Bottoms Up ft. GP13 (Originating 101) | Ep 51

This one's going to specifically focus on giving you guys an overview guys and gals , an overview of bottom up bedding . So since my podcast with Jfar , I've gotten a lot of DMs about what is bottom up bedding . Just describe it . That was in the comments of the YouTube video and a lot of DMs like what is bottom up bedding ? How do I get started ?

What are some examples of it ? So since usually when I got a lot of DMs about something I realize it's probably a good topic to make a podcast about , and since this is on a lot of our minds after the Jfar episode , then seems like a good time to do it , so I want to jump right into it .

So bottom up bedding is essentially the cousin , the other side of the coin to top down . Top down is what a lot of people listening to know as plus CV bedding , and it's essentially just comparing discrepancies between different sports books and using those discrepancies to try and find a price .

So you're not trying to come up with your own price , but you're trying to use the sports books , determine who might be better at pricing of the sports books and then use their price . Bottom up is you're creating your own price , you're creating your own bet and because actually not all bottom up means you're creating your own price .

What I would call bottom up is it's if every sports book had the same price , you would never find a bet in top down . So if literally , like the cheese first Eagles was minus three , minus 110 across every single sports book , top down bedding would never find a bet there . Bottom up bedding would occasionally .

It's probably not but bottom up bedding would have a chance of finding a bet . That's how I would differentiate it . So top down bedding , if every single sports book has the same line , then there will not be a bet found . Bottom up is any strategy that could find a bet , even if every sports book had the same line .

So the reason I say that is because I think there are . I split bottom up that into two categories in my head and the two categories are modeling and angles . So I'll touch on modeling first .

Modeling is what I think most people think about when they think about bottom up Modeling is it's an ambiguous term , but generally the idea is you take some data from a sport right , you try and organize it , clean it , structure it , which is the hard part of modeling .

Definitely the hard part is getting all the right data , getting it to work with each other , building a database so you can manipulate it , etc . That's the tough part .

And then using that data to inform some kind of price on a bet , some kind of fair value on a bet , the same fair value that you would get from your top down strategy where you're using other sports books . And in that situation you're using other sports books models , right .

So the same way that you're getting a fair value from the models of other sports books , in modeling as a bottom up practice , you're getting a fair value from your own model , your own inputs . So I heard there's someone who said on Twitter this is Botide Better , who's one of my absolute favorite followers on Twitter if you don't follow that account .

But they said I think we had a little talk on Twitter about what is pure originating and what is top down and he said pure , pure originating would be someone locked in a room who has no idea of what's going on in the market and just spitting out a number which Doesn't , yeah , and he said , which doesn't exist and would be stupid to do anyway .

And I was like , yeah , that that does make sense . Because the reason he said that and the reason we're having this conversation is part of modeling is Spitting out your own number and then seeing where the market's at and adjusting based on where the market is .

Usually you're regressing a little bit to the market , from your model to the market , and it's not in a vacuum . So when I say bottom up , I mean bottom up . The best bottom up better is understand top down and and use top-down strategies in their betting . All right , so what are some skills you need for modeling ? The first skill is research .

You need to be able to quickly find information that you're missing , because a lot of , a Lot of anything is just looking stuff up on Google , being good at googling , being good at you know , expediting the process by finding information out there that you can quickly use , instead of Spending hours and hours trying to come up with it yourself .

So , like , let's say , you're looking for you know , oh , how many times has so and so gone over this line . Well , if there's already like a code repository somewhere that has this exact thing , then set of like trying to build it yourself , you could go and just use that very basic example .

But it's , you know , always go look for something before spending hours and hours trying to build it yourself and always , you know , try and figure out answers to your question Questions as efficiently as possible .

As someone who switched from Excel to Python , I've done a ton of this research , looking it up , using other people's stuff , asking friends , you know , just being scrappy and using what's out there and trying to learn along the way .

So research is very important for modeling , I would say domain knowledge , and this one I say some domain knowledge that , that being said , I still think I might even go above some .

I wrote some , but now I might go to a lot of them in knowledge , because you need to know what to look for and you can't just back test , you can't just optimize for historic data and expect to win on it .

So you need to have some kind of thesis about something that changed or something that's not being taken into account by the market , and you want to have a model that connects with reality instead of trying to fit something perfectly to the historic data , which has shown in many different markets to be a surefire way to lose a lot of money .

Now , why I said some was because if you think about someone like Pickham Expert , who's really locked in on correlation , I think that's an example where he started out , didn't have a ton of domain knowledge and still built a winning model because he was very good at basically modeling , correlation and gathering data and doing all the important things like database

management and all that stuff , and then he learned along the way . He started to follow the sports and gain some more domain knowledge . But he's a good example of someone who came in and started winning without much domain knowledge . So it's possible .

Database management is super important , so that's like manipulating , cleaning , gathering data , so that could be scraping . That could be finding APIs that you can connect with and interact with . That can be creating your own data . That's kind of low key , a very valuable thing in sports betting . If you're creating good , unique data .

That could be something that you're the only one leveraging . Programming . Important skill Now , I said , can be Excel . That's where I started and we did very well with Excel for a long period of time , so I'm not knocking Excel .

That being said , I've transitioned the whole basically everything I do to Python over the last couple of months and it's been night and day . I would never go back to Excel after learning Python . A lot of people program in R for sports betting modeling that's another good language and for the majority I see some some split of Excel , r and Python .

So if you already know Excel , maybe start there . Or if you want to learn , if you don't know any of them , I would maybe choose R or Python and learn while making some kind of basic model . And then the last skill you need is it's basic understanding of statistics . So your models are going to spit out R squared or whatever they're going to .

Your models are going to be based on some type of could be a linear regression , something as simple as that , where you spit out an R squared and you have to know what that means is . Actually , you don't have to go take a class on statistics , but it's important to kind of understand the terms and how models work .

You know , talk about training data and test data and kind of intricacies of what's happening , and if you have domain knowledge . It's helpful because you won't get too lost in the sauce . You know you won't , you won't be too swayed by by just like your model outputs .

But if you have both domain knowledge and some understanding of what the model is saying , it gives you a way to , kind of idiot test the model so you don't lose a ton of money . So those are the skills that you would need for modeling . And now let's just do a quick example of what a sports betting model could be in practice .

So what you could be doing is you could be looking at , so you want to go into a model thinking about , like , what am I trying to predict here ? What market is this model going to be for ? Now you could be trying to predict something very niche , like second half assists , or something you could be trying to predict a main line market , like NFL spreads .

But what you want to do is you want to think about okay , what am I trying to predict ? So let's just go with second half assists because it's more accessible , it's more realistic as something you might be able to model and win at than NFL spreads , which is going to be much more difficult . So if you were thinking about how to predict second half assists .

What would you need to know ? Well , you probably need to know . You're doing this for each player , so you need to know , kind of just a baseline what are their second half assist numbers , maybe what are their full game assist numbers ? Then you'd have to know something like their lineup how many assists do they get ?

What does certain players being in and out , in fact their assists ? What happens in the first half ? What happens in the first half , in fact their assists ? Is there someone that they split time with that ? You need to maybe do an input on their total fouls in the first half . That might impact how much playing time that person gets in the second half , right ?

So basically , you're just trying to determine how much this player is going to play , who they're going to play with , and then , in those circumstances , give them an efficiency rating on assists . So you would try and gather a couple different sets of data , combine them all together and spit out some kind of prediction on second half assists . And then you would .

This point and this is where the top down comes in is you go look around the market , see what's posted out there , and you might regress your model . You know , 20% to the market or something , and there's no . It's kind of an art , because first you wouldn't regress it to just the market .

There's certain sports books that you would respect more on their second half assists , and that comes from just following them , you know . So you would do some kind of composite or just pick one or two of the more respectable ones , the ones that the market tends to move to , and just use those and then how much you want to regress towards the market .

I mean , I don't know the perfect way to do this . I think I've heard 20% , I've heard even higher . I do different variations myself and it's more of like a finger to the wind , but the main point is that you have an idea of where the market's at and you move a little bit towards it to protect yourself against being really wrong .

So in that example , we'd check the sharp books for assists , we'd have our assist numbers and we would move our assist numbers slightly towards the assist numbers of the sharp books and then , if we see any value , it's the same thing . You got a fare , you got a line and you make your bet . So that's just a basic example of modeling .

Okay , and now the second half of bottom-hook up-betting , which I break down in my head to modeling and to angles , is angles . So when I think about angles I'm thinking about I still bet that doesn't show value in a top-down sense , but it's not a pure output of a model . It's not .

You have a bunch of data that comes in to your model and it spits out the bar on 27.5 points and you see the line's 30 and you know the probability difference between one point and you say , oh , so he's 60% to go under . I can take the under . That would be from a model .

An angle is you're not going to have necessarily a fair price , but you will have very strong evidence that the bet is a good bet .

These are going to be based on more heavily on domain knowledge , understanding sports book rules correlation , of course , is going to play a role in a lot of these and you're going to make a bet where you don't necessarily have a fair but you know it's a really good bet .

I'm actually going to give an example of this first , because I think it's not as easy for everybody to picture , but a couple of examples . One would be the McHale Bridges situation . We all remember that game where he was preserving his Ironman streak and he played a couple minutes and then sat out .

A lot of people not a lot , but some people who knew the NBA very , very well understood the coach talk , understood Bridges Ironman streak and understood that while they were going to sit , some starters that he was probably going to play just enough to preserve the streak and then leave the game and kind of rest with some of the rest of the starters .

What happened in that situation is people knew that this was likely to happen . They didn't know the exact fair value . How do you put a price on that ? How do you put a price on guessing that someone's going to try and preserve their Ironman streak ?

What the Sharp Betters did was they took a lot of same game parlays correlated on Bridges Unders , basically saying a ton of Unders parlayed together that he's not going to get X amount of minutes but he will set foot on the court to preserve the streak .

It's like he will play , so the bets will stand , but he will not play enough to get any of his stats up to the line that the books have set for them . That's a great example of one of the more recent popular angles which people want a lot of money on Golf correlation .

Another one you might not know exactly the correlation between two golfer scores , but you know that the overs and unders are correlated

. If the course plays hard or easy , you're willing to place a bet by knowing that about golf , that a course could play hard that day . Play easy because of the conditions , because of the course . You're willing to place a bet without necessarily knowing a fair or having a model tell you this is the exact probability of these two golfers going over .

But you can know the game well enough to know that you can place that bet . The skills you need to do this are great research skills like beyond great . Think about the Bridges situation . You have to know about his Ironman streak . You have to heard the coach talk about sitting starters .

You have to remember that it happened last year , so been paying attention long enough or researched it enough to know it happened last year . You need to have great domain knowledge . Great domain knowledge . Bridges situation again , you're not going to know that unless you know basketball , you're not going to be able to .

You're not going to have followed the Bridges situation long enough to remember what happened last year , to follow the coach speak enough to understand what to do and to know how many minutes is he actually going to play . If you told me that , I'd be like well , I still don't . I don't get it .

Does that mean he's going to actually play the whole game to reserve a streak ? I know nothing about basketball , so I'm still . Even if you told that to me , I still wouldn't be as effective of putting it into practice as somebody who has really good domain knowledge and knows exactly how this situation is going to play out .

Most of the time To play these angles , you have to have great research , great domain knowledge or great understanding of a sports book's rules . There's plenty of situations where rules have created an opportunity for betters to make profitable bets if you read the terms and conditions of the sports book .

To finish up , I wanted to give an example of an angle that relies on knowing the house rules . So , like a house rules angle , this falls under research . Always learn the rules of the market you're betting in , no matter what . You've found plenty of gold in the TSCs and the rules of the house rules of a sports book over the years .

So always check the rules . There might be something in there that either makes you a lot of money or helps you avoid losing a lot of money and not being able to get it back because you made a mistake . That was clearly stated in the rules .

So there is a good article on Unabated , written by a professional better named Tilemicus model , about an MLB situation that happened I think it was in 2022 on Caesar's sports and the situation was about ninth inning money line .

So Caesar's would offer inning money lines for each inning and a game of baseball and usually the money lines would be like pretty close , depending on whatever the relative strength of the team . But he offers an example in his article where he shows the seventh , eighth and ninth inning money lines of a Rockies at Miami Marlins game .

So each inning in the seventh inning , the Rockies are plus 115 , marlins are minus 135 . In the eighth inning , rockies are plus 120 , marlins are minus 140 . In the ninth inning , rockies are minus 220 , marlins are plus 180 . So that's a total flip because Marlins have gone from being about minus 130 , minus 140 to plus 180 . And he says why is that ?

We always have to ask why . And it's because in baseball , if the home team's winning , they will probably not get to bat in the ninth . So fair enough , the Rockies would basically just get to free roll the Marlins a large percentage of the time and those numbers look right .

But instead of just seeing that and kind of whatever , going on to the next thing , he decided to take a look at the rules of the sportsbook , the rules of Caesar sportsbook , and he found a section called innings betting . And these are the two rules . Rule number one predict the team that will score the most runs in a specified inning .

The full inning must be completed for bets to stand . In the event of a tie this is rule two bets will be void unless odds are quoted for the tie . The main thing that matters there , and you might already realize this , is the full inning must be completed for bets to stand .

So the situation is the Rockies are no longer free rolling the Marlins because if the Marlins have won the game and the Rockies have scored in the ninth but haven't scored enough to beat the Marlins in the actual baseball game , then the bet will void .

So that whole ninth inning advantage that the way team has is taken away because the bet voids if the inning is not completed . So what ends up happening is the ninth inning money line should look very similar to the eighth inning money line , but you're getting a price that's twice as good on it . You're getting plus 180 instead of minus 140 .

And the only reason this is a good bet is because of the House rule is that , say , the full inning must be completed . On other books the full inning probably didn't have to be completed and they were giving the line out as minus 220 plus 180 . But on Caesars it did .

So they ran with this angle for a while , made a lot of money and it wouldn't have happened if they didn't dig into the TOCs and the House rules . So anyway , those are some examples of what I would call angles .

Another bottom-up strategy takes good research , great domain knowledge and a willingness to place a bet without knowing the exact fair and yeah , and just keep digging . So anyway , that's my 30 , 25-minute overview on bottom-up . It's very basic , but I just wanted to give everybody a taste of what's out there in the space .

What are some different ways you can start with bottom-up ? What are the ways that might be more suited to your strengths or your weaknesses ? How much time do you have ? That all factors into how you would want to attack it , build it into your process .

But the good part about being a top-down better is you already understand how the market works , how to price shop , how steam chasing works . This is all valuable stuff that makes your better bottom-up . So I would just look at it as a natural progression of your sports betting skill your sports betting journey , as people say . So I hope that was helpful .

Thanks everyone again for listening and I will see you on the next episode .