Machine Learning Model for NBA Sports Betting

[Eternad]

I just saw this Machine Learning for NBA Sportbetting that being use on other forum which you can have your own model to create an analysis on past statistics of the team and create a betting analysis for you.

The Machine Learning Description by the developer

A machine learning AI used to predict the winners and under/overs of NBA games. Takes all team data from the 2007-08 season to current season, matched with odds of those games, using a neural network to predict winning bets for today's games. Achieves ~75% accuracy on money lines and ~58% on under/overs. Outputs expected value for teams money lines to provide better insight.

This is the actual topic which some users used the machine learning to create model for him.

https://www.reddit.com/r/sportsbetting/comments/131pf84/machine_learning_model_nba_predictions_42823/

He has 65% winning percentage on all of his bets that he recorded on his spreadsheet https://docs.google.com/spreadsheets/d/1LxPq1ZF3k6BreB8p50vESP-7BTZXVrFs6abU7qxaq2c/edit?usp=drivesdk


I’m planning to give this tool a try since manually digging the stats of team is a hassle job to do. Just sharing in case someone wants to experiment.

[Eternad]

He has 65% winning percentage on all of his bets that he recorded on his spreadsheet https://docs.google.com/spreadsheets/d/1LxPq1ZF3k6BreB8p50vESP-7BTZXVrFs6abU7qxaq2c/edit?usp=drivesdk

Update: The moneyline and Under/Over result is pretty bad maybe because most of the result on this playoffs is an upset since machine learning is only using stats to create an algorithm for choosing a bet.

This kind of match which the team that newly complete the full line-up can easily destroy the credibility of the machine learning since it doesn’t consider the player star factor on the winning probability of the team. This is the reason why it’s better place bet manually if you are aware on the drastic changes on the team during playoffs.

[danherbias07]

I checked the link in Reddit and he said it himself at his Q&A, it is not predicted using updated factors like injuries, rosters, and so on. That means a big change could happen. It's only using a history basis so it's a risk.
A good example is the game last night Denver Nuggets versus Phoenix Suns. The Nuggets are known to score 120+ points and it's on paper that they win most of their game at that score. No hard injuries happened, just CP3 hurting his groin but the known offensive players are still available.
The starting line Over and Under was 228.5 if I remember it right. I didn't take that bet after what I saw in Game 1 where the Suns only scored 100.
Thankfully, I was right, they only scored 87 yesterday, Game 2.
It could help for a good basis, but for those who will bet, always check the rosters first and analyze how the game could end up especially now in the playoffs where defense is hardcore and refs are not calling soft fouls which means the time keeps on ticking down.

[Eternad]

I checked the link in Reddit and he said it himself at his Q&A, it is not predicted using updated factors like injuries, rosters, and so on. That means a big change could happen. It's only using a history basis so it's a risk.
A good example is the game last night Denver Nuggets versus Phoenix Suns. The Nuggets are known to score 120+ points and it's on paper that they win most of their game at that score. No hard injuries happened, just CP3 hurting his groin but the known offensive players are still available.
The starting line Over and Under was 228.5 if I remember it right. I didn't take that bet after what I saw in Game 1 where the Suns only scored 100.
Thankfully, I was right, they only scored 87 yesterday, Game 2.

This is true. This is the disadvantage of using this kind machine since the parameters of prediction is set by human and it needs to be updated regularly. The problem is if the developer doesn’t bother updating all those factor that you mention that will gives a huge impact in the game.

I don’t think the presence of CP3 will not gonna give that much impact on the game for Suns to win. They are still managing to fight Nuggets side by side on 3rd quarter until some point on Q4. Their scorer is just shutdown badly by Nuggets defense. Anyway this is already for the NBA thread.

It could help for a good basis, but for those who will bet, always check the rosters first and analyze how the game could end up especially now in the playoffs where defense is hardcore and refs are not calling soft fouls which means the time keeps on ticking down.

Agree, we should always check the actual condition of the game before we made any bet on the matches. I experience this in the past which I place pre match bet with the Bucks. I’m too excited to place bet since Bucks has a higher odds while they are the top team. I just realized during the game that Giannis is injured that’s why the odds is so high for them.

[Casdinyard]

This is a potent tool if it at least provides a higher than 52% chance of success. Meaning it's up to something.

I think this is a great tool to help gamblers predict which team to bet on, NBA's not really anybody's game so there's bound to be teams that will statistically rise to the top which the Machine's Algorithm's probably getting a hold of. For other types of betting that are pretty circumstantial this wouldn't really work well. But for reliably predicting who's gonna win? I think that's more than enough. I say you take it for a spin OP. Let us know how your win rate's gone and make it a little scientific too so we can really say for certain that  the algorithm is working, perhaps make a small bet without using the algorithm vs using it in a set span of time, and then let us know how it did.

[Zilon]

Taking data from 2007/08 season till current data can make very good predictions. The more the data the better the model's predictive scores. But one thing is it will have to require steady update on the trained and test data sets to make sure the model learns from more recent datas.

It won't be a bad idea giving it a try but make sure the model is updated before risking your funds because outcomes in gambling are never static and things changes for various reasons. So the more recent the last value in the data set is the better the prediction and the better the outcome as well.

[bisdak40]

Taking data from 2007/08 season till current data can make very good predictions. The more the data the better the model's predictive scores. But one thing is it will have to require steady update on the trained and test data sets to make sure the model learns from more recent datas.

It won't be a bad idea giving it a try but make sure the model is updated before risking your funds because outcomes in gambling are never static and things changes for various reasons. So the more recent the last value in the data set is the better the prediction and the better the outcome as well.

For me, no amount of data can make a game in the NBA more predictable because if indeed that "data" thing can make us predict the outcome of the game by more than 50 percent, for sure bookies will stop doing their business as they won't be gaining profit anymore.

But if you will give it a try OP, I would be interested to follow the outcome if you post the result in this thread or create another one dedicated to that.

[robelneo]

This kind of match which the team that newly complete the full line-up can easily destroy the credibility of the machine learning since it doesn’t consider the player star factor on the winning probability of the team. This is the reason why it’s better place bet manually if you are aware on the drastic changes on the team during playoffs.

This is one of the big issues in AI or machine learning I consider this an AI too because it is a computer program to give an output, it can provide accurate data, and its prediction is based on old and fresh provided data but when it comes to big changes or sudden changes it is lost on how to give input while we humans have insight to rely upon, besides AI when testing it for the best time will tell you that they are an AI and they based their output on what's on their database, so they are limited.

[btc_angela]

I just saw this Machine Learning for NBA Sportbetting that being use on other forum which you can have your own model to create an analysis on past statistics of the team and create a betting analysis for you.

The Machine Learning Description by the developer

A machine learning AI used to predict the winners and under/overs of NBA games. Takes all team data from the 2007-08 season to current season, matched with odds of those games, using a neural network to predict winning bets for today's games. Achieves ~75% accuracy on money lines and ~58% on under/overs. Outputs expected value for teams money lines to provide better insight.

This is the actual topic which some users used the machine learning to create model for him.

https://www.reddit.com/r/sportsbetting/comments/131pf84/machine_learning_model_nba_predictions_42823/

He has 65% winning percentage on all of his bets that he recorded on his spreadsheet https://docs.google.com/spreadsheets/d/1LxPq1ZF3k6BreB8p50vESP-7BTZXVrFs6abU7qxaq2c/edit?usp=drivesdk


I’m planning to give this tool a try since manually digging the stats of team is a hassle job to do. Just sharing in case someone wants to experiment.

If I'm not mistaken, some modelling prediction is already been used by some websites or those who give tips to NBA fans. So I'm not surprised that with the advent of AI, even NBA games are going to be used to create a 'perfect' machine learning to predict the winners.

However, as you have said, there could be upsets, just like in the Heat vs Bucks or Memphis vs Lakers that AI learning can't see. Everything is just base on numbers and statistics and if is broken then obviously it will fail at some point.

[ImThour]

I am really interested in ML models for sports betting however there is something I do not agree with.
They use previous data to predict the future just like any other model, fair enough. But it's not how it should be.

ML models should have human-like brain and should be able to compete each and every individual player and it's last match performance and then generate a result.
I am sure the ML model behind ChatGPT i.e. GPT4 will be a really great tool for this.

[Pierre 2]

Even though it doesn't include very important details like injuries, its still very handful of winning percentage. I feel like AI can be improved to levels beyond our comprehensions because it can practically have all data in world if you work harder enough. Data is there. Only thing developers will do to add algorithms, ways to let it embrace. Also I feel like lack of independent inspiration of biological machines (humans) still kill purpose of AI. I mean AI should be able to do wild guesses as well. I guess it will happen in couple of years anyways.

[Yatsan]

I am really interested in ML models for sports betting however there is something I do not agree with.
They use previous data to predict the future just like any other model, fair enough. But it's not how it should be.

ML models should have human-like brain and should be able to compete each and every individual player and it's last match performance and then generate a result.
I am sure the ML model behind ChatGPT i.e. GPT4 will be a really great tool for this.

Human like brain would be impossible at the present. Even with AI projects, existing data are only the basis of the information or the limit eith the information, such software would be processing. With ML in gambling, I am not really into it. I would more enjoy losing because of my decision than to rely with the generated data in particular with the outcome. Also, losing is more certain in gambling and I won't be happy winning because a tool just helped me. It could also promote an advantage over those who do not have the tesources to access MLs. Not considered as cheating but it somehow bypasses the idea of what gambling actually is.

[ryzaadit]

I believe, I also have a similiar topic but using "ChatGPT".

However, most of machine learning are taken the data from internet and most of their data are the conclusion. The result on other topic, the AI not using updating data but they're using old data from the last season.

So, honestly the data/result are not quite good.

[Fortify]

I just saw this Machine Learning for NBA Sportbetting that being use on other forum which you can have your own model to create an analysis on past statistics of the team and create a betting analysis for you.

The Machine Learning Description by the developer

A machine learning AI used to predict the winners and under/overs of NBA games. Takes all team data from the 2007-08 season to current season, matched with odds of those games, using a neural network to predict winning bets for today's games. Achieves ~75% accuracy on money lines and ~58% on under/overs. Outputs expected value for teams money lines to provide better insight.

This is the actual topic which some users used the machine learning to create model for him.

https://www.reddit.com/r/sportsbetting/comments/131pf84/machine_learning_model_nba_predictions_42823/

He has 65% winning percentage on all of his bets that he recorded on his spreadsheet https://docs.google.com/spreadsheets/d/1LxPq1ZF3k6BreB8p50vESP-7BTZXVrFs6abU7qxaq2c/edit?usp=drivesdk


I’m planning to give this tool a try since manually digging the stats of team is a hassle job to do. Just sharing in case someone wants to experiment.

This is actually a very interesting idea and could close the gap between the average user versus the sophisticated bet analysis software used by bookmakers to offer bets. Even if it's just able to tease out better performance or possibly compare more data points in the long run. Even boosting the winning prediction percentage by one or two points extra could be very beneficial for players and start to balance the odds of sportbooks. It will definitely be interesting to follow a long someone making predictions that are actually based on an intelligent model, rather than random guessing by someone who follows the sport but is susceptible to blind spots in their knowledge like everyone else.

[coin-investor]

He has 65% winning percentage on all of his bets that he recorded on his spreadsheet https://docs.google.com/spreadsheets/d/1LxPq1ZF3k6BreB8p50vESP-7BTZXVrFs6abU7qxaq2c/edit?usp=drivesdk

Update: The moneyline and Under/Over result is pretty bad maybe because most of the result on this playoffs is an upset since machine learning is only using stats to create an algorithm for choosing a bet.

This kind of match which the team that newly complete the full line-up can easily destroy the credibility of the machine learning since it doesn’t consider the player star factor on the winning probability of the team. This is the reason why it’s better place bet manually if you are aware on the drastic changes on the team during playoffs.

This is one of the times that AI employment is not accurate because the AI did not consider sudden changes in the lineup, they only analyze the team based on their last performance, so if you take out two or three strong members of the team and it was replaced by weak players the results will not be the same, to get good results there should be no sudden changes.

In the case of the NBA, there's always trade involved, there's always last-minute replacement because of injury or the contract expire
human analysis in this case is still favorable for bettors.

[Oasisman]

He has 65% winning percentage on all of his bets that he recorded on his spreadsheet https://docs.google.com/spreadsheets/d/1LxPq1ZF3k6BreB8p50vESP-7BTZXVrFs6abU7qxaq2c/edit?usp=drivesdk

Update: The moneyline and Under/Over result is pretty bad maybe because most of the result on this playoffs is an upset since machine learning is only using stats to create an algorithm for choosing a bet.

This kind of match which the team that newly complete the full line-up can easily destroy the credibility of the machine learning since it doesn’t consider the player star factor on the winning probability of the team. This is the reason why it’s better place bet manually if you are aware on the drastic changes on the team during playoffs.

This is the first thing that came to my mind when I hear about this machine or AI that generates stats from the past up to the present. The only credible source that we get from this machine was only the stats, everything else may not be available unless this machine is well maintained and updated every time there are changes from the teams in the NBA in general.
One important factor to look at when placing a bet is the injury updates and player availabilities.
This machine could also be a good source to analyse a match but of course it could be accompanied with manual research to make it more effective.

[wheelz1200]

I just saw this Machine Learning for NBA Sportbetting that being use on other forum which you can have your own model to create an analysis on past statistics of the team and create a betting analysis for you.

The Machine Learning Description by the developer

A machine learning AI used to predict the winners and under/overs of NBA games. Takes all team data from the 2007-08 season to current season, matched with odds of those games, using a neural network to predict winning bets for today's games. Achieves ~75% accuracy on money lines and ~58% on under/overs. Outputs expected value for teams money lines to provide better insight.

This is the actual topic which some users used the machine learning to create model for him.

https://www.reddit.com/r/sportsbetting/comments/131pf84/machine_learning_model_nba_predictions_42823/

He has 65% winning percentage on all of his bets that he recorded on his spreadsheet https://docs.google.com/spreadsheets/d/1LxPq1ZF3k6BreB8p50vESP-7BTZXVrFs6abU7qxaq2c/edit?usp=drivesdk


I’m planning to give this tool a try since manually digging the stats of team is a hassle job to do. Just sharing in case someone wants to experiment.

I'm sure this is decent to use.  The only caution I have is that all regular season stats get thrown out the door and can't be a good predictor in the playoffs.  Coaches coach differently and teams use different rotations in the playoffs that aren't used during the regular season.  Every tool is a helpful piece but I wouldn't put all my chips on what this thing spits out.

[dothebeats]

NBA sports betting with AIs will be a mess. There are a lot of variables that come into play that isn't really predictable like injuries and whatnot. Sure, you may have the stats and the historical data to use as a baseline for your predictions, but are these sufficient enough to be used by AIs infallibly? 65% win rate is honestly an impressive number for an AI that cannot take into account the injuries and other such stuff. I'd use this AI for player stats, but not for moneylines because that is an area that is largely affected by a lot of variables that the AI cannot take into account.