Creating a Sports Betting Model 101 - Intro to Linear Regression (The simplest model ever created!)

[@Burps___]

Wow...this is a lot more work than just saying “ The Jets got rocked last weekend, so there’s.no way they’re winning this Sunday.” 😂 🏈 Thank you for your videos.

[@aidenward103]

Seeing this 11 months later, that probably wasnt a bad strategy lol

[@GoldGunsandGolf]

One thing I've learned about doing stats is how important strength of schedule is... stats are useless unless you know how hard or easy it was for a team to obtain those stats.

[@amyrg360]

Is there a fast way you got the past year's schedule into excel? or did you hand input?

[@kylejones3649]

you can export into excel from sports reference

[@zyzyvlad1942]

This works out, If you wanna bet right you ll need to have a high level of statistics and econometric knowledge

[@jschwets11]

Not necessarily true.

[@joshcolbert5613]

wrong

[@Garycarlyle]

Its actually not that hard. Some trades are actually quite obvious. Prices drift a lot for nonsensical reasons sometimes. Just like trading the financial markets.

[@TopFearless]

You for sure don’t need a high level understanding of statistics to follow something like this linear programming model, but if you want to actually build a model that can arbitragely make you money, of course you need to be one hell of a mathematician

[@bbsara0146]

They should teach horse gambling and sports betting in schools. young people will learn all their stats and math without even realizing it because they are having fun gambling on horses

[@mariellezarraga1467]

Hi William, I´m looking at your video but don´t understand in minute 3 where did your get away and home from? is not on the website you´re showing. Could you help me please?

[@briancox6926]

Love the content - quick question, where can I pull all results for college bball games this season? Sports-reference doesnt seem to have that. Thanks in advance!

[@williamleiss4]

You're going to have to get creative with more advanced scraping methods. Data is the hardest part about all this, and its something you have to keep secret. If you go around telling everyone where you get your data, your data sources will enact anti scraping measures, forcing you to look elsewhere.

[@sueannjones4524]

printing money from sportsmodelanalytics dot com picks

[@khizarabid8414]

Many thanks, I've been looking for "how does plus minus work in betting" for a while now, and I think this has helped. Have you heard people talk about - Keton Jenackenzie Equalizer - (do a google search ) ? Ive heard some amazing things about it and my m8 got great success with it.

[@fizzfox8886]

i came for betting and stayed for excel

[@edgarz7330]

There are a few components to sports picks. One plan I discovered that successfully combines these is the Simple Sports Goldmine (google it if you're interested) definately the no.1 system that I have ever heard of.Check out all the incredible info .

[@majestyd3015]

I guess its part of the game

[@Garycarlyle]

Do you have a sheet with the table headings definitions you can share, please?. I want to use this for a college project in the UK but I'm English and I don't understand those references I get the math and Excel however.

[@jessesviridoff5942]

How did you find the APF/HPF?

[@ThermalWalnut09]

Will this work for current schedule? And fill automatically as games are played per week

[@ryanclarke2948]

How did you get pro football reference to get you scores for Home and Away, for me it only has score for winning team vs losing team?

[@1Kayhan]

Great video I just have a question. I used a few stats for this and when I did the regression on them individually they all had very low P values, but when I did them all together they got very large. How do I interpret that?

[@jaimenenclares6953]

there might be some autocorrelation between the variables, the more variables there are the more noise there is

[@bailey6143]

The lower the pvalue and the higher the r squared is the better. I don’t like my pvalue higher then .05 and my average is always around the 99.9999% mark

[@gazza4770]

did u do that manual or is sit an easy way to do that? 2:44

[@TheMotorcycleboy25]

Hello SBT. I have question: how many independent variables do you usually has in your models? How high should be your R square to start test model? Best regards, thanks!

[@williamleiss4]

My college basketball model uses 7 and has an R Squared of 50.

[@TheMotorcycleboy25]

@@williamleiss4 Thank you. R squared doesnt have value from 0 to 1? Do you mean 0,5? What is minimum in your opinion to treat model serious?

[@jublitz72]

I’d also like to know pls ^

[@Joe-cy6cl]

Hey great content I appreciate your help and time. One thing maybe you could help me with. I get the ASRS perfectly fine, but the HSRS I am getting N/A for a few of the teams. When I go to my schedule I copy that home teams name and replace it with the same exact name in my SRS tab and now the HSRS is coming up, but now the ASRS is getting N/A for that team. I dont get it because it's spelled the exact same for the SRS tab, Home column and Away column. Tried troubleshooting myself, but can not find anything. Thanks!

[@Joe-cy6cl]

EDIT: I figured it out was a silly mistake. For anyone having the same problem look at the rows with the outcome N/A and check to see if there are any extra spaces at the end of the team name for whatever team name is coming up N/A. Make sure the spelling is the same as well in your SRS tab and also make sure there are no extra spaces.

[@ethanschwab9670]

How would you enter a statistic like defensive yards allowed, where a lower number is better than a higher number? Do I make the numbers all negative?

[@derekroberts7196]

Which stats do you like to look at for NFL? I'm getting good p and r values, but I cant quite get to better than a .5 intercept.

[@williamleiss4]

Unfortuantely I have never built out an NFL model I have used to track or bet with. I believe NFL is a bit too hard to beat even with a model. With that being said the only number you should try to get as high as possible is the R-Squared value. If R-Squared is high, the P-Values(that matter) will be high also. It's not the end of the world if the intercept isnt in line with home advantage - the intercept is simply the starting point of the trend plot. As far as NFL stats I would look at were I to build a model, it would probably be drive efficiency, success rate, and offensive line metrics like adjusted line yards. Football Outsiders has some stats you can use, if not you can try and create them via play by play.

[@derekroberts7196]

@@williamleiss4 The highest I've gotten it is .52 r-value over about 8 independent variables - but the intercept was around -11. Only a few of the p values were under .05 though so that had me worried. I have used a couple of FO stats but also using some of the play by play stuff from warren sharps stats. i did not try offensive or defensive line metrics but that is a great idea!

[@willhartley7631]

@@derekroberts7196 what did you use to get yours all the way to .52? cant seem to get mine above .45

[@chikomufc]

Thanks for the video and especially the website!! Everything I need to start building some models

[@milesmcdonald8132]

For all of my data analysis, I keep getting a positive intercept even tho I am using the same stats and everything as you. Does this mean that the away team has a ≈2 point advantage

[@ianharkins2172]

How would something that is not a team specific metric, such as weather, be input into something like this?

[@Iamiami409]

do u mlb or nba model. or just football?

[@Dimensionaut_]

excellent video. even tho it's linear regression and basic in concept, this is still a bit difficult to put together. do you have any models available for download online or something? maybe something more advanced? i keep track of my bets in excel and have analytics, but don't have a model to make my picks

[@williamleiss4]

I do not have anything available for download, my own models are secret and the only thing I make public is whats in these videos

[@kiltaa]

Hey, would you manage to create a model for A Virtual Sports betting site? The games are played after every 3 minutes(8 games/ 16 teams). I'm guessing it is based on a certain algorithm

[@jesuscolina2366]

I've been profiting like crazy off sportsmodelanalytics.com and their advanced forecast models. the best 24 bucks I ever spent

[@kylecool7022]

Could you make a follow up video just explaining how to apply our model week to week in the nfl for example?

[@daviddowell7729]

Great video man

[@alejandrobotran4068]

TCU!!!!!! yeah good memories from Ft Worth

[@de0o0]

What if there is no any "rating" for teams, like how do u get which team is better on paper from another. I just simply have matches history and statistics about them

[@AlexA-nr4bj]

I did this for College football with the HSRS and ASRS and no line fit plot was outputted for HSRS. However all the numbers I got looked amazing. R_squared was .71, Intercept was -2.98, P-Values and Significance F were all low. Am I fine?

[@williamleiss4]

Those are good values and there is nothing wrong with using SRS as a component but to base an entire model off of a widely available public statistic is going to be -EV in the long run. I'd find more statistics to pair it with

[@joshpenney6544]

My model is coming back with ridiculous numbers...I double checked the stats were inputted correctly. Can’t figure out the problem. Any ideas? Also my p-values look really bad for stats that should be more predictive

[@williamleiss4]

I would need more context to be able to help. Also if you aren't opponent adjusting and neutralizing your stats, they wont be predictive. SRS(the stat I used in this video) is an opponent adjusted/neutralized stat

[@kippkoren5476]

Hi, I tried analyzing some other stats, like the following listed below, and the intercept was '-42' , would I not use this value then when multiplying and adding coefficients ? I get a really large number. Away_Def_SRS , Home_Def_SRS, Away_Points_Allowed, Home_Points_Allowed

[@bailey6143]

Intercept means nothing really

[@sax7q4]

What is the best place to go to easily import mens college basketball schedule into excel?

[@jahsehonfroy8792]

Same question ^

[@jordanfrkovich6849]

Please help. I am trying to make a model for the Bundesliga. I have plugged in soooo many stats/advanced stats and can't find anything with an R-Squared over 25%. I have 4 years worth of games, and my dependent variable is the margin. Is the margin harder to predict? Is that why all of the stats I am using are 17 to 25%? I am running out of ideas.

[@williamleiss4]

Wish I could help you on soccer but I never watch soccer unless its the US men in the World Cup. I know nothing about the sport.

[@jonathonsawyer3470]

If I use this formula for a bunch of different stats and continue to put scores into the sheet will the different stats change throughout the season from the games I’ve entered? For example I’m doing it for NHL and if I continue entering scores each day will things like goals against per game change as I enter more and more scores?

[@williamleiss4]

Yes, the scores will change

[@adammoore5980]

This may be a dumb question but how did you pull the scores in home/away format

[@williamleiss4]

It's something you have to rig yourself in excel.

[@arlenestanton9955]

At pro- football reference.com does not list, team schedules, away vs home. This make it difficult to copy and paste. Any suggestions how to format this?

[@williamleiss4]

You have to do it yourself on the backend. I use a couple of excel formuals to do it.

[@jc3394]

Thank you ive subscribed and will be watching some more videos thanks again

[@tommyproductions891]

Great video but that’s not at all what the R^2 value means. It’s a measure of how close a line is, not quite to do with any percentage of datapoints “explained” by the line.

[@TexasRangersReportbyCFM]

Is that a Vincero watch?

[@williamleiss4]

Yes

[@tywes23]

I did everything the same way you did but just for NBA, ran this equation and got -2.25425 for Clippers @ Suns. Does the - in front mean the Clippers are 2.25 favorites or underdogs?

[@williamleiss4]

Underdogs

[@j.p.marcos6417]

Where did you find the home/away schedule? Can’t seem to find it anywhere

[@williamleiss4]

Sports Reference but you have to manipulate it yourself within excel

[@j.p.marcos6417]

@@williamleiss4 ah ok got it thanks for the great content

[@Il-matic]

Hello where do you find all games played schedule in Excel ? What web site please ?

[@williamleiss4]

Sports Reference is a good one

[@calebmynatt6946]

If you wouldn't mind explaining, why do your values (such as p-value) change for SRS between the 2 examples? You obviously added more variables in the second example, but why does that change the values of the H and A SRS in that example? Is the same stat not being measured? Very new to this all, so thank you for any help you can provide.

[@dadamnicholson]

The p-value is the probability that that outcome is the result of random chance. Each parameter will have a different p-value and it will change if you add or remove other parameters from the equation. The accepted threshold is typically p < .05, which means there's less than a 5% probability that the outcome is due to random chance. Hope that helps!

[@dadamnicholson]

Also, the coefficients will change when you add other parameters, because now they're working together and may explain more or less of the outcome. For example, if you have age in a model predicting income, it will partly explain that relationship. However, if you add education, the effect of age will likely decrease because it was capturing part of the effects of education (the older you are, the more likely you are to have completed more years of education, to a point).

[@rickettpicks3534]

So if my R square is .942102 it is saying it is 94% accurate in explaining the margin of victory?

[@williamleiss4]

It means your regression equation explains 94% of the data points in your dataset.

[@rickettpicks3534]

@@williamleiss4 Thanks, now I need to make a predictive formula to match the theories. It's a start I suppose. Your videos are great thank you for all the hard work.

[@Ballsohard2324]

if the number comes back positive is that the spread for the home or away team ?

[@williamleiss4]

Depends on which side is being subtracted from who in the margin calcuation

[@chrismaisonet5717]

What College courses would I be able to take to learn and recreate these formulas?

[@williamleiss4]

Statistics and Computer Science

[@chrismaisonet5717]

@@williamleiss4 Thank you! I am currently in Elementary Stats but idk if that'll cover anything

[@depressedessendonfan5702]

IMO you just exaggerate the losses if you get this wrong. Can remove critical context then create more confidence. Needs to be taken with independent analysis

[@francisharte1774]

Have you back tested this? Curious what the results were.

[@jasoncihanowyz8886]

Could you do one of these for NBA? I want to create a linear regression model for it but don’t know what stats to use and what y value to use from what data to predict ML or spread

[@williamleiss4]

Look at my Excel series for NBA

[@AlexPerez-ud6iq]

What’s the apf and hpf

[@callumrichter7466]

When I try to calculate the ASRS I get #N/A

[@newwell321]

me too

[@newwell321]

something to do with the settings, I'm trying to figure it out

[@gardnertemkin6463]

Did you manually input all 256 games?

[@williamleiss4]

No

[@arushichaturvedi4941]

Amazing , i like it

[@playbyplaypodcast5167]

I looked at the score to the Dallas vs Detroit game in 2018. Cowboys won by 2

[@dallee2325]

If ASRS comes up as N/A, have any idea what the problem could be ?

[@williamleiss4]

N/A generally means a non numeric value is in one of the cells

[@knowwhere4185]

If we random forest... will it be better...make a video on that..also is there any software where i can run my model

[@angelnunez8760]

I've been profiting like crazy off sportsmodelanalytics dot com and their projection models. the best 24 bucks I ever spent

[@joejekpenyong]

Question: Is this model basic?

[@williamleiss4]

Very

[@ease3715]

This year they are using linear equations for nba betting. It’s surprising not a lot of people are aware.

[@jamesf3583]

what are the primary stats you use for game to game handicapping?

[@williamleiss4]

Those are my secrets

[@alleykat6171]

🤣🤣🤣🤣

[@HIGOLF]

Great video

[@romanluna4069]

Can you help with that ASRS formula i keep on getting N/A and im doing everything you do cant seem to find the issue

[@Ballsohard2324]

make sure your spelling for your teams are the same in SRS tab and Schedule tab

[@Joe-cy6cl]

im actually getting the same problem. I get the ASRS perfectly fine, but the HSRS I am getting N/A for a few of the teams. When I go to my schedule I copy that teams name and replace it with the same exact name in my SRS tab and now the HSRS is coming up, but now the ASRS is getting N/A. I dont get it because its spelled the exact same for the SRS tab, Home column and Away column. If anyone can help me out I would appreciate it. Tried troubleshooting myself, but can not find anything.

[@Joe-cy6cl]

hey go up and look at my comment to him i have an explanation

[@alleykat6171]

You guys are real morons, this guy is a double talking con artist.

[@KETODIETFREE]

amazing video

[@erikjohansson2300]

I prefer writing it out on paper.

[@DingBong6]

When doing multiple regression, you need to look at Adj R^2, not R^2, right?

[@williamleiss4]

Yes

[@wolfmib]

merci beaucoup pour cette video. Bon !

[@alivstyson1864]

confused this guys implies he can turn a profit then there are other video where he states he could NOT do it

[@paraschaurs]

Could have zoomed the video to understand the points

[@adammoore5980]

Also, Why does my coefficient get so large the more stats I add in?

[@williamleiss4]

What do you meam by large?

[@adammoore5980]

@@williamleiss4 initial coefficient sitting around -18 or so after loading more stats in. R value at .39 but the charts aren’t working for anything except the Home Power rating.

[@adammoore5980]

@@williamleiss4 could that be from the fact the only data for “Margin” is the first 12 weeks of the season?

[@williamleiss4]

@@adammoore5980 ignore the coefficients... R value is the only thing you need to pay attention to. I think 39 is too low

[@adammoore5980]

@@williamleiss4 oh okay thank you. I thought the coefficient was the home/away advantage that needed to be included in the process. So my main focus should be finding stats to get the R value as high as possible?

[@Ricatellez682]

Thanks so much pal

[@jadynfinch1685]

Hey there, this is an awesome video. I was wondering if there was a way you imported the scores at the beginning of the video. Thanks in advance

[@williamleiss4]

I exported the CSV from pro-football-reference.com's 2018 schedule page. I did a little clean up on it before the video to get rid of the columns I didn't need.

[@jadynfinch1685]

William Leiss Thank you, I was at that step but can’t figure out how to sort it so it has home team on one side and away on the other as opposed to winning on one and losing on the other

[@williamleiss4]

@@jadynfinch1685 It's something you have to do manually in Excel. =IF(F1="@",E1,G1) and copy down(Assuming the start of the file is at A1)

[@tjmorris86]

Are you a data scientist now?

[@johnsuarez6166]

How much different would a UFC or boxing betting model be than your traditional model? and how could I change your model to accommodate for UFC or boxing?

[@quinrizer6143]

You could use a logistic regression, with a response being 1 for victory and 0 for defeat. It’s a little more complicated

[@kylesprague2298]

Has anyone tried using something like an AHP to choose picks?

[@troydink18]

love your channel, I have a couple questions in regards to scaling your bankroll that i hope you can answer: - What's the maximum bankroll you can have if your betting 2.5% per play? -How do you deploy a huge bankroll of lets say $7million if your doing 2.5% per play like do you spread out the 2.5% with different sportsbooks? what's a good strategy?

[@troydink18]

I ask because i heard sportsbooks ban winners and/or limit them. thanks

[@_rellorT]

I wish I had your problem lmao

[@BringBiggieBack]

@@_rellorT 🤣🤣

[@Lifelongloser]

Great video. And what it shows above all else is just how utterly tedious betting singles on sport truly is . Makes you wonder why anyone bothers. There must be easier ways to make some money.

[@gtalio]

Simple is relative

[@cesaralvarez8558]

Would yards per play differential be a good stat for regression model?

[@williamleiss4]

As long as its opponent adjusted, yes

[@schumzy]

press F4 to lock cell...

[@goonie79]

or just add "$" if F4 doesn't work.

[@vita2200]

How about one using R! Excel is the devil!

[@juz882010]

that website is so slow now from all the scraping lol

[@samshankar9413]

Those biceps though

[@beattheodds2477]

Good intel. #cheers

[@mediapc4747]

"Predicative"? It's "predictive"

[@legendarhy1323]

did you use the wrong srs value at 9:00 ? Dont you use the one on the other page?

[@williamleiss4]

Not quite sure what you mean, but it looks right to me

[@jimmytouchdown7146]

ty

[@Lyashin55]

I appreciate your effort a lot. But cant you make same kind of analysis for soccer in one of major leauges? For instance it can be total goals over 2,5 . It will be usefu for non-us bettors because sports like baseball, football, basketball mostly famous in us. Soccer is the core betting of all around the world.

[@Ricatellez682]

Iam economist but, do you have any reference about betting sports and econometrics methods...

[@luisrodrigoulloaanguiano1163]

I have a program created by me that I want to exploit is NFL, hockey and NBA. If you are interested, I can send you the proof. highly effective, only I am the problem. I lose control and bet on everything. I'm economist too

[@talkwradio]

What’s the formula if you add additional stats to this model?

[@williamleiss4]

You add additional coefficients to the regression equation

[@talkwradio]

Sports Betting Truth thank you. So is it the additional coefficient times the stat? Example Intercept + (stat * coefficient of that stat + stat * coefficient of that stat) or just (statx1 * coefficientx1 + coefficientx2 + coefficientx3). Hope that makes sense what I’m asking. Thanks again!!

[@williamleiss4]

@@talkwradio the coefficients and stat each get their own set of parenthesis

[@talkwradio]

Sports Betting Truth ah, gotcha! Thank you for responding and these videos

[@simonkalis9635]

Nice bait man, have you ever been in contact with econometri? Using panel data like time series. But lets say its fine what about autocorrelation, heteroskedasticity, non stacionarity and fake regression?

[@kejdim777]

Šimon Kališ Yeah I'm pretty surprised that he didn't think to mention any of the conditions required for accurate linear regression, absolutely no mention of normality or variance of the residuals, independence, etc.

[@diankane9613]

What is Episoketren System? I've noticed several awesome things about this popular training program.

[@damyeonrasheedmosley7032]

This dude is in it for the views. If you wanna do all this mess and you understand decimals and reference sports. Just use Sagarin. This model is completely WRONG! Power Rating, Effectiveness Rating, Offensive & Defensive Ratings. If you don't know handicapping, you want figure it out this way. Excel is workbook to keep track of statistics, Period! It'll never calculated for injury or players just being out for rest. Good luck! Dude slow down!! You're teaching people, not a masters class! People are trying to keep up and your moving for yourself.

[@justinking5964]

Gosh why are you so handsome.You could be a film star.

[@RipCityPeezyRx]

I’m lost

[@clmco36]

These are perfect world snapshot stats. He doesn't take into account injuries, starting lineups, weather, days off, morning, afternoon, prime time televised games, the officiating crew etc... My dad used to call a guy like this an 'educated idiot'

[@ilooodesigns8617]

this is the worst regression model ever seen. r squared of 0,32 is awful, and all p-values nearly above 0,5

[@mikenyc1589]

Its better than if your doing this with horse racing....the most I can get is around .24

[@ilooodesigns8617]

@@mikenyc1589 still wont get you profit

[@petshopb]

Basic... ha! :D

[@joseferreirar]

my bookie is gonna try and shut down www.sportsmodelanalytics.com for sure before he goes broke.. lol

[@armin4594]

Many thanks, been searching for "martingale betting system simulator" for a while now, and I think this has helped. You ever tried - Ameliaten Bankify Bigshot - (do a search on google ) ? Ive heard some amazing things about it and my mate got great success with it.

[@julianbogart126]

jesus

[@youpsces642]

ok, I'm convinced www.sportsmodelanalytics.com saw the future with this game before it even started