WEBVTT - Data Golf 0:00:00.160 --> 0:00:03.240 Welcome back to another edition of the Frida Egg Podcast. 0:00:03.520 --> 0:00:06.400 My name is Garrett Morrison and this episode is brought 0:00:06.400 --> 0:00:10.840 to you by ourselves. So we're having a Black Friday 0:00:10.960 --> 0:00:13.800 sale that is this Friday at our pro shop on 0:00:13.840 --> 0:00:16.720 our website, so you can find it at proshop dot 0:00:16.800 --> 0:00:20.640 Thefrida Egg dot com and it's an automatic twenty percent 0:00:20.720 --> 0:00:26.200 discount off of everything that includes hats, shirts, headcovers, turvice tumblers. 0:00:26.720 --> 0:00:30.560 But I wanted to talk specifically about our photography prints. 0:00:30.640 --> 0:00:34.520 So these are drone shots of great golf courses. Most 0:00:34.520 --> 0:00:37.239 of the shots were taken by Andy Johnson. You can 0:00:37.280 --> 0:00:41.120 get them framed or mounted on metal. We've got pictures 0:00:41.200 --> 0:00:45.400 of bally Neil, Sand Valley, Prairie Dunes, Pasa Tiempo, and 0:00:45.479 --> 0:00:47.040 we've got a ton of new ones on the way. 0:00:47.520 --> 0:00:49.959 They look great. I've got one on my wall myself, 0:00:50.159 --> 0:00:54.560 and they make really good gifts. So this Friday, November 0:00:54.600 --> 0:00:58.120 twenty seventh, twenty percent off the whole store, including our prints. 0:00:58.440 --> 0:01:01.360 That's at the pro Shop, pro show dot Thefrida Egg 0:01:01.880 --> 0:01:06.880 dot com. All right, so on with our episode. My 0:01:06.959 --> 0:01:10.880 guest today is Matt Korshane along with his brother Will 0:01:10.959 --> 0:01:14.640 Korshane Matt founded Data Golf, which you can find at 0:01:14.720 --> 0:01:18.400 datagolf dot com. So Data Golf is really part of 0:01:18.440 --> 0:01:21.920 what has been a revolution in golf over the past 0:01:21.920 --> 0:01:25.960 couple of decades, and that has been this infusion of 0:01:26.280 --> 0:01:31.560 advanced data, statistics and analytics into golf, especially golf at 0:01:31.600 --> 0:01:36.200 the highest professional levels. Now part of that story is Shotlink, 0:01:36.319 --> 0:01:39.360 which provides a great deal of raw data that simply 0:01:39.400 --> 0:01:43.200 wasn't available in the nineteen nineties. And another part of 0:01:43.240 --> 0:01:47.040 it is the advent of strokes gained statistics, which were 0:01:47.080 --> 0:01:52.040 pioneered by the Columbia Business School professor Mark Brody. And basically, 0:01:52.040 --> 0:01:56.720 what strokes gained does is it compares a player's performance 0:01:56.760 --> 0:02:00.600 to the rest of the field, shot by shot, and 0:02:00.680 --> 0:02:03.280 in this way you can really isolate the value or 0:02:03.400 --> 0:02:07.360 quality of each shot a player hits. Now, in strokes gained, 0:02:07.440 --> 0:02:10.920 with each shot the player gains or loses a certain 0:02:11.040 --> 0:02:14.480 amount on the field. And what this has done is 0:02:14.520 --> 0:02:17.239 it's just given us a much better way to analyze 0:02:17.240 --> 0:02:20.000 the different skills that a player has. We really have 0:02:20.040 --> 0:02:23.320 a better idea of who the great drivers and iron 0:02:23.360 --> 0:02:27.240 players actually are because we've been able to zero in 0:02:27.320 --> 0:02:30.480 on the quality of each strike of the ball. But 0:02:30.760 --> 0:02:34.040 you know, strokes gained is like still in its infancy 0:02:34.280 --> 0:02:37.000 and it's just begun to shift our understanding of golf. 0:02:37.040 --> 0:02:40.480 I think where we're just scratching the surface, and that's 0:02:40.600 --> 0:02:44.360 the exciting arena in which Data Golf is working right now. 0:02:44.440 --> 0:02:48.320 Matt and Will Korshane are very young, but also very qualified, 0:02:48.600 --> 0:02:51.400 and they're doing a lot of interesting work. One of 0:02:51.480 --> 0:02:54.280 the things that I think Data Golf could change is 0:02:54.320 --> 0:02:58.560 how we view golf courses in golf course design now, 0:02:58.960 --> 0:03:01.840 I always think that golf course design will be an 0:03:01.960 --> 0:03:05.440 art and assessing golf course design will be an art 0:03:05.480 --> 0:03:08.880 as well. Know, no, data, I don't think can tell 0:03:08.960 --> 0:03:11.800 us if a golf course is good, but it can 0:03:11.840 --> 0:03:15.040 definitely tell us how. Data can definitely tell us how 0:03:15.160 --> 0:03:19.680 different tournament venues demand different skill sets from players. And 0:03:19.720 --> 0:03:21.680 I find that to be a really compelling question. And 0:03:21.720 --> 0:03:24.679 so that's what I wanted to talk with Matt Corshane 0:03:24.720 --> 0:03:27.920 about because Data Golf, the website that he and his 0:03:27.919 --> 0:03:31.920 brother have developed, has some great tools that use strokes 0:03:31.960 --> 0:03:35.040 gained data to tell you something about the golf courses 0:03:35.080 --> 0:03:38.920 on the PGA Tour, specifically this week's host of the 0:03:38.960 --> 0:03:43.640 Mayacoba Golf Classic, El kamal Ane Golf Club outside of Cancun, Mexico, 0:03:43.840 --> 0:03:47.280 is one of the biggest outliers in Data Golf's model. 0:03:47.760 --> 0:03:51.080 It's such a strange course. It just demands this extremely 0:03:51.320 --> 0:03:55.200 unusual type of performance from the top players. So I 0:03:55.200 --> 0:03:57.600 thought it was a good time to get at a 0:03:57.720 --> 0:04:00.520 question that I've had for a long time with the 0:04:00.520 --> 0:04:04.280 help of Matt Course Shane, what can statistics tell us 0:04:04.560 --> 0:04:07.720 about golf course architecture on the PGA Tour. I hope 0:04:07.760 --> 0:04:10.360 you enjoy I miss the green for example, I'm already 0:04:10.440 --> 0:04:12.480 upset when I find my ball in the bunker, I'm 0:04:12.480 --> 0:04:13.200 really upset. 0:04:13.280 --> 0:04:13.760 And when I. 0:04:13.720 --> 0:04:16.160 Find my ball in a brid egg Friday egg, the 0:04:16.240 --> 0:04:19.719 dreaded Frida egg fridagg Frida egg egg Frida egg bride 0:04:19.720 --> 0:04:45.080 egg Lie, I'm about ready to run off the golf course. So, 0:04:45.279 --> 0:04:47.599 just to find out a little bit about your background 0:04:47.600 --> 0:04:51.320 in golf, you were a pretty excellent golfer. You got 0:04:51.360 --> 0:04:54.400 down to a scratch handicap, You played some college golf 0:04:54.400 --> 0:04:58.200 at Queen's University as a competitive golfer. Did data ever 0:04:58.760 --> 0:04:59.400 enter your life? 0:05:00.080 --> 0:05:01.000 No, it really didn't. 0:05:01.320 --> 0:05:04.240 Yeah, I wasn't logging like the Strokes game stats or anything. 0:05:04.240 --> 0:05:06.520 I was not a and I'm still not. I'm very 0:05:06.600 --> 0:05:08.599 much a I mean, I'm a vivid head case on 0:05:08.640 --> 0:05:10.599 the golf course. Now, I think I'm very much a 0:05:10.600 --> 0:05:12.400 field player. I guess I would say I'm not like 0:05:12.480 --> 0:05:15.640 a yeah, technical robotic player by any means. 0:05:16.040 --> 0:05:20.320 That's an interesting contrast to me. So like if data 0:05:20.360 --> 0:05:23.440 were you know, some of these analytics were around back then, 0:05:23.480 --> 0:05:25.960 but they just weren't as commonly used, I guess, especially 0:05:25.960 --> 0:05:28.480 at the college golf level. Do you think anything would 0:05:28.480 --> 0:05:29.160 have changed. 0:05:28.880 --> 0:05:31.440 For you looking back when I was actually a serious 0:05:31.480 --> 0:05:33.600 player and didn't have because honestly, the way I think 0:05:33.640 --> 0:05:35.440 about it now is I just have on the list 0:05:35.480 --> 0:05:37.839 of the top ten things I need to improve in golf, 0:05:38.520 --> 0:05:41.040 understanding how to better use data is pretty far down 0:05:41.040 --> 0:05:43.440 that list. I need to stop like blocking my opening 0:05:43.480 --> 0:05:46.200 TV all ob before I start worrying about that stuff. 0:05:46.600 --> 0:05:48.880 But like, yeah, no, back in the day when I 0:05:48.920 --> 0:05:50.719 was good, it would be Yeah, it'd be super informalve 0:05:50.720 --> 0:05:52.839 just to get the basic numbers like the Strokes game stuff, 0:05:52.880 --> 0:05:55.719 just using Brodie's app, Like just understanding what you're losing 0:05:55.760 --> 0:05:58.839 strokes is super valuable, and I think it's often for 0:05:58.920 --> 0:06:01.479 most people it's counter intuitive. I think probably a lot 0:06:01.480 --> 0:06:03.719 of people would end myself included, would probably realize they're 0:06:03.760 --> 0:06:06.599 losing most shots, like on the long game, whenever you 0:06:06.600 --> 0:06:08.719 play poorly, or always thinking about all these pots you miss, 0:06:09.440 --> 0:06:12.159 et cetera. But really everybody misses a lot of potts, 0:06:12.160 --> 0:06:14.320 so I think, yeah, But anyway to answer the question, 0:06:14.440 --> 0:06:16.200 I think I would have taken advantage of this stuff. 0:06:16.200 --> 0:06:18.280 It was more easily accessible in ten years ago. 0:06:18.560 --> 0:06:21.720 Well. Hearing golfers talk about it, it's almost like, in 0:06:21.720 --> 0:06:25.880 addition to providing knowledge, these stats sometimes provide a kind 0:06:25.880 --> 0:06:29.560 of peace of mind, like you know what's wrong, yeah, 0:06:29.839 --> 0:06:33.960 or or you can contextualize your bad performances as just 0:06:34.000 --> 0:06:36.279 like this is what happens. You know, it's not a 0:06:36.360 --> 0:06:38.880 judgment on me as as a player. This is just 0:06:38.960 --> 0:06:39.760 what happens. 0:06:40.200 --> 0:06:40.400 Yeah. 0:06:40.400 --> 0:06:42.800 And it's also I kind of think that is the 0:06:42.880 --> 0:06:44.960 right way to think about it, because it's also debatable 0:06:45.000 --> 0:06:48.159 how like actionable this stuff is. Like if I find 0:06:48.160 --> 0:06:50.680 out that I am lacking, well, if I'm losing small 0:06:50.760 --> 0:06:52.640 good example of if I'm losing strokes, offer te it's 0:06:52.680 --> 0:06:55.120 like okay, like, yeah, if I hit. 0:06:55.000 --> 0:06:56.720 It further, that'd be great. But I mean there are 0:06:56.720 --> 0:06:58.000 examples of players. 0:06:57.640 --> 0:06:59.720 Who have like doing for Telly last week was getting 0:06:59.720 --> 0:07:02.880 a lot of attention for he's not a massive guy, 0:07:02.920 --> 0:07:05.080 and he managed to pick up some club at speech. 0:07:05.080 --> 0:07:06.200 So there are things you can do. 0:07:06.279 --> 0:07:08.000 But I've always thought, I think my brother and I've 0:07:08.000 --> 0:07:10.560 always thought that, Yeah, it's more like descriptive. It's nice 0:07:10.600 --> 0:07:13.000 to know where you're losing strokes and what is the 0:07:13.000 --> 0:07:15.280 difference between myself and like the number one player in 0:07:15.280 --> 0:07:15.640 the world. 0:07:15.640 --> 0:07:18.120 But I don't know how actionable a lot of it is. 0:07:19.280 --> 0:07:22.760 So right now you're doing a PhD at the University 0:07:22.800 --> 0:07:26.840 of British Columbia in economics. Where did your interest in 0:07:26.880 --> 0:07:27.800 economics come from? 0:07:28.360 --> 0:07:31.480 So my grandfather is, well, he's still alive. 0:07:31.560 --> 0:07:33.200 He was a pretty stop working but he was a 0:07:33.200 --> 0:07:35.800 pretty well known economist, so there was always that connection. 0:07:36.400 --> 0:07:40.360 I actually did my undergrad in biochemistry, which was I mean, 0:07:40.400 --> 0:07:43.200 I was just a classic. I just went to university 0:07:43.240 --> 0:07:44.680 not really knowing what to do, and I was like, oh, 0:07:44.840 --> 0:07:46.360 I mean why not medical school? 0:07:46.360 --> 0:07:47.880 Like that sounds reasonable. 0:07:47.920 --> 0:07:50.400 It's embarrassing looking back on it, but then I eventually 0:07:50.440 --> 0:07:52.920 realized I think it's like a few Coon courses and 0:07:52.960 --> 0:07:56.400 just sort of liked the general approach to things. Just 0:07:56.760 --> 0:07:59.880 using statistical methods, which are particularly the ones that are 0:08:00.160 --> 0:08:02.200 prominent in economics, Just using those as a way to 0:08:02.280 --> 0:08:04.760 understand the world is. It's just changed the way I 0:08:04.760 --> 0:08:06.520 think about a lot of things, and I've really become 0:08:07.200 --> 0:08:09.200 I really enjoy looking at the world through that lens. 0:08:09.960 --> 0:08:13.800 So at some point you combined your background in golf 0:08:14.480 --> 0:08:19.640 with your interest in statistics and informed data golf. How 0:08:19.680 --> 0:08:22.200 did the idea for data golf come about? 0:08:22.560 --> 0:08:23.840 Well, I guess the first thing I just say is, 0:08:24.000 --> 0:08:26.360 so data Golf is run by myself from my younger brother, Will, 0:08:26.440 --> 0:08:29.600 who's so he's two years younger. Yeah, he was working 0:08:29.640 --> 0:08:31.720 at that point, and I was just in my second 0:08:31.760 --> 0:08:32.760 year of the PhDs. 0:08:32.840 --> 0:08:34.720 This is like four or five years ago. And then 0:08:34.720 --> 0:08:36.199 the Yeah, the PJ Tour used to. 0:08:36.160 --> 0:08:39.400 Have an academic program where you could access the shot 0:08:39.440 --> 0:08:42.080 link data for free, and so we got that data 0:08:42.120 --> 0:08:42.920 and we just started. 0:08:43.360 --> 0:08:45.080 Yeah, we just started a Twitter account. 0:08:44.840 --> 0:08:48.680 Started a WordPress blog and started doing some basic analyses 0:08:48.880 --> 0:08:50.960 with the data, trying to just sort of answer. 0:08:51.559 --> 0:08:52.400 It's funny to look back. 0:08:52.400 --> 0:08:54.640 We still have all these blogs somewhere on our old website. 0:08:54.640 --> 0:08:56.280 It's funny to look back on them. But yeah, just 0:08:56.280 --> 0:08:59.440 answering questions like do players in fact, there's the same 0:08:59.440 --> 0:09:01.199 wording out you shoot a really good round, it's hard 0:09:01.200 --> 0:09:02.640 to follow it up, So we were sort of just 0:09:02.880 --> 0:09:06.040 checking the degree to which that's all or true and yeah, 0:09:06.040 --> 0:09:07.760 and then it's sort of just progressed from there. 0:09:08.240 --> 0:09:12.439 Right for sure. So it's really evolved well past the 0:09:12.880 --> 0:09:16.160 original blog. So the original Data Golf blog was basically 0:09:16.160 --> 0:09:20.559 a word Press basic WordPress blog, and now it's this 0:09:20.920 --> 0:09:25.440 very complex website with not only blog posts, but a 0:09:25.520 --> 0:09:28.480 number of predictive models and interactive tools that kind of 0:09:28.520 --> 0:09:32.400 make sense of the huge amount of data that's coming 0:09:32.400 --> 0:09:35.560 out of professional golf right now. And so I wanted 0:09:35.559 --> 0:09:39.360 to focus actually on the interactive tools and visualizations. So 0:09:39.640 --> 0:09:42.480 you know, to someone like me, I find these very 0:09:42.600 --> 0:09:46.360 useful and very informative just in general, How do you 0:09:46.400 --> 0:09:48.600 decide what kinds of tools to build? 0:09:49.400 --> 0:09:51.960 Yeah, so I think a lot of the tools come 0:09:52.080 --> 0:09:54.920 naturally out of the model that we have, So really 0:09:54.960 --> 0:09:56.920 like the centerpiece of the website and what we do 0:09:56.960 --> 0:09:59.040 with Data Golf is this predictive model. We have of 0:09:59.360 --> 0:10:01.840 golfer perform, where essentially the output of that model is 0:10:01.840 --> 0:10:04.599 we're trying to, like, ultimately what matters in golf is 0:10:04.640 --> 0:10:06.760 strokes gain, how many strokes you're beating the field by. 0:10:06.880 --> 0:10:09.080 So that's ultimately what we're trying to predict with this model. 0:10:09.440 --> 0:10:11.800 And along the way there will be things that I 0:10:11.800 --> 0:10:13.199 don't know, sort to pop out of the model that 0:10:13.240 --> 0:10:14.280 we can make into a webpage. 0:10:14.280 --> 0:10:16.600 So a good example is true strokes gained. 0:10:16.679 --> 0:10:19.680 So true strokes gained is it's regular strokes gain, which 0:10:19.720 --> 0:10:22.080 is how many strokes beat the feeld by, except we're gonna. 0:10:21.880 --> 0:10:24.520 Adjust that for the strength of that field. 0:10:24.600 --> 0:10:27.320 So, for example, on the web dot com, if we 0:10:27.480 --> 0:10:29.560 estimate that those fields are like a shot worse on 0:10:29.640 --> 0:10:33.040 average than TJ tour field, So if you beat a 0:10:33.040 --> 0:10:35.120 web dot com field by two shots, that's gonna be 0:10:35.120 --> 0:10:37.520 worth like a true strokes gin of one because we're 0:10:37.520 --> 0:10:39.480 saying this field is one shot. 0:10:39.240 --> 0:10:41.079 Worse than average. So that's something that sort of comes 0:10:41.080 --> 0:10:41.680 out of our model. 0:10:41.720 --> 0:10:44.720 Another example of a page would be the course history 0:10:44.760 --> 0:10:47.800 stuff or the course fit tool. We have that output 0:10:47.800 --> 0:10:50.480 in our in our model, and at some point we say, wow, 0:10:50.520 --> 0:10:52.680 this is interesting. I think general golf fans would like this. 0:10:52.800 --> 0:10:55.160 Let's turn it into like an intuitive page that people 0:10:55.160 --> 0:10:56.280 can can get something out of. 0:10:56.320 --> 0:10:57.920 So where that's where a lot of them come from. 0:10:58.720 --> 0:11:00.600 So I did want to dig a little bit deeper 0:11:00.800 --> 0:11:05.280 into what you call the course fit tool, because I 0:11:05.320 --> 0:11:07.280 think it does have a lot to say about what 0:11:07.679 --> 0:11:10.319 one of our primary interests at the fried Egg is, 0:11:10.440 --> 0:11:13.880 and that's golf course design and set up and how 0:11:14.679 --> 0:11:18.120 design and set up influence play on the professional tours. 0:11:18.160 --> 0:11:20.960 But I mean, I think maybe before getting into course 0:11:21.000 --> 0:11:24.960 fit specifically, maybe we should lay the groundwork with this 0:11:25.240 --> 0:11:29.319 related but very different notion of course history. Can you 0:11:29.400 --> 0:11:32.040 tell me what course history is and some of the 0:11:32.080 --> 0:11:33.520 issues that you've had with it. 0:11:34.280 --> 0:11:34.480 Yeah. 0:11:34.520 --> 0:11:36.440 So, course istuy in the most general sense, is just 0:11:36.559 --> 0:11:38.840 how a player has performed at a specific course in 0:11:38.840 --> 0:11:41.640 the past. And then the way we analyze it is 0:11:42.160 --> 0:11:44.600 not just how like a player has performed in an 0:11:44.640 --> 0:11:47.920 absolute sense at a course, because, for example, Tiger, if 0:11:47.960 --> 0:11:50.319 you just define course a player's course history as just 0:11:50.360 --> 0:11:52.760 their total stroke seeing at that course, and somebody like 0:11:52.800 --> 0:11:54.880 Tiger is going to have the best course history everywhere. 0:11:55.320 --> 0:11:58.760 So course history, the way we define it is performance 0:11:58.800 --> 0:12:02.959 relative to expectation or baseline. So and by expectation or 0:12:03.000 --> 0:12:05.840 baseline that's going to be intuitally, that's like how a 0:12:05.880 --> 0:12:09.160 player performs across all courses in a given year, that's 0:12:09.160 --> 0:12:11.679 going to be their baseline. So it might be there 0:12:11.720 --> 0:12:14.280 might be a plus one strokes gained player. And then 0:12:14.600 --> 0:12:17.880 if there's a specific specific course, say I guess the National, 0:12:17.880 --> 0:12:20.560 where they've performed at an average of plus two strokes gained, 0:12:21.160 --> 0:12:23.240 we're gonna say they have good course history, dick, because 0:12:23.240 --> 0:12:26.640 they performed above their baseline. The main issue with course 0:12:26.679 --> 0:12:30.240 history in terms of trying to use it for actionable 0:12:30.280 --> 0:12:33.000 things like predicting performance, is that it's generally a small sample. 0:12:33.120 --> 0:12:35.240 If a guy has only played four rounds of the 0:12:35.280 --> 0:12:37.600 course and he's played really well, maybe he won the tournament, 0:12:37.679 --> 0:12:41.400 which is not not too uncommon, what can we really 0:12:41.440 --> 0:12:43.719 say from that, Like from our analysis, we've found like 0:12:43.800 --> 0:12:46.520 if a guy is performing one stroke better than expected 0:12:46.559 --> 0:12:49.760 at a course, historically, we're only going to improve his 0:12:49.920 --> 0:12:52.839 predictive performance going forward by maybe like two or three 0:12:52.880 --> 0:12:53.960 percent per round. 0:12:54.000 --> 0:12:55.800 Let's say, so, well, if he's if he's done that 0:12:55.840 --> 0:12:56.240 for four. 0:12:56.160 --> 0:12:58.480 Rounds, we would probably bump up his predicted skill by 0:12:58.520 --> 0:13:01.680 like zero point zero five if he's been one stroke planner. 0:13:01.720 --> 0:13:04.840 So it's a pretty small adjustment, but those things can matter. 0:13:05.720 --> 0:13:08.120 Yeah, it seems like your opinion on this has has 0:13:08.160 --> 0:13:11.960 shifted a little bit over time, but the general critique 0:13:12.040 --> 0:13:15.280 of how maybe the lay person uses course history is 0:13:15.320 --> 0:13:18.600 still valid. You know, when when you hear people talking 0:13:18.640 --> 0:13:21.480 about how a certain player has done well at a 0:13:21.520 --> 0:13:23.840 course in the past and is using that fact to 0:13:23.840 --> 0:13:27.200 say this player will definitely do well this week at 0:13:27.200 --> 0:13:30.760 the same course, that that's a vast oversimplification of how 0:13:30.800 --> 0:13:31.960 things actually work. 0:13:31.840 --> 0:13:33.120 Right, Yeah, it is. 0:13:33.600 --> 0:13:37.000 In general, making strong claims about predicting performance in golf 0:13:37.080 --> 0:13:40.080 is a bad idea, just because there's so much there's 0:13:40.080 --> 0:13:42.439 so much randomness in golf as we like, even just 0:13:42.520 --> 0:13:45.160 last week, like like Bryson losing to Lang, right, I guess, 0:13:45.200 --> 0:13:47.800 so like obviously this is a very unexpected outcome, but 0:13:47.840 --> 0:13:50.280 those things happen in golf. So but to your point 0:13:50.320 --> 0:13:52.679 about us changing our opinion, that's definitely true. I mean 0:13:52.920 --> 0:13:54.760 when we first started out on Twitter, we were more 0:13:54.840 --> 0:13:57.440 maybe rash and like combat of I guess to some 0:13:57.520 --> 0:13:59.720 debdeon and we took a pretty hard stance on course 0:13:59.800 --> 0:14:02.160 history saying it didn't matter. But I think we've definitely 0:14:02.200 --> 0:14:05.800 softened that. There are just examples like Phil Augusta. He 0:14:05.840 --> 0:14:08.560 now has like he's played like so many rounds there, 0:14:08.600 --> 0:14:12.840 like like sixty seventy rounds, and he's averaging almost a 0:14:12.920 --> 0:14:15.520 shot better than than expected Augusta, which is a huge 0:14:15.520 --> 0:14:17.880 that's a huge difference, and it can't be you can't 0:14:17.920 --> 0:14:20.080 just wave that away by saying, oh, that's just randomness. 0:14:20.120 --> 0:14:22.840 That's pretty meaningful. So, yeah, course history, when you get 0:14:22.840 --> 0:14:24.280 big examples, can matter for sure. 0:14:25.000 --> 0:14:27.880 Now, so course history can matter a little bit. There's 0:14:27.960 --> 0:14:30.080 there's a small adjustment that you can make for a 0:14:30.120 --> 0:14:32.880 player's course history when you're trying to predict their performance 0:14:32.920 --> 0:14:35.760 in a given week. It seems to me that course 0:14:35.920 --> 0:14:39.360 fit in a lot of ways makes up for some 0:14:39.480 --> 0:14:43.680 of the deficiencies of course history. Is that basically right? 0:14:44.040 --> 0:14:45.040 Yeah, that's that's definitely right. 0:14:45.080 --> 0:14:46.720 You get you sort of with So with course fit, 0:14:47.240 --> 0:14:50.640 what we're gonna look at is which skills a given 0:14:50.680 --> 0:14:53.200 course favors. So like the skills, we're gonna focus on 0:14:53.200 --> 0:14:56.760 our driving distance, driving accuracy and strokes and putting around 0:14:56.760 --> 0:14:59.680 the green approach, and by focusing on skills, you kind 0:14:59.720 --> 0:15:02.960 of bypass this sample size problem. So if we know 0:15:03.400 --> 0:15:07.280 through our fancy statistical methods that Augusta favors people hit 0:15:07.320 --> 0:15:10.120 it further. And even if there's a guy who a 0:15:10.120 --> 0:15:12.560 first time round Augusta, so we have no course history 0:15:12.640 --> 0:15:14.240 date on him, but if we know he hits it 0:15:14.280 --> 0:15:16.600 above average distances, we can use course fit to make 0:15:16.760 --> 0:15:19.040 some judgments about how he's going to perform in that course. 0:15:19.080 --> 0:15:23.440 So yeah, by focusing on characteristics instead of specific players, 0:15:23.480 --> 0:15:25.360 which is what you do with course history, of course, 0:15:25.400 --> 0:15:27.120 it can be a lot more meaningful. Yeah, because you 0:15:27.160 --> 0:15:28.880 don't have a we have a ton of data to 0:15:29.000 --> 0:15:31.920 estimate how how much driving distance is favored at Augusta, 0:15:32.000 --> 0:15:33.600 we can use every player to understand that. 0:15:34.120 --> 0:15:37.320 Right. Yeah, So course fit just basically defined, is is 0:15:37.400 --> 0:15:40.600 kind of the degree to which a given golf course 0:15:40.880 --> 0:15:45.560 favors a specific skill set, Right, And those the skills 0:15:45.560 --> 0:15:48.720 that you measure are you know, maybe it can help 0:15:48.720 --> 0:15:52.960 me out here driving distance, driving accuracy, strokes gained, approach, 0:15:53.120 --> 0:15:56.640 strokes gained around the green, and strokes gained putting. Yeah, 0:15:57.000 --> 0:15:59.120 those are the kind of the five skills that you 0:15:59.520 --> 0:16:02.000 measure when you're assessing whether a course kind of fits 0:16:02.000 --> 0:16:06.360 a particular kind of player. So has anything kind of 0:16:06.400 --> 0:16:09.400 jumped out and surprised you since you started looking at 0:16:09.760 --> 0:16:10.280 course fit. 0:16:11.000 --> 0:16:13.680 I think the I mean, the main takeaways I think 0:16:13.680 --> 0:16:16.080 from course fit are like with the data that we have, 0:16:16.280 --> 0:16:17.920 it seems like the way that court, the way that 0:16:18.440 --> 0:16:21.640 PGA tour courses differ is along the dimensions of how 0:16:21.720 --> 0:16:25.040 much they favored driving distance and driving accuracy. Certainly we 0:16:25.080 --> 0:16:27.840 can make intuitiveur arguments for why different courses favor putting 0:16:27.960 --> 0:16:29.720 or around the green, but I think the reality is, 0:16:29.760 --> 0:16:32.200 like the data, especially the putting, it's just so there's 0:16:32.240 --> 0:16:35.080 so much variance that it's hard to it's hard to 0:16:35.080 --> 0:16:37.920 really say whether or not a course favors putting more 0:16:37.960 --> 0:16:40.480 than the average, and that sort of reflected in the tool. 0:16:40.560 --> 0:16:43.080 I think, I guess my takeaway like, I think I was. 0:16:43.080 --> 0:16:46.520 A bit surprised. Maybe how still important? Again, there are 0:16:46.520 --> 0:16:48.720 a lot of caveats here, but how important driving accuracy 0:16:48.800 --> 0:16:51.440 is just because there is the narrative of I mean, 0:16:51.440 --> 0:16:53.240 this is how I'm looking at it. Somebody else could 0:16:53.240 --> 0:16:55.000 look at it and say, oh, I'm surprised how much 0:16:55.120 --> 0:16:58.080 that distance is the most favored skill, And just for 0:16:58.080 --> 0:17:00.520 people who aren't looking at this thing. And basically the 0:17:00.600 --> 0:17:03.400 hierarchy is the way we have it is basically distance 0:17:03.560 --> 0:17:06.000 at the average course is the most particular power, along 0:17:06.000 --> 0:17:09.040 with approaching approach, and then it sort of goes. 0:17:09.760 --> 0:17:10.639 I guess driving. 0:17:10.400 --> 0:17:13.480 Accuracy, putting around the green are all pretty similar. Around 0:17:13.480 --> 0:17:15.399 the green is probably the smallest, but that's sort of 0:17:15.400 --> 0:17:18.520 the hierarchy. And then there's all sorts of interesting things 0:17:18.560 --> 0:17:21.280 that specific courses that you could get into. 0:17:21.440 --> 0:17:23.359 Yeah, and we'll get into that a little bit in 0:17:23.400 --> 0:17:26.720 a minute here, But first thing, maybe you could just 0:17:26.760 --> 0:17:29.840 describe the course fit tool a little bit. I think 0:17:29.840 --> 0:17:32.560 people should go and see if the visualization is really 0:17:32.600 --> 0:17:35.840 simple and effective, and it would be probably hard to 0:17:35.880 --> 0:17:39.199 describe over the radio basically what it looks like, but 0:17:39.240 --> 0:17:41.280 could you just tell me basically what the course fit 0:17:41.359 --> 0:17:43.520 tool is and how you built it. 0:17:44.200 --> 0:17:46.480 So the course fit tool, it's essentially what we're just 0:17:46.520 --> 0:17:49.520 trying to visualize, as you said earlier, the degree to 0:17:49.560 --> 0:17:52.160 which each course on the PGA Tour favors each skill, 0:17:52.560 --> 0:17:56.000 and so what we have on it is a basically 0:17:56.119 --> 0:17:58.760 have these shapes. Each shape represents the degree to which 0:17:58.840 --> 0:18:01.679 the course favors skills and so we have the average 0:18:01.680 --> 0:18:03.880 PGA Tour course on there, and then you can sort 0:18:03.880 --> 0:18:06.360 of easily see when we overlay another course on top 0:18:06.400 --> 0:18:09.040 of that whether or not that overlaid course favors distance 0:18:09.119 --> 0:18:11.880 more than the average course. Same with accuracy, et cetera. 0:18:12.320 --> 0:18:13.560 And yeah, I don't know if we want to get 0:18:13.560 --> 0:18:16.639 into the details of how these numbers are actually calculated. 0:18:16.720 --> 0:18:19.880 Is it's somewhat complicated, but the intuitive all we're really 0:18:19.880 --> 0:18:22.800 doing with this tool is like, if you take driving distance, 0:18:22.840 --> 0:18:25.320 for example, we're gonna before each event let's play on 0:18:25.320 --> 0:18:27.560 the PGA Tour or each round, we're gonna have an 0:18:27.600 --> 0:18:30.520 estimate of every player's driving distance and every player's driving accuracy, 0:18:30.560 --> 0:18:33.880 et cetera. And to estimating how much a course favors 0:18:33.960 --> 0:18:37.320 driving distance, we're basically just gonna compare two players, and 0:18:37.320 --> 0:18:39.720 this is important, two players who are similar in every 0:18:39.720 --> 0:18:41.000 dimension except distance. 0:18:41.160 --> 0:18:41.600 So there's one. 0:18:41.560 --> 0:18:43.280 Player, say that hits it ten years further than the other. 0:18:43.720 --> 0:18:46.480 So we're gonna take those two players, compare their stroke 0:18:46.520 --> 0:18:48.920 scheme in that round, and then you do that many 0:18:48.920 --> 0:18:50.560 times and eventually you're able to say, Okay, if you 0:18:50.640 --> 0:18:53.360 hit a ten yards further than somebody, that translates into 0:18:53.400 --> 0:18:56.639 whatever point three point four strokes gained around over that 0:18:56.680 --> 0:18:58.280 person holding everything else fixed. 0:18:59.000 --> 0:18:59.760 And it's right. 0:18:59.800 --> 0:19:02.520 You know, the visualization makes this very clear because this 0:19:02.560 --> 0:19:05.080 is an octagon. Basically, right, there are five skills, so 0:19:05.720 --> 0:19:08.159 there are there are five points on the octagon, and 0:19:08.359 --> 0:19:11.479