1
00:00:01,040 --> 00:00:03,680
Speaker 1: All right, and welcome back to the Betting Pros Podcast.

2
00:00:03,720 --> 00:00:06,000
Speaker 1: We are under one month away from the NFL regular

3
00:00:06,040 --> 00:00:09,000
Speaker 1: season kicking off, so today we'll be breaking down how.

4
00:00:08,880 --> 00:00:11,959
Speaker 2: To build your own sports betting model for the NFL.

5
00:00:12,119 --> 00:00:14,320
Speaker 1: My name is Seth Wilcock, and today we'll be tapping

6
00:00:14,360 --> 00:00:17,079
Speaker 1: into the big brain of our data scientists here at

7
00:00:17,079 --> 00:00:17,680
Speaker 1: Fantasy Pro.

8
00:00:17,840 --> 00:00:20,200
Speaker 2: Sam Hoppin' Sam, what's hopping today?

9
00:00:20,239 --> 00:00:20,400
Speaker 3: Man?

10
00:00:20,440 --> 00:00:22,759
Speaker 1: I know you're just back from doing some world traveling

11
00:00:22,760 --> 00:00:23,360
Speaker 1: over the weekend.

12
00:00:24,400 --> 00:00:28,840
Speaker 4: Yeah, I'm back. I'm doing great. Excited to dust off

13
00:00:28,880 --> 00:00:32,360
Speaker 4: some of the podcasting cobwebs, if you will. It's been

14
00:00:32,840 --> 00:00:36,000
Speaker 4: been a little while since I've been on the recording

15
00:00:36,080 --> 00:00:40,120
Speaker 4: side of things, So happy to be here. Excited to gosh.

16
00:00:40,200 --> 00:00:44,280
Speaker 4: I think we're almost exactly a month away from the

17
00:00:44,320 --> 00:00:49,040
Speaker 4: first NFL regular season games, so getting prepped for that

18
00:00:49,159 --> 00:00:51,599
Speaker 4: and excited to get back into it.

19
00:00:52,240 --> 00:00:54,720
Speaker 1: Yeah. Glad we could break you out from your laboratory

20
00:00:54,760 --> 00:00:56,800
Speaker 1: and bring you here front and center for the great

21
00:00:56,840 --> 00:00:59,240
Speaker 1: people on the Betting Pros Podcast. I want to start

22
00:00:59,280 --> 00:01:02,360
Speaker 1: by talking about the basics of sports betting modeling. Sam,

23
00:01:02,480 --> 00:01:04,160
Speaker 1: can you tell us a little bit about yourself and

24
00:01:04,480 --> 00:01:07,080
Speaker 1: your inner nerd and what you got you into creating

25
00:01:07,120 --> 00:01:08,360
Speaker 1: your own sports betting models?

26
00:01:08,400 --> 00:01:13,400
Speaker 4: Man, Yeah, of course. So I've been doing this for

27
00:01:13,720 --> 00:01:19,520
Speaker 4: a couple of years now and have a non traditional background,

28
00:01:19,600 --> 00:01:23,120
Speaker 4: like a lot of people, did study marketing in undergrad

29
00:01:23,160 --> 00:01:26,319
Speaker 4: but then went back to get my masters and data science,

30
00:01:26,360 --> 00:01:30,959
Speaker 4: so learned a lot of what I use in that program.

31
00:01:31,800 --> 00:01:35,399
Speaker 4: Finished that up a couple of years ago and have

32
00:01:36,319 --> 00:01:40,720
Speaker 4: been doing a variety of sports. I joined betting Pros

33
00:01:40,840 --> 00:01:45,880
Speaker 4: last year, but I've been doing some NFL stuff for them,

34
00:01:46,400 --> 00:01:52,880
Speaker 4: also focusing on the NBA and WNBA, and previously had

35
00:01:52,960 --> 00:01:57,240
Speaker 4: done some work on Formula one models as well.

36
00:01:58,280 --> 00:02:00,480
Speaker 1: Sick, you're doing the lord's work out here pros and

37
00:02:00,520 --> 00:02:04,080
Speaker 1: Fantasy pros, so we appreciate that. Man, what exactly, from

38
00:02:04,080 --> 00:02:07,280
Speaker 1: an elementary perspective is a sports betting model and what

39
00:02:07,360 --> 00:02:09,239
Speaker 1: makes it useful for sports betters?

40
00:02:09,240 --> 00:02:12,240
Speaker 4: Would you say? It can really mean a lot of things, right.

41
00:02:12,240 --> 00:02:17,640
Speaker 4: It can be your classic regression or classification model predicting. Sure,

42
00:02:17,680 --> 00:02:20,280
Speaker 4: you know sort of who will win the game, how

43
00:02:20,360 --> 00:02:23,400
Speaker 4: many rushing yards a player is going to have, but

44
00:02:23,480 --> 00:02:26,320
Speaker 4: there are some other techniques you can use as well,

45
00:02:26,440 --> 00:02:29,520
Speaker 4: you know, just using some sampling which is at a

46
00:02:29,600 --> 00:02:35,080
Speaker 4: high level, taking games with similar characteristics and calculating different

47
00:02:35,160 --> 00:02:39,720
Speaker 4: betting derivatives based on that sample. I think you can

48
00:02:39,760 --> 00:02:44,000
Speaker 4: also create mental models as well. You know, sports betting

49
00:02:44,280 --> 00:02:46,840
Speaker 4: is as much of an art as it is as science,

50
00:02:47,600 --> 00:02:51,040
Speaker 4: and so one thing I've done been working on recently

51
00:02:51,440 --> 00:02:54,720
Speaker 4: is coming up with my own NFL team ratings. So

52
00:02:54,760 --> 00:02:58,760
Speaker 4: I'm going and grading each unit within a team, applying

53
00:02:58,919 --> 00:03:03,240
Speaker 4: a specific way to it, and then creating a final

54
00:03:03,320 --> 00:03:07,440
Speaker 4: team grade, which I then convert into a projected spread

55
00:03:07,520 --> 00:03:14,000
Speaker 4: against sort of a neutral team. So it's not, you know, again,

56
00:03:14,040 --> 00:03:17,320
Speaker 4: a regression model or anything like that. I'm using my

57
00:03:17,440 --> 00:03:21,520
Speaker 4: domain knowledge, some other sources of information, historical data, things

58
00:03:21,560 --> 00:03:25,720
Speaker 4: like that, but using it to create an output that

59
00:03:25,880 --> 00:03:29,200
Speaker 4: can eventually be used for handicapping games.

60
00:03:29,960 --> 00:03:32,440
Speaker 1: Okay, and Sam, there's a lot of different variables that

61
00:03:32,480 --> 00:03:34,880
Speaker 1: go into sports betting. In your model, how do you

62
00:03:34,920 --> 00:03:37,240
Speaker 1: decide what do you put in, what do you throw out?

63
00:03:37,440 --> 00:03:40,000
Speaker 2: What kind of brings that that good bull of stew Altogether?

64
00:03:40,040 --> 00:03:44,120
Speaker 4: For you, it's a couple of things. Obviously, the most

65
00:03:44,440 --> 00:03:48,520
Speaker 4: important thing if you are building you a linear regression

66
00:03:48,600 --> 00:03:52,960
Speaker 4: or a classification model is doing exploratory data analysis, So

67
00:03:53,120 --> 00:03:57,280
Speaker 4: understanding sort of the shape of the data you're working with,

68
00:03:57,400 --> 00:04:01,480
Speaker 4: it's characteristics, all that sort of stuff. Seeing what again

69
00:04:01,560 --> 00:04:06,280
Speaker 4: correlates with your target variable. Doing all of that is

70
00:04:07,080 --> 00:04:11,200
Speaker 4: crucial in understanding again what's going to contribute to the

71
00:04:11,280 --> 00:04:14,560
Speaker 4: results of your model at the end. And then another

72
00:04:14,600 --> 00:04:17,760
Speaker 4: part of it too is just domain knowledge and knowing

73
00:04:18,400 --> 00:04:21,800
Speaker 4: you know when certain data points are going to be

74
00:04:22,160 --> 00:04:24,640
Speaker 4: a little bit off, or you know, if you're coming

75
00:04:24,720 --> 00:04:29,800
Speaker 4: up with interaction variables, understanding how those work together as

76
00:04:29,839 --> 00:04:33,640
Speaker 4: well is huge and you get that from having domain

77
00:04:33,720 --> 00:04:38,320
Speaker 4: knowledge in whatever sport you're working on. So again there's

78
00:04:37,920 --> 00:04:41,600
Speaker 4: the technical side of things of you know, looking at

79
00:04:41,640 --> 00:04:45,919
Speaker 4: the data you're using, but also understanding you know, what

80
00:04:46,040 --> 00:04:48,279
Speaker 4: parts of a sport are going to be captured best

81
00:04:48,320 --> 00:04:49,560
Speaker 4: in data and things like that.

82
00:04:50,240 --> 00:04:52,000
Speaker 2: Okay, awesome, glad to hear it.

83
00:04:52,040 --> 00:04:54,840
Speaker 1: And before we dive deeper into the sports betting modeling,

84
00:04:54,880 --> 00:04:57,520
Speaker 1: I want to remind everyone about betting pros betting systems.

85
00:04:57,560 --> 00:05:00,719
Speaker 1: Betting pros Betting systems are designed to help you winning

86
00:05:00,800 --> 00:05:04,240
Speaker 1: trends and make smarter beds, creat and customize your system

87
00:05:04,279 --> 00:05:07,359
Speaker 1: by sport, bet type, timeframe, and more to find the

88
00:05:07,360 --> 00:05:11,279
Speaker 1: most profitable betting opportunities for you follow systems to track

89
00:05:11,320 --> 00:05:14,560
Speaker 1: profitability and tail upcoming beds are system track profitabilities for

90
00:05:14,600 --> 00:05:17,159
Speaker 1: both the NFL and the NBA, as well as the

91
00:05:17,240 --> 00:05:19,839
Speaker 1: MLB and college football. So download and use the Betting

92
00:05:19,839 --> 00:05:22,320
Speaker 1: Pros app today to find the most profitable betting systems

93
00:05:22,320 --> 00:05:22,680
Speaker 1: for you.

94
00:05:23,080 --> 00:05:24,560
Speaker 2: Now, all right, and.

95
00:05:24,560 --> 00:05:26,640
Speaker 1: I want to get into building the foundations of your

96
00:05:26,640 --> 00:05:29,440
Speaker 1: model here, Sam, What tools are software do you generally

97
00:05:29,480 --> 00:05:32,479
Speaker 1: try to use when building your specific NFL betting.

98
00:05:32,200 --> 00:05:36,520
Speaker 4: Model, so it can be a wide range of things.

99
00:05:36,560 --> 00:05:41,080
Speaker 4: I tend to use different coding languages, whether it's Our

100
00:05:41,520 --> 00:05:46,680
Speaker 4: or Python. Certainly possible to make a sports betting model

101
00:05:46,760 --> 00:05:50,920
Speaker 4: in Excel as well. There are tons of resources out

102
00:05:50,960 --> 00:05:56,080
Speaker 4: there that cover using both Excel or these different coding languages.

103
00:05:56,440 --> 00:06:00,240
Speaker 4: You know. Obviously there's another big aspect of things that

104
00:06:00,400 --> 00:06:03,840
Speaker 4: is trending right now, which is AI. Whether it's using

105
00:06:03,880 --> 00:06:07,159
Speaker 4: something like chat, GPT or things like that. And I

106
00:06:07,200 --> 00:06:12,000
Speaker 4: will say they are useful, but I would caution when

107
00:06:12,040 --> 00:06:14,000
Speaker 4: you are using it, you know, don't just enter a

108
00:06:14,839 --> 00:06:19,040
Speaker 4: prompt of build me an NFL sides and TOIL model

109
00:06:19,160 --> 00:06:23,159
Speaker 4: like it's not going to give you the results you need.

110
00:06:23,920 --> 00:06:26,559
Speaker 4: You also need to know sort of what you're doing

111
00:06:26,600 --> 00:06:32,600
Speaker 4: when you're asking these AI assistants for different help, because

112
00:06:33,279 --> 00:06:36,320
Speaker 4: they might spit you wrong information and you need to

113
00:06:36,320 --> 00:06:39,720
Speaker 4: be able to know when when to sift through that.

114
00:06:40,920 --> 00:06:43,679
Speaker 2: Okay, so a I'm hearing AI, I'm hearing coding.

115
00:06:44,440 --> 00:06:48,159
Speaker 1: Is there anything that's free more accessible to people that

116
00:06:48,520 --> 00:06:51,120
Speaker 1: might not want to invest heavily into software like that?

117
00:06:51,200 --> 00:06:51,640
Speaker 2: Quite yet?

118
00:06:51,680 --> 00:06:54,119
Speaker 1: Is there anything simply you can do? Can you use Excel?

119
00:06:54,200 --> 00:06:56,080
Speaker 1: Can you use different spreadsheets things like that?

120
00:06:56,440 --> 00:07:01,239
Speaker 4: Yeah, you certainly can. And Excel has a solver add

121
00:07:01,279 --> 00:07:05,880
Speaker 4: on that I believe is free and it'll create different

122
00:07:05,960 --> 00:07:10,239
Speaker 4: regression models for you as well. You can do that. Again,

123
00:07:10,280 --> 00:07:13,120
Speaker 4: I think Excel is good for some of those quote

124
00:07:13,160 --> 00:07:15,680
Speaker 4: unquote mental models, if you will, of just putting a

125
00:07:15,720 --> 00:07:19,120
Speaker 4: lot of data into one place and using things that way.

126
00:07:19,200 --> 00:07:22,560
Speaker 4: So there's a time and place for all of this,

127
00:07:22,720 --> 00:07:26,240
Speaker 4: for sure, And I think if you're not a coding expert,

128
00:07:26,360 --> 00:07:28,800
Speaker 4: don't have a lot of experience, Excel is a good

129
00:07:28,800 --> 00:07:29,440
Speaker 4: place to start.

130
00:07:30,200 --> 00:07:32,840
Speaker 2: How do we sort through all the bs and really

131
00:07:32,880 --> 00:07:35,440
Speaker 2: filter that? Like, how are you collecting it? How are

132
00:07:35,440 --> 00:07:36,080
Speaker 2: you managing it?

133
00:07:36,160 --> 00:07:38,320
Speaker 1: Is it a different software we're using or is it

134
00:07:38,400 --> 00:07:39,960
Speaker 1: just Hey, you got to put the time in it

135
00:07:40,120 --> 00:07:42,120
Speaker 1: and you know, really capture that data.

136
00:07:42,360 --> 00:07:44,480
Speaker 4: Yeah, Like you said, we are in an age where

137
00:07:44,840 --> 00:07:49,440
Speaker 4: data is more available than ever, and it takes a

138
00:07:49,440 --> 00:07:52,720
Speaker 4: lot of work to you know, understand what is out

139
00:07:52,760 --> 00:07:56,920
Speaker 4: there and what's accurate and what you might need. You know,

140
00:07:56,960 --> 00:08:00,000
Speaker 4: obviously looking in the right places. Google is a very

141
00:08:00,320 --> 00:08:04,559
Speaker 4: helpful tool to figure out where all that is. Again,

142
00:08:04,640 --> 00:08:08,360
Speaker 4: I use a combination of both coding languages that I

143
00:08:08,520 --> 00:08:11,760
Speaker 4: mentioned for my data cleaning. It just helps speed it

144
00:08:11,880 --> 00:08:15,640
Speaker 4: up and make it more efficient. But again Excel is

145
00:08:16,200 --> 00:08:20,440
Speaker 4: a wonderful tool for that as well. You know, I

146
00:08:20,480 --> 00:08:23,720
Speaker 4: think it's the big part of it is just understanding

147
00:08:23,760 --> 00:08:27,480
Speaker 4: the data that you're working with for sure, okay.

148
00:08:27,640 --> 00:08:29,520
Speaker 1: And kind of on that same note of a lot

149
00:08:29,560 --> 00:08:31,640
Speaker 1: of noise being out there right now, what are some

150
00:08:31,680 --> 00:08:35,200
Speaker 1: common mistakes that people you think first make when building models.

151
00:08:35,720 --> 00:08:37,720
Speaker 1: Is there data you think they're factoring that maybe they

152
00:08:37,760 --> 00:08:39,839
Speaker 1: shouldn't be, or data maybe they're missing out on.

153
00:08:40,480 --> 00:08:44,360
Speaker 4: I think a big mistake and one that I've learned

154
00:08:44,360 --> 00:08:46,600
Speaker 4: a lot of lessons from is assuming that the data

155
00:08:46,640 --> 00:08:52,880
Speaker 4: that you're getting is correct, okay, and understanding how different

156
00:08:52,920 --> 00:08:56,560
Speaker 4: stats are calculated. Now, rushing yards, passing yards, like those

157
00:08:56,600 --> 00:09:00,840
Speaker 4: are pretty cut and dry with you know how accurate

158
00:09:00,920 --> 00:09:03,200
Speaker 4: they are. But you know, I've had some I had

159
00:09:03,200 --> 00:09:05,560
Speaker 4: some discussions with someone a couple of weeks ago about

160
00:09:05,679 --> 00:09:08,760
Speaker 4: air yards in the NFL, which is the distance that

161
00:09:09,240 --> 00:09:12,679
Speaker 4: the ball travels from the line of scrimmage to the

162
00:09:13,280 --> 00:09:16,760
Speaker 4: catchpoint or target point. And a bunch of sites have

163
00:09:16,800 --> 00:09:20,480
Speaker 4: a bunch of different numbers for that, depending on you know,

164
00:09:20,520 --> 00:09:24,440
Speaker 4: whether they include plays with penalties that don't actually count

165
00:09:24,480 --> 00:09:28,079
Speaker 4: as a play or things like that. So doing some

166
00:09:28,160 --> 00:09:32,280
Speaker 4: sense checking of your data is huge. Cleaning data in

167
00:09:32,760 --> 00:09:37,000
Speaker 4: any model is going to be paramount most of your time.

168
00:09:37,400 --> 00:09:39,360
Speaker 4: It's it's not super fun, but most of your time

169
00:09:39,440 --> 00:09:42,040
Speaker 4: is going to be spent on on data collection and

170
00:09:42,160 --> 00:09:46,080
Speaker 4: data cleaning. Again, no two sources are going to use

171
00:09:46,679 --> 00:09:50,480
Speaker 4: the same nomenclature either. I mean, if you try try

172
00:09:50,559 --> 00:09:55,800
Speaker 4: merging NFL team abbreviations, some have it as HOU for Houston,

173
00:09:55,960 --> 00:10:01,880
Speaker 4: some have it as HST. So understanding reasons for that.

174
00:10:02,559 --> 00:10:05,839
Speaker 4: And again a big part of model building is if

175
00:10:05,880 --> 00:10:08,199
Speaker 4: you put crap in, you're going to get crap out.

176
00:10:08,320 --> 00:10:11,360
Speaker 4: So if the data you're putting in is not good,

177
00:10:11,400 --> 00:10:14,240
Speaker 4: it's not accurate. The results of your model are not

178
00:10:14,280 --> 00:10:16,000
Speaker 4: going to be good or accurate either.

179
00:10:16,720 --> 00:10:18,679
Speaker 1: Kind of sounds like eating fast food a little bit,

180
00:10:18,720 --> 00:10:20,439
Speaker 1: you know, when you're on the road, when you're traveling,

181
00:10:20,600 --> 00:10:22,400
Speaker 1: sometimes you got to bite the bullet, maybe get a

182
00:10:22,400 --> 00:10:25,160
Speaker 1: little nick danks and you and might not feel great

183
00:10:25,200 --> 00:10:27,880
Speaker 1: the next day. So I like that analogy from you there, Sam.

184
00:10:28,000 --> 00:10:29,640
Speaker 1: Let's keep it moving here, But first I want to

185
00:10:29,640 --> 00:10:31,840
Speaker 1: give everyone a quick note that if you're looking to

186
00:10:31,840 --> 00:10:34,920
Speaker 1: score a signed Amari Cooper jersey for free courtesy for

187
00:10:34,920 --> 00:10:37,360
Speaker 1: our friends over at Pristine Auction dot com, all you

188
00:10:37,440 --> 00:10:40,240
Speaker 1: gotta do is head over to Bettingpros dot com slash contest,

189
00:10:40,480 --> 00:10:42,960
Speaker 1: complete the form and either a download the Betting Pros

190
00:10:42,960 --> 00:10:45,400
Speaker 1: app b leave a review for the Betting Pros podcast

191
00:10:45,480 --> 00:10:48,160
Speaker 1: this podcast, or subscribe to our social media channels at

192
00:10:48,160 --> 00:10:51,199
Speaker 1: Betting Pros on x and TikTok or at Betting Pros

193
00:10:51,320 --> 00:10:54,480
Speaker 1: NFL on Instagram. The more actions you complete, the more

194
00:10:54,600 --> 00:10:57,200
Speaker 1: entries you will receive, and we'll be announcing the winner

195
00:10:57,280 --> 00:10:59,280
Speaker 1: right here on the Betting Pros YouTube channel. So make

196
00:10:59,280 --> 00:11:01,200
Speaker 1: sure you turn on those certification so you can be

197
00:11:01,240 --> 00:11:03,200
Speaker 1: alerted when new episodes.

198
00:11:02,760 --> 00:11:04,920
Speaker 2: Are up and claim your prize.

199
00:11:05,360 --> 00:11:07,360
Speaker 1: All right, I want to get into more of the

200
00:11:07,400 --> 00:11:10,200
Speaker 1: thick here testing and tweaking your model, Sam, how do

201
00:11:10,200 --> 00:11:12,959
Speaker 1: you test your model engages accuracy? Are you kind of

202
00:11:13,000 --> 00:11:15,520
Speaker 1: sending it all in and just hey, week one, I

203
00:11:15,600 --> 00:11:17,599
Speaker 1: got to see what I have here and pushing the

204
00:11:17,679 --> 00:11:19,240
Speaker 1: chips in the middle, or are you just laying some

205
00:11:19,280 --> 00:11:22,200
Speaker 1: small unit wagers to kind of test that off the rip.

206
00:11:23,679 --> 00:11:26,440
Speaker 4: Yeah, So there are a number of ways you can

207
00:11:26,720 --> 00:11:29,880
Speaker 4: test a model. Obviously, back testing is a huge part

208
00:11:29,920 --> 00:11:32,840
Speaker 4: of any sports betting model. You have your sort of

209
00:11:32,840 --> 00:11:39,800
Speaker 4: classical excuse me, classic model technique. So for a classification model, accuracy, precision,

210
00:11:40,040 --> 00:11:43,320
Speaker 4: things like that. For a regression model, you've got our

211
00:11:43,360 --> 00:11:46,440
Speaker 4: square mean, absolute error, all those sorts of things. You

212
00:11:46,480 --> 00:11:50,240
Speaker 4: can can search more about that. But another big part

213
00:11:50,320 --> 00:11:55,600
Speaker 4: of a sports betting model is comparing it to historical odds.

214
00:11:55,640 --> 00:12:00,520
Speaker 4: So comparing you know, past predictions or or what your

215
00:12:00,559 --> 00:12:05,480
Speaker 4: model would have predicted against the lines and see if

216
00:12:05,480 --> 00:12:10,920
Speaker 4: it's a profitable model over a large enough sample. If

217
00:12:11,320 --> 00:12:15,760
Speaker 4: if your model predicts something at an eighty percent accuracy rate,

218
00:12:15,960 --> 00:12:19,080
Speaker 4: but the sports book predicts an eighty five percent, it's

219
00:12:19,120 --> 00:12:22,720
Speaker 4: not going to be beneficial as beneficial to you and

220
00:12:22,800 --> 00:12:26,280
Speaker 4: it'll help you identify if you can overcome the big

221
00:12:26,559 --> 00:12:31,040
Speaker 4: in certain markets. I think if you don't have past odds.

222
00:12:31,080 --> 00:12:34,320
Speaker 4: You know, some sports like the NFL have a litany

223
00:12:34,480 --> 00:12:40,920
Speaker 4: of sports betting odds available for historical games, but some

224
00:12:41,120 --> 00:12:46,240
Speaker 4: like the WNBA or smaller niche sports aren't as really available.

225
00:12:46,320 --> 00:12:49,720
Speaker 4: So if you don't have past odds, I think paper

226
00:12:49,760 --> 00:12:52,800
Speaker 4: trading is a good way to do that. And what

227
00:12:52,840 --> 00:12:56,200
Speaker 4: I mean by that is just writing down bets that

228
00:12:56,280 --> 00:13:01,319
Speaker 4: your model would theoretically make in certain games and situations,

229
00:13:01,320 --> 00:13:03,760
Speaker 4: don't actually have to put a wager on it, and

230
00:13:03,800 --> 00:13:07,160
Speaker 4: then track that. Once you get sort of a large

231
00:13:07,240 --> 00:13:10,920
Speaker 4: enough sample I would say, depending on the market, probably

232
00:13:10,960 --> 00:13:13,360
Speaker 4: at least one hundred bets I think is a good sample,

233
00:13:14,400 --> 00:13:17,199
Speaker 4: then you can understand again if you truly have an

234
00:13:17,360 --> 00:13:20,800
Speaker 4: edge in that market with the model your betting. So

235
00:13:21,440 --> 00:13:24,520
Speaker 4: again it's not fun because if if your model wins

236
00:13:24,520 --> 00:13:27,079
Speaker 4: all one hundred bets, it's it's not ideal. But you

237
00:13:27,440 --> 00:13:30,800
Speaker 4: want to make sure before you get super deep into

238
00:13:31,760 --> 00:13:35,360
Speaker 4: risking a ton of units, to confirm that your model

239
00:13:35,360 --> 00:13:37,120
Speaker 4: has a bit of an edge. I like that.

240
00:13:37,200 --> 00:13:39,200
Speaker 1: I like that, you know, kind of putting your toes

241
00:13:39,240 --> 00:13:41,240
Speaker 1: in the water, testing it out. Before you go full

242
00:13:41,280 --> 00:13:44,040
Speaker 1: send on it. I like that idea, Sam, Especially in

243
00:13:44,080 --> 00:13:46,559
Speaker 1: the NFL, things are getting a little wacky at times.

244
00:13:46,559 --> 00:13:49,360
Speaker 1: I mean, we just saw the Hall of Fame game

245
00:13:49,480 --> 00:13:51,200
Speaker 1: end in the third quarter due to weather a couple

246
00:13:51,200 --> 00:13:53,520
Speaker 1: of weeks ago. How do you count for like just

247
00:13:53,600 --> 00:13:57,320
Speaker 1: crazy variables, unexpected things happening here in the world of

248
00:13:57,400 --> 00:13:58,120
Speaker 1: NFL gambling.

249
00:13:59,720 --> 00:14:03,400
Speaker 4: Yeah, this is part of the handicapping process, right. It's

250
00:14:03,720 --> 00:14:08,680
Speaker 4: understanding what parts of a market or projection your model

251
00:14:09,320 --> 00:14:12,520
Speaker 4: doesn't account for. I know we're talking about the NFL here,

252
00:14:12,600 --> 00:14:16,200
Speaker 4: but to borrow an example from some of the Formula

253
00:14:16,200 --> 00:14:19,640
Speaker 4: one modeling I was doing, you know, each between races,

254
00:14:19,720 --> 00:14:24,600
Speaker 4: cars can make certain upgrades, and that's not you know,

255
00:14:24,880 --> 00:14:28,360
Speaker 4: sure you could go back and collect a lot of

256
00:14:28,360 --> 00:14:31,640
Speaker 4: that data. It's not readily available, but you know that's

257
00:14:31,640 --> 00:14:34,800
Speaker 4: not necessarily a feature that you might want to include

258
00:14:34,880 --> 00:14:37,880
Speaker 4: in a model or know how to code into a model.

259
00:14:38,000 --> 00:14:41,440
Speaker 4: So you know, your projection might have a certain driver

260
00:14:41,640 --> 00:14:47,040
Speaker 4: or team better than expected, but maybe they also made

261
00:14:47,120 --> 00:14:49,640
Speaker 4: an upgrade to the car something like that. In the NFL,

262
00:14:50,800 --> 00:14:55,200
Speaker 4: maybe it's weather, maybe it's the team made a change

263
00:14:55,440 --> 00:15:00,120
Speaker 4: of who the play color is. In understanding it and

264
00:15:00,680 --> 00:15:02,880
Speaker 4: your model is not necessarily going to capture that, but

265
00:15:03,040 --> 00:15:07,840
Speaker 4: understanding what impact that might have on the e output

266
00:15:07,840 --> 00:15:11,720
Speaker 4: of your model. Okay, I think again that's that's the

267
00:15:11,800 --> 00:15:15,440
Speaker 4: part of the the art of sports betting is is

268
00:15:15,560 --> 00:15:19,560
Speaker 4: understanding what goes into your model, what's captured, and then

269
00:15:20,160 --> 00:15:24,040
Speaker 4: taking the other things and applying it in the handicapping process.

270
00:15:24,240 --> 00:15:26,280
Speaker 1: Yeah, it's funny and f one, you can't account for

271
00:15:26,840 --> 00:15:28,960
Speaker 1: some drivers who just don't want to don't want to

272
00:15:29,000 --> 00:15:31,280
Speaker 1: win the race and have to pit because their team

273
00:15:31,320 --> 00:15:32,840
Speaker 1: is yelling at them, just like we saw a couple

274
00:15:32,880 --> 00:15:35,600
Speaker 1: of weeks ago with Lando Norris and that that squad

275
00:15:35,640 --> 00:15:39,320
Speaker 1: over there. Some final tips here for the sports gamblers

276
00:15:39,320 --> 00:15:42,160
Speaker 1: out there, Sam, how do you update your model or

277
00:15:42,160 --> 00:15:44,360
Speaker 1: how often do you update your model on based on

278
00:15:44,440 --> 00:15:46,600
Speaker 1: new data and trends and how do you best spot

279
00:15:46,640 --> 00:15:50,000
Speaker 1: those trends as well? Throughout the NFL season, which again

280
00:15:50,120 --> 00:15:52,840
Speaker 1: there's things coming at us left and right. Things are changing,

281
00:15:52,880 --> 00:15:54,480
Speaker 1: So do you have time to kind of go back

282
00:15:54,520 --> 00:15:56,600
Speaker 1: and tweak that throughout or is it like, hey, by

283
00:15:56,600 --> 00:15:59,200
Speaker 1: week one, if my model isn't set, you know, I

284
00:15:59,280 --> 00:16:00,320
Speaker 1: just kind of have to roll with it.

285
00:16:02,080 --> 00:16:07,160
Speaker 4: Yeah. I think between seasons for sports is sort of

286
00:16:07,200 --> 00:16:11,760
Speaker 4: a good benchmark for updating and retraining a model. It

287
00:16:11,840 --> 00:16:15,600
Speaker 4: sort of depends on the sport too. You know, with

288
00:16:15,720 --> 00:16:18,680
Speaker 4: the NFL this season, we're going to see a different

289
00:16:18,720 --> 00:16:22,800
Speaker 4: type of kickoff, and while that may not have a

290
00:16:22,920 --> 00:16:26,520
Speaker 4: massive impact, it's going to have some impact on you know,

291
00:16:26,600 --> 00:16:32,000
Speaker 4: different models, depending on again what you're projecting. But I

292
00:16:32,040 --> 00:16:36,000
Speaker 4: think again, understanding again what your your model is capturing.

293
00:16:36,040 --> 00:16:38,840
Speaker 4: Like if you're not simulating play by play for the

294
00:16:39,000 --> 00:16:44,840
Speaker 4: NFL and simulating the kickoffs specifically, like you need to

295
00:16:44,960 --> 00:16:48,360
Speaker 4: understand how how that works. I think it's it's also

296
00:16:48,440 --> 00:16:52,520
Speaker 4: important not only like how often you set it up,

297
00:16:52,520 --> 00:16:54,080
Speaker 4: but the way that you set it up. You know,

298
00:16:54,600 --> 00:16:57,800
Speaker 4: when you're going through the initial build of a model,

299
00:16:57,840 --> 00:17:02,080
Speaker 4: setting it up so that the code or the process

300
00:17:02,200 --> 00:17:04,800
Speaker 4: is done in a way that it's simple to retrain it.

301
00:17:04,840 --> 00:17:06,320
Speaker 4: You don't want to have to start from scratch at

302
00:17:06,400 --> 00:17:08,679
Speaker 4: right a single time you want to retrain it. You

303
00:17:08,720 --> 00:17:11,480
Speaker 4: want to properly label all the data that you're working

304
00:17:11,520 --> 00:17:14,240
Speaker 4: with and things like that, so it's not a huge

305
00:17:14,280 --> 00:17:15,879
Speaker 4: burden to retrain it.

306
00:17:16,240 --> 00:17:18,359
Speaker 1: And what advice would you give to someone Sam who's

307
00:17:18,400 --> 00:17:20,960
Speaker 1: just kind of starting out here in sports betting modeling.

308
00:17:21,240 --> 00:17:23,080
Speaker 2: Is there any does and do not you have? Kind

309
00:17:23,080 --> 00:17:24,360
Speaker 2: of right off the top of the dome.

310
00:17:24,640 --> 00:17:26,679
Speaker 4: I would say it's going to be a trial and

311
00:17:26,840 --> 00:17:30,959
Speaker 4: error process sort of seeing what works. I think, you know,

312
00:17:31,000 --> 00:17:35,560
Speaker 4: there is some debate as to whether it's you know,

313
00:17:35,680 --> 00:17:39,240
Speaker 4: better to focus on a sport that you know or not,

314
00:17:39,480 --> 00:17:42,920
Speaker 4: because if you're familiar with the NFL, you're going to

315
00:17:43,000 --> 00:17:46,280
Speaker 4: have some inherent biases on oh yeah, what you think

316
00:17:46,680 --> 00:17:48,840
Speaker 4: should be included in a model or what shouldn't And

317
00:17:49,520 --> 00:17:53,080
Speaker 4: letting the data speak to you is a huge part

318
00:17:53,119 --> 00:17:56,560
Speaker 4: of that, but it also will help you spot potential

319
00:17:56,680 --> 00:18:00,360
Speaker 4: errors in the data that you're working with, So there's

320
00:18:00,359 --> 00:18:03,639
Speaker 4: a little bit of a balance there. I think understanding

321
00:18:03,680 --> 00:18:07,520
Speaker 4: too that projections are just projections, like that's it. They're

322
00:18:07,560 --> 00:18:11,520
Speaker 4: not picks per se. It takes more than that. I

323
00:18:11,560 --> 00:18:15,560
Speaker 4: know I've mentioned this a bunch of times, but I

324
00:18:15,560 --> 00:18:19,919
Speaker 4: would never advocate for taking your projections or someone else's

325
00:18:19,920 --> 00:18:24,479
Speaker 4: projections and just blindly betting them by comparing them to

326
00:18:24,760 --> 00:18:27,560
Speaker 4: whatever the sports books lines are there are you know,

327
00:18:27,600 --> 00:18:29,679
Speaker 4: you're not going to be able to capture everything in

328
00:18:29,800 --> 00:18:35,280
Speaker 4: a sports betting model that the market in a sports

329
00:18:35,280 --> 00:18:38,520
Speaker 4: bet is going to capture as well, so you can

330
00:18:38,680 --> 00:18:42,440
Speaker 4: use those projections as sort of a guiding light into

331
00:18:43,080 --> 00:18:46,359
Speaker 4: where things might be going. And then again use the

332
00:18:46,400 --> 00:18:51,200
Speaker 4: handicapping of understanding what's not in your model to figure

333
00:18:51,240 --> 00:18:56,640
Speaker 4: out if what you aren't capturing is enough to sort

334
00:18:56,640 --> 00:19:00,399
Speaker 4: of overcome the big if you will, And then vastly

335
00:19:01,080 --> 00:19:04,880
Speaker 4: using simulations, you can you create sort of expected value

336
00:19:04,880 --> 00:19:08,360
Speaker 4: of a bet. Using that, I think that's super important

337
00:19:08,480 --> 00:19:14,000
Speaker 4: to track and making sure that you're really on the

338
00:19:14,080 --> 00:19:16,760
Speaker 4: right track and provides you sort of a range of

339
00:19:16,760 --> 00:19:19,640
Speaker 4: outcomes for the specific bet you're going to make.

340
00:19:20,080 --> 00:19:22,440
Speaker 1: Awesome, Awesome. I appreciate that advice. I'm sure the great

341
00:19:22,440 --> 00:19:24,560
Speaker 1: folks here at Betting Pros will as well. Last question,

342
00:19:24,560 --> 00:19:26,640
Speaker 1: I want to leave you with what communities or other

343
00:19:26,720 --> 00:19:29,160
Speaker 1: resources can you recommend for people who are just kind

344
00:19:29,160 --> 00:19:31,720
Speaker 1: of getting into the sports betting modeling world.

345
00:19:33,359 --> 00:19:36,240
Speaker 4: There are a couple of things. I think one underrated

346
00:19:36,600 --> 00:19:39,760
Speaker 4: aspect is like reading scholarly papers. And I know this

347
00:19:40,440 --> 00:19:44,119
Speaker 4: isn't college, this isn't not necessarily a fun part, but

348
00:19:44,520 --> 00:19:46,520
Speaker 4: there are a ton of published papers out there on

349
00:19:46,640 --> 00:19:50,560
Speaker 4: different sports topics, even if they're not betting related, that

350
00:19:51,160 --> 00:19:53,840
Speaker 4: will give you a sense of what has or hasn't

351
00:19:53,880 --> 00:19:58,840
Speaker 4: worked for other groups and give you maybe a starting

352
00:19:58,880 --> 00:20:01,760
Speaker 4: point too, or a different way of thinking about approaching

353
00:20:01,800 --> 00:20:07,720
Speaker 4: a specific problem. Again, I mentioned chat, GPT or AI earlier.

354
00:20:07,920 --> 00:20:10,159
Speaker 4: Is you know, being a little bit wary of it,

355
00:20:10,200 --> 00:20:13,720
Speaker 4: but it is a useful tool for helping with automation

356
00:20:13,880 --> 00:20:16,399
Speaker 4: and things like that. You know, I'm in the Betting

357
00:20:16,440 --> 00:20:19,000
Speaker 4: Pros discord and on Twitter as well, like, feel free

358
00:20:19,040 --> 00:20:22,119
Speaker 4: to reach out to me and ask questions of people

359
00:20:22,200 --> 00:20:25,879
Speaker 4: like It's it's daunting because there's a lot and some

360
00:20:25,920 --> 00:20:29,639
Speaker 4: people are not necessarily willing to share everything. So asking

361
00:20:29,680 --> 00:20:33,199
Speaker 4: the right questions is important because if someone has an

362
00:20:33,240 --> 00:20:36,200
Speaker 4: edge with a sports betting model, they're likely not going

363
00:20:36,240 --> 00:20:38,760
Speaker 4: to be sharing it because they want to keep that

364
00:20:38,880 --> 00:20:42,880
Speaker 4: edge to themselves. But if you can ask questions about

365
00:20:42,960 --> 00:20:47,360
Speaker 4: you know, feature engineering or things like that and not

366
00:20:47,880 --> 00:20:50,680
Speaker 4: you know, don't ask hey can you like build me

367
00:20:50,720 --> 00:20:52,600
Speaker 4: a model? Or hey, can you give me your output?

368
00:20:53,160 --> 00:20:55,879
Speaker 4: That's those are the sort of questions you want to avoid.

369
00:20:55,920 --> 00:20:58,880
Speaker 4: But you know, have someone have questions that will help

370
00:20:58,960 --> 00:21:01,760
Speaker 4: lead you to water so that you can further build

371
00:21:01,800 --> 00:21:02,440
Speaker 4: your model out.

372
00:21:02,600 --> 00:21:05,000
Speaker 1: Well, we appreciate you leaving us to water here today,

373
00:21:05,080 --> 00:21:07,040
Speaker 1: Sam and that's gonna be all for us. So we

374
00:21:07,080 --> 00:21:10,520
Speaker 1: hope everyone got a solid foundation into sports betting modeling.

375
00:21:10,560 --> 00:21:14,000
Speaker 1: In this video, he's a scientist at Sam Hoppin on Twitter,

376
00:21:14,040 --> 00:21:17,000
Speaker 1: I'm Seth Wilcock apt between Underscore seth FF on Twitter.

377
00:21:17,119 --> 00:21:19,440
Speaker 1: I'm just here so I don't get fine. Thanks for watching, y'all.

378
00:21:19,480 --> 00:21:20,400
Speaker 1: We'll catch you next time.

379
00:21:20,840 --> 00:21:23,440
Speaker 3: Thanks for listening to the Betting Pros podcast. If you

380
00:21:23,520 --> 00:21:25,879
Speaker 3: love the show, the best freeway to support us is

381
00:21:25,880 --> 00:21:29,280
Speaker 3: by leaving a positive review on Apple Podcasts or Spotify.

382
00:21:29,600 --> 00:21:32,199
Speaker 3: Follow us on x and TikTok at Betting Pros and

383
00:21:32,320 --> 00:21:35,960
Speaker 3: Instagram at Betting Pros NFL. Also subscribe to our YouTube

384
00:21:36,040 --> 00:21:38,520
Speaker 3: channel at YouTube dot com slash Betting Pros