stats. 598 00:30:10,000 --> 00:30:13,680 Speaker 1: How you use the tracking and if people that use 599 00:30:13,720 --> 00:30:17,400 Speaker 1: it the moving dots on the screen to come up 600 00:30:17,440 --> 00:30:19,200 Speaker 1: with some of these numbers in the stats that you 601 00:30:19,200 --> 00:30:20,240 Speaker 1: guys do for next Gen. 602 00:30:21,000 --> 00:30:25,040 Speaker 3: Yeah, So, just to set the stage, every player is 603 00:30:25,040 --> 00:30:27,000 Speaker 3: wearing two chips in their shoulder pads and we get 604 00:30:27,000 --> 00:30:33,480 Speaker 3: their location and speed ten times per second. There's an 605 00:30:33,480 --> 00:30:39,040 Speaker 3: eventing team that is run by the Zebra Technologies groups 606 00:30:39,080 --> 00:30:42,040 Speaker 3: the ones who's collected this data, and a QA team 607 00:30:42,120 --> 00:30:44,760 Speaker 3: that goes through and gets those events to bring context 608 00:30:44,760 --> 00:30:47,080 Speaker 3: to that tracking data. So you can have just a 609 00:30:47,080 --> 00:30:48,920 Speaker 3: bunch of players running around, but if you don't know 610 00:30:48,920 --> 00:30:50,920 Speaker 3: when the ball is snapped and when the ball is 611 00:30:50,960 --> 00:30:53,760 Speaker 3: thrown and when a player is tackled, then you know, 612 00:30:53,840 --> 00:30:56,680 Speaker 3: it's really hard to infer anything from that. So really 613 00:30:57,040 --> 00:31:01,000 Speaker 3: the eventing is the first key piece of context that 614 00:31:01,160 --> 00:31:03,440 Speaker 3: allows us to build everything else and triggers all of 615 00:31:03,480 --> 00:31:06,920 Speaker 3: our models and algorithms in order to build those stats. 616 00:31:07,200 --> 00:31:09,680 Speaker 3: A lot of our stats originally were built with logic, 617 00:31:09,840 --> 00:31:12,440 Speaker 3: you know, just distance based how far away was this 618 00:31:12,480 --> 00:31:16,400 Speaker 3: defender when the pass arrived, how far away was the 619 00:31:16,440 --> 00:31:19,120 Speaker 3: nearest pass rusher when the quarterback through the ball. But 620 00:31:19,240 --> 00:31:23,560 Speaker 3: over the years we found that in terms of like 621 00:31:24,440 --> 00:31:27,800 Speaker 3: classifying the complex the complexity of the game, it's a 622 00:31:27,800 --> 00:31:32,200 Speaker 3: lot better done with machine learning models, so specifically transformer models. 623 00:31:32,200 --> 00:31:33,880 Speaker 3: They're able to pick up a lot more on these 624 00:31:33,920 --> 00:31:37,520 Speaker 3: patterns of movement, so we can tell you now, you know, 625 00:31:37,840 --> 00:31:40,920 Speaker 3: based on usually training models, based off of our own 626 00:31:41,000 --> 00:31:46,000 Speaker 3: labels that are hand charted, the machine starts to pick 627 00:31:46,080 --> 00:31:49,160 Speaker 3: up these kind of overall movement patterns and everything, so 628 00:31:49,200 --> 00:31:52,520 Speaker 3: we can tell you in near real time now whether 629 00:31:52,720 --> 00:31:55,680 Speaker 3: a defense is running a cover two or cover six scheme, 630 00:31:56,120 --> 00:32:00,280 Speaker 3: whether the you know corner the field corner is the 631 00:32:00,280 --> 00:32:03,320 Speaker 3: curl flat or he's you know, playing deep outer third. 632 00:32:03,960 --> 00:32:05,840 Speaker 3: Starting to build up these kind of player roles one 633 00:32:05,880 --> 00:32:08,400 Speaker 3: by oneting team schemes, and then you start to just 634 00:32:08,520 --> 00:32:10,760 Speaker 3: builter the data every which way. You know, how does 635 00:32:10,760 --> 00:32:13,040 Speaker 3: a quarterback do against the blitz? How does the receiver 636 00:32:13,120 --> 00:32:16,240 Speaker 3: do against pressman coverage? And all the we I think 637 00:32:16,280 --> 00:32:18,800 Speaker 3: have over five hundred stats that are calculating in near 638 00:32:18,840 --> 00:32:23,800 Speaker 3: real time once the plays evented and validated, which takes 639 00:32:23,840 --> 00:32:27,840 Speaker 3: about a minute after a play is completed. And that's 640 00:32:27,960 --> 00:32:31,280 Speaker 3: kind of when we're working with that data to send 641 00:32:31,280 --> 00:32:34,520 Speaker 3: it out across all of our broadcast partners, you know, 642 00:32:35,000 --> 00:32:38,360 Speaker 3: supporting each broadcast by sending over kind of that research 643 00:32:38,440 --> 00:32:39,120 Speaker 3: as it comes in. 644 00:32:40,120 --> 00:32:41,080 Speaker 2: Yeah, that makes sense. 645 00:32:41,240 --> 00:32:43,200 Speaker 1: I was one thing for pressures because I remember there 646 00:32:43,240 --> 00:32:45,120 Speaker 1: was a game I might have been two years ago 647 00:32:45,800 --> 00:32:48,920 Speaker 1: where a guy had like two quarterback hits in the 648 00:32:48,960 --> 00:32:52,280 Speaker 1: game on the official sheet, but they had no pressures, right, 649 00:32:52,720 --> 00:32:57,320 Speaker 1: And I know, and I'm just wondering, is that is 650 00:32:57,320 --> 00:32:59,480 Speaker 1: that based on how far he was away when the 651 00:32:59,520 --> 00:33:02,160 Speaker 1: throw was made? And then maybe that's why I didn't register. 652 00:33:02,440 --> 00:33:04,960 Speaker 1: How do you quantify whether or not there's an offensive 653 00:33:05,000 --> 00:33:07,240 Speaker 1: lineman between the rusher and the quarterback and whether or 654 00:33:07,280 --> 00:33:10,560 Speaker 1: he's actually pressuring How do you figure out the pressure component, 655 00:33:10,560 --> 00:33:13,120 Speaker 1: because I do think that's something that can be very 656 00:33:13,240 --> 00:33:15,440 Speaker 1: much even for the people that do this, you know, 657 00:33:15,560 --> 00:33:18,360 Speaker 1: hand do every game from one person to the next, 658 00:33:18,600 --> 00:33:20,840 Speaker 1: so soone's opinion on whether or not it's a pressure 659 00:33:20,960 --> 00:33:22,640 Speaker 1: can be very very different. 660 00:33:22,760 --> 00:33:25,320 Speaker 3: Right right, Actually, I just had a very interesting conversation 661 00:33:25,760 --> 00:33:28,880 Speaker 3: about this with someone from one of those other firms 662 00:33:28,920 --> 00:33:31,120 Speaker 3: that has their own pressure model. But it is a 663 00:33:31,160 --> 00:33:34,840 Speaker 3: model that basically for every time slice every tenth of 664 00:33:34,880 --> 00:33:38,880 Speaker 3: a second, predicts whether a player is generating pressure or 665 00:33:38,960 --> 00:33:42,040 Speaker 3: not based on play level labels. So we take a 666 00:33:42,160 --> 00:33:46,800 Speaker 3: set of from twenty eighteen to twenty twenty three when 667 00:33:46,800 --> 00:33:51,440 Speaker 3: we train the model, you know, labels for each pass 668 00:33:51,520 --> 00:33:53,880 Speaker 3: rusher whether they got pressure or not, and then use 669 00:33:53,960 --> 00:33:57,200 Speaker 3: the tracking data to try and infer whether they affected 670 00:33:57,200 --> 00:34:00,160 Speaker 3: the quarterback before he threw or forced the quarterback off 671 00:34:00,160 --> 00:34:03,760 Speaker 3: the spot. And we do take into account features such 672 00:34:03,840 --> 00:34:07,400 Speaker 3: as a blocker interference, so how many blockers are in 673 00:34:07,480 --> 00:34:11,239 Speaker 3: between the quarterback and the pass rusher. They're overall kind 674 00:34:11,280 --> 00:34:15,320 Speaker 3: of speed vectors towards the quarterback. Are they closing the gap? 675 00:34:15,640 --> 00:34:19,520 Speaker 3: Is the quarterback now moving based on this incoming pass rusher? 676 00:34:19,920 --> 00:34:22,600 Speaker 3: So we you know, I think four years ago and beyond, 677 00:34:22,600 --> 00:34:24,759 Speaker 3: it was just did the pass rusher get within two 678 00:34:24,760 --> 00:34:27,440 Speaker 3: and a half yards or two yards of the quarterback 679 00:34:27,520 --> 00:34:30,880 Speaker 3: at any point during the dropback. We've evolved to a 680 00:34:30,960 --> 00:34:34,520 Speaker 3: much more comprehensive one that really takes into account the 681 00:34:34,600 --> 00:34:38,640 Speaker 3: quarterback's movement and reaction to pressure more than anything that 682 00:34:38,680 --> 00:34:41,759 Speaker 3: we were capturing before. And you know, there are certainly 683 00:34:42,200 --> 00:34:46,240 Speaker 3: going to be some disagreements between us and a certain 684 00:34:46,320 --> 00:34:48,680 Speaker 3: charting servers or a different model, and that's stuff you 685 00:34:48,719 --> 00:34:50,839 Speaker 3: need to live with. But we do like the fact 686 00:34:50,840 --> 00:34:53,560 Speaker 3: that it is objective and it's not relying on a 687 00:34:53,560 --> 00:34:56,400 Speaker 3: hand charting, and it's also immediate more than anything. You know, 688 00:34:56,440 --> 00:34:58,839 Speaker 3: we get to have that in real time and feed 689 00:34:58,880 --> 00:34:59,919 Speaker 3: that to the broadcasters. 690 00:35:00,520 --> 00:35:02,880 Speaker 1: Yeah, Nook, I think that makes a lot of sense. 691 00:35:03,640 --> 00:35:05,759 Speaker 1: How about separation this is I know you guys have 692 00:35:05,760 --> 00:35:07,880 Speaker 1: have done a lot more with r I'm measuring corners 693 00:35:07,880 --> 00:35:10,640 Speaker 1: and wide receivers and how open wide receivers are. How 694 00:35:10,680 --> 00:35:13,359 Speaker 1: has that process kind of moved along as you try 695 00:35:13,360 --> 00:35:16,200 Speaker 1: to figure out how well wide receivers can create separation, which, 696 00:35:16,320 --> 00:35:18,200 Speaker 1: let's be honest, getting open and then catching the ball 697 00:35:18,280 --> 00:35:20,480 Speaker 1: is really the most important skill a wide receiver. 698 00:35:20,280 --> 00:35:22,560 Speaker 3: Can have, right Yeah, And I still don't think we 699 00:35:22,600 --> 00:35:24,640 Speaker 3: do a great job of that, to be honest. I 700 00:35:24,680 --> 00:35:28,399 Speaker 3: think the separation stats that you'll see on the next 701 00:35:28,400 --> 00:35:32,800 Speaker 3: Gen Stats public website are usually just how much separation 702 00:35:32,960 --> 00:35:35,120 Speaker 3: did you have when the pass was thrown, but it 703 00:35:35,120 --> 00:35:38,000 Speaker 3: has no context for where the pass was thrown Underneath. 704 00:35:38,000 --> 00:35:40,480 Speaker 3: Passes are just generally going to have more separation. Was 705 00:35:40,480 --> 00:35:43,600 Speaker 3: it against man or was it against zone? And how 706 00:35:43,640 --> 00:35:47,960 Speaker 3: did the receiver specifically create that separation throughout the route? 707 00:35:47,960 --> 00:35:51,360 Speaker 3: What would you expect it based on this route versus 708 00:35:51,400 --> 00:35:56,520 Speaker 3: coverage versus leverage overall. I do think that, you know, 709 00:35:56,600 --> 00:35:59,319 Speaker 3: we rarely kind of use that stat on its own. 710 00:35:59,360 --> 00:36:03,319 Speaker 3: It needs con text, so that's something that we tell 711 00:36:03,320 --> 00:36:06,200 Speaker 3: our analysts to do. But one of the cool things 712 00:36:06,200 --> 00:36:08,520 Speaker 3: that came out of our stat development project last year 713 00:36:09,120 --> 00:36:12,160 Speaker 3: was we now have coverage matchups on every play, so 714 00:36:12,239 --> 00:36:16,160 Speaker 3: who is guarding who, and from that we can tell 715 00:36:16,160 --> 00:36:19,640 Speaker 3: you what the average separation is throughout the route for 716 00:36:19,719 --> 00:36:22,840 Speaker 3: each matchup and also how much timer is a cornerback 717 00:36:22,840 --> 00:36:25,520 Speaker 3: spending in tight coverage, which we define is less than 718 00:36:25,520 --> 00:36:27,879 Speaker 3: a yard away from the receiver, So is a guy 719 00:36:27,960 --> 00:36:31,600 Speaker 3: like really sitting in his back pocket? Of course, the 720 00:36:31,640 --> 00:36:34,680 Speaker 3: pressman corners are going to lead the NFL in that. 721 00:36:34,800 --> 00:36:36,920 Speaker 3: But you know you can filter based on those stats 722 00:36:36,960 --> 00:36:39,600 Speaker 3: and see, really what corners played the stickiest coverage in 723 00:36:39,640 --> 00:36:42,080 Speaker 3: the league, And usually they're pretty good ones like Trent 724 00:36:42,120 --> 00:36:45,440 Speaker 3: McDuffie or Patrick Certain or Riley Moss or whoever that 725 00:36:45,520 --> 00:36:45,799 Speaker 3: might be. 726 00:36:46,280 --> 00:36:48,080 Speaker 1: I only got two more for you. I went past 727 00:36:48,120 --> 00:36:50,680 Speaker 1: what we were supposed to. But this is I'm fascinated, 728 00:36:50,719 --> 00:36:54,360 Speaker 1: Thank you. This has been great. Yards over expected, whether 729 00:36:54,440 --> 00:36:55,760 Speaker 1: for a running back or receiver. 730 00:36:55,840 --> 00:36:56,959 Speaker 2: How do you guys figure that out? 731 00:36:57,600 --> 00:37:01,279 Speaker 3: So we have an expected rush yards model which is 732 00:37:01,360 --> 00:37:06,040 Speaker 3: based off at handoff how much yards is the running 733 00:37:06,080 --> 00:37:08,560 Speaker 3: back or would average running back be expected to gain 734 00:37:08,640 --> 00:37:11,640 Speaker 3: in this situation based on where they are moving and 735 00:37:11,680 --> 00:37:15,799 Speaker 3: where every offensive blocker and every defender is moving and 736 00:37:15,840 --> 00:37:18,800 Speaker 3: their leverage to the running back and where he's running. 737 00:37:19,800 --> 00:37:23,040 Speaker 3: Do I think that it is a great stat for 738 00:37:23,200 --> 00:37:26,520 Speaker 3: measuring what a running back does on a and isolating 739 00:37:26,520 --> 00:37:30,040 Speaker 3: the running backs performance. I actually don't necessarily because you 740 00:37:30,080 --> 00:37:33,799 Speaker 3: know a guy could at handoff look like he's doing 741 00:37:33,800 --> 00:37:36,200 Speaker 3: well on a block and then lose it two seconds 742 00:37:36,280 --> 00:37:39,600 Speaker 3: later and lead to a tackle for a loss. One thing, 743 00:37:39,680 --> 00:37:41,759 Speaker 3: one of the parts that we're hoping to get out 744 00:37:41,760 --> 00:37:43,839 Speaker 3: of the run blocking matchups is to then go back 745 00:37:43,840 --> 00:37:46,840 Speaker 3: and improve our expected rush yards model to kind of 746 00:37:46,960 --> 00:37:50,480 Speaker 3: better give out that division of credit between the offensive 747 00:37:50,480 --> 00:37:53,160 Speaker 3: line and the running back. I would say expected rush 748 00:37:53,239 --> 00:37:55,920 Speaker 3: yards and rushing yards ever expected at this point is 749 00:37:55,960 --> 00:37:58,879 Speaker 3: more of a status of overall run game health than 750 00:37:58,920 --> 00:38:01,759 Speaker 3: any individual player. If you break off a huge run, 751 00:38:01,880 --> 00:38:04,480 Speaker 3: you're gonna lead the NFL and rushing yards ever expected 752 00:38:04,480 --> 00:38:07,319 Speaker 3: because most runs go for three four yards and then 753 00:38:07,360 --> 00:38:09,319 Speaker 3: every once in a while you get eighty. That's plus 754 00:38:09,400 --> 00:38:14,280 Speaker 3: seventy six, and you're good. From a receiving yards over expected. 755 00:38:14,960 --> 00:38:21,840 Speaker 3: That takes into account the completion probabilities, how difficult the 756 00:38:21,920 --> 00:38:24,799 Speaker 3: catch might have been, which again we know there might 757 00:38:24,840 --> 00:38:27,359 Speaker 3: be some issues with in terms of catch probability versus 758 00:38:27,360 --> 00:38:30,600 Speaker 3: completion probability, as I mentioned earlier. And also they expected YAK, 759 00:38:30,600 --> 00:38:33,160 Speaker 3: which I think is probably the most isolated of all those. 760 00:38:33,560 --> 00:38:36,480 Speaker 3: Based on when you catch the ball, where are the 761 00:38:36,520 --> 00:38:38,879 Speaker 3: nearest defenders, Where are they heading and how much YAK 762 00:38:38,880 --> 00:38:41,479 Speaker 3: would you be expected to get? Here's a probability curve 763 00:38:41,480 --> 00:38:43,960 Speaker 3: of all the different yards totals you can get and 764 00:38:44,360 --> 00:38:46,080 Speaker 3: how much WO did you outgain that? So there are 765 00:38:46,280 --> 00:38:48,880 Speaker 3: I think the YAK over expected is a much better 766 00:38:49,360 --> 00:38:53,719 Speaker 3: metric isolating receiver performance than rushing yards over respected would 767 00:38:53,760 --> 00:38:57,400 Speaker 3: be for running backs because there relies on way less pieces. 768 00:38:57,560 --> 00:39:01,200 Speaker 3: Usually when a receivers catching it, it's all the yards 769 00:39:01,200 --> 00:39:03,360 Speaker 3: after the catch that he's gaining are either from the 770 00:39:03,360 --> 00:39:07,080 Speaker 3: accuracy of the quarterback, which we're gaining a little bit 771 00:39:07,160 --> 00:39:10,799 Speaker 3: in the you know where he's throwing him to. Uh, 772 00:39:11,040 --> 00:39:13,680 Speaker 3: and the models picked the expectations picking that up a 773 00:39:13,719 --> 00:39:15,640 Speaker 3: little bit. But once he has the ball in his hands, 774 00:39:15,760 --> 00:39:17,680 Speaker 3: what does he do with it? That's breaking tackles, that's 775 00:39:17,719 --> 00:39:19,239 Speaker 3: making guys miss and that's gonna be a lot more 776 00:39:19,320 --> 00:39:20,560 Speaker 3: receiver based all. 777 00:39:20,480 --> 00:39:23,400 Speaker 2: Right, Final one, how do you dodge miss tackles? 778 00:39:24,320 --> 00:39:25,920 Speaker 1: I know you're probably just going off as you go 779 00:39:25,960 --> 00:39:28,200 Speaker 1: into closest of dots on the you know dots on 780 00:39:28,239 --> 00:39:30,320 Speaker 1: the tracking How do you figure out what you qualify 781 00:39:30,360 --> 00:39:32,040 Speaker 1: as a as a force misstackle? 782 00:39:32,560 --> 00:39:35,719 Speaker 3: Yeah, so that's another model we built a few years ago, 783 00:39:35,760 --> 00:39:39,520 Speaker 3: where essentially we know if a player had a tackle 784 00:39:39,640 --> 00:39:42,280 Speaker 3: or a mistackle on a play. We train a model 785 00:39:42,320 --> 00:39:46,640 Speaker 3: to determine based on you know, the overall dots two 786 00:39:46,680 --> 00:39:50,799 Speaker 3: dimensional movement, whether a player attempted to tackle or not, 787 00:39:50,920 --> 00:39:53,799 Speaker 3: and based on whether they attempted to tackle and the 788 00:39:53,880 --> 00:39:57,480 Speaker 3: proximity to other defenders. So like, if you miss a tackle, 789 00:39:57,520 --> 00:39:59,239 Speaker 3: but he goes right into a group tackle like you 790 00:39:59,320 --> 00:40:03,080 Speaker 3: start the you slow him down and the defense is 791 00:40:03,120 --> 00:40:05,120 Speaker 3: able to pursue and stop the ball carrier right there, 792 00:40:05,120 --> 00:40:06,600 Speaker 3: and that might not be a mistackle. But in the 793 00:40:06,640 --> 00:40:09,080 Speaker 3: open field you dive at the shoe strings of a 794 00:40:09,120 --> 00:40:11,799 Speaker 3: guy and then he's able to get you know, an 795 00:40:11,800 --> 00:40:13,799 Speaker 3: extra couple of yards after that, that it's most likely 796 00:40:13,800 --> 00:40:16,720 Speaker 3: going to be a mistackle. Another one of our models 797 00:40:16,719 --> 00:40:19,560 Speaker 3: that I think we could improve in the future with 798 00:40:19,640 --> 00:40:22,400 Speaker 3: better label set, and that's something that you know, to 799 00:40:22,400 --> 00:40:24,480 Speaker 3: get really nerdy, you might want to get more frame 800 00:40:24,560 --> 00:40:28,839 Speaker 3: level labels rather than play labels where you're guessing where 801 00:40:28,840 --> 00:40:34,400 Speaker 3: the tackle attempt happened. But overall, I think that's another 802 00:40:34,480 --> 00:40:38,040 Speaker 3: area where you know, if we get very usable hawkeye 803 00:40:38,080 --> 00:40:40,600 Speaker 3: tracking data and scouts tracking data, we'll be able to 804 00:40:40,600 --> 00:40:45,200 Speaker 3: get much better approximations of what a actual tackle attempt 805 00:40:45,280 --> 00:40:47,880 Speaker 3: is and not just did you get close enough to 806 00:40:47,960 --> 00:40:52,080 Speaker 3: a guy and your overall kind of velocity vector was 807 00:40:52,080 --> 00:40:55,160 Speaker 3: heading towards the ball carrier fast enough in order to 808 00:40:55,280 --> 00:40:57,080 Speaker 3: qualify as a tackle attempt or not. 809 00:40:57,840 --> 00:41:00,080 Speaker 1: And then finally, Keegan, one of our partners here I 810 00:41:00,120 --> 00:41:02,120 Speaker 1: know they're a big partner of you guys over at 811 00:41:02,160 --> 00:41:05,440 Speaker 1: NFL dot com slash iq specifically is Amazon. How are 812 00:41:05,440 --> 00:41:06,920 Speaker 1: you working with them in order to kind of put 813 00:41:06,920 --> 00:41:09,760 Speaker 1: all this together and bring this that at your broadcast 814 00:41:09,760 --> 00:41:12,360 Speaker 1: partners and the public to make the game more accessible 815 00:41:12,400 --> 00:41:12,840 Speaker 1: for fans. 816 00:41:13,239 --> 00:41:16,480 Speaker 3: Yeah, Amazon is a great partner at every level of this. 817 00:41:17,000 --> 00:41:19,200 Speaker 3: They're where we host all of our data and all 818 00:41:19,239 --> 00:41:23,120 Speaker 3: of our models that processes data in real time. Their 819 00:41:23,280 --> 00:41:26,759 Speaker 3: Amazon Quick is how we build all our internal dashboards 820 00:41:26,800 --> 00:41:31,600 Speaker 3: and the dashboards you see on nfl iq. The Amazon 821 00:41:31,960 --> 00:41:34,560 Speaker 3: Pro Serve team slash you know, they change their name 822 00:41:34,600 --> 00:41:37,239 Speaker 3: every year, but like their data science experts basically are 823 00:41:37,280 --> 00:41:39,720 Speaker 3: the ones that help us with our new stat development 824 00:41:39,760 --> 00:41:41,920 Speaker 3: project every year. So the ones I've talked about in 825 00:41:41,960 --> 00:41:45,120 Speaker 3: the past, and then run blocking schemes and run blocking 826 00:41:45,120 --> 00:41:49,919 Speaker 3: matchups for this offseason that heavily relies on their data 827 00:41:49,960 --> 00:41:53,280 Speaker 3: science expertise. And then you know when you're watching Sunday 828 00:41:53,360 --> 00:41:56,680 Speaker 3: Night Football or the Fox A game or Thursday Night 829 00:41:56,680 --> 00:42:00,280 Speaker 3: Football and you see the sponsored Next Stats segment powered 830 00:42:00,280 --> 00:42:02,960 Speaker 3: by AWS. AWS is the one that's you know, promoting 831 00:42:03,040 --> 00:42:07,480 Speaker 3: us and helping us, you know, get that time during 832 00:42:07,480 --> 00:42:10,200 Speaker 3: a game to really, you know, show what we can 833 00:42:10,239 --> 00:42:11,680 Speaker 3: do in a real live setting. 834 00:42:12,200 --> 00:42:12,439 Speaker 2: Key. 835 00:42:12,440 --> 00:42:13,920 Speaker 1: I know we mentioned a couple of times, but just 836 00:42:13,960 --> 00:42:15,840 Speaker 1: one last time, tell the fans where he could find 837 00:42:16,120 --> 00:42:18,280 Speaker 1: all these great numbers, all the stats, all the information 838 00:42:18,320 --> 00:42:21,120 Speaker 1: that you guys are putting out there in conjunction with Amazon. 839 00:42:21,320 --> 00:42:24,640 Speaker 3: Yeah, NFL i Q, NFL dot Com, slash iq is 840 00:42:24,680 --> 00:42:26,960 Speaker 3: this whole off season project that we're working on. And 841 00:42:27,000 --> 00:42:29,600 Speaker 3: then during the season, if you ever want to, you know, 842 00:42:29,800 --> 00:42:34,080 Speaker 3: watch stats directly linked to film. NFL Pro, which is 843 00:42:34,120 --> 00:42:37,279 Speaker 3: available through NFL Plus is a great platform for that, 844 00:42:37,360 --> 00:42:39,839 Speaker 3: and you'll see all of the insights that our great 845 00:42:39,840 --> 00:42:42,719 Speaker 3: research team puts together every week, the same insights that 846 00:42:42,760 --> 00:42:45,319 Speaker 3: we're given to Tom Brady and Greg Olsen and all 847 00:42:45,320 --> 00:42:47,879 Speaker 3: the other broadcasters during the week. 848 00:42:48,000 --> 00:42:50,200 Speaker 2: So, yeah, awesome stuff. Kick it. 849 00:42:50,200 --> 00:42:51,520 Speaker 1: Thanks so much for the time, man, It's been a 850 00:42:51,520 --> 00:42:53,400 Speaker 1: pleasure talking to you and good luck the rest of 851 00:42:53,400 --> 00:42:53,960 Speaker 1: the off season. 852 00:42:54,400 --> 00:42:55,600 Speaker 3: Thanks John, appreciate it. 853 00:42:55,680 --> 00:42:57,400 Speaker 1: We look forward to seeing where that data come out 854 00:42:57,480 --> 00:42:58,960 Speaker 1: right here on the Johns Little Podcast, brought to you 855 00:42:59,000 --> 00:43:01,640 Speaker 1: by Citizens Official Bank at the Giants from the Hackensack 856 00:43:01,640 --> 00:43:04,359 Speaker 1: Marinehill Podcast Studio. Keep getting better, We'll see you next time. 857 00:43:04,400 --> 00:43:04,840 Speaker 2: Everybody,