Running in Google Collab #376
Replies: 4 comments 19 replies
-
|
I suggest you don't bother running this , this model isn't accurate . Its extremely misleading and well underperforming your probably better off just asking chatgpt whose going to win |
Beta Was this translation helpful? Give feedback.
-
|
Yes excatly and iv been saying this is a issue with this model. Iv always said treat this as a tool in a toolbelt. |
Beta Was this translation helpful? Give feedback.
-
|
@aaaclones someone did track some results from 2022-2023 season : https://docs.google.com/spreadsheets/d/1LxPq1ZF3k6BreB8p50vESP-7BTZXVrFs6abU7qxaq2c/edit#gid=1616360111 . I am not sure how you are using chatgpt to do a prediction when you haven't backtested it . I did some testing on chatgpt using some advanced data from nba stats and you simply cant do it due to resource constrained. I did used chatgpt to hep me with some code and run it on my local computer and it does work that way. I looked at how they backtest it and it was based on last 10 years vs closing line . If this model has an accuracy of 54% , it is beating the vegas odd with vig. I do agree with the weirdness of the confidence level that the model churn out . I am also seeing some random confidence when i ran on my own custom dataset. does this model account for injury or lineup change ? Yes and no. If lebron is out for the whole season , the team stat should reflect that . This model will not account for short term injury / lineup changes so you will need to make manual adjustment to the output. |
Beta Was this translation helpful? Give feedback.
-
|
For the year the model is getting crushed . The model doesn't account for anything outside the data points in the columns it's being ran on . Kyle's little sidekick always seems to think the model is being over fitted which is pure comedy . So no the model doesn't account for lineup change and to your question sure if lebron is out for the season the team stat will reflect it your right but what if he's out for just that game. ? or what if it's not lebron whose out but just Anthony davis or player 3 ? this model doesn't account for any of that .. not only that it spews out a statistical confidence level in the 90% and 80% meanwhile it's getting smashed to smithereens . If you prompt chatgpt properly with the proper scripts it will put this shitty model to shame . |
Beta Was this translation helpful? Give feedback.
-
|
please note at 53% sure your turning a small profit but is its way to volatile and stressful to deal with at 53% to comfortably say your beating it you would need to be somewhere in the 58% or better to be sitting comfortable and feeling you have the best of it . Now if your beating the lines and coming back on the number then sure you will have way the best of it but the books will never let you keep beating the number and middling the line . For example if you bet over 225 and now the number is 227 and you bet under 227 you will lose vig on one side and for the year the games land in between the lines roughly 50-60 times so if you keep beating the number and coming back on it with a hedge you will lose vig one way but be +50 units on the low end most certainly you will feel good about a bet you either scoop or lose vig . Goodluck to you man I advise you just to use the machine learning script Kyle has to managed to share he does deserve a lot of credit for being open source . The only value this repository has is the framework in the machine learning it provided but outside of that please dont follow the information to bet and don't let his little sidekick convince you otherwise he will claim over fitting and underfitting . Theres machine learning models that compiles and process 1000's of data points to come up with a prediction a far more accurate one than sports outcomes . |
Beta Was this translation helpful? Give feedback.
-
|
I dont think you understand how efficient the nba market is . Getting 1-2% edge over closing with vig is crushing . If you can come up with 58% model and you are able to put down money to the book without getting limited , props to you. Your example with the line movement clearly show you dont understand how sport betting work. I use market data + the output of my models to give me an idea which side i should bet. Yes , if the line move against me , it either means that there are some new information came out such as last minute lineup changes / injury. Please dont discredit the effort Kyle put into this. It is a pretty good starter model and with tuning with more advanced data , it can probably get maybe .5-.1% more accurate. |
Beta Was this translation helpful? Give feedback.
-
|
stfu you don't understand how it works , at 1% edge you think it's good
enough to keep putting your money on 1% and riding it ? go ahead and do it
then you dummy don't try and start a conversation being condescending you
obviously have no clue on how to beat sports otherwise you would know this
repository is trash , if the line moves against you chances are your on the
wrong side and you can't middle . Go fan out with Kyle you dummy . Get lost
and dont talk to me anymore your not worthy
…On Mon, Dec 4, 2023 at 12:19 PM albeng8 ***@***.***> wrote:
I dont think you understand how efficient the nba market is . Getting 1-2%
edge over closing with vig is crushing . If you can come up with 58% model
and you are able to put down money to the book without getting limited ,
props to you.
Your example with the line movement clearly show you dont understand how
sport betting work. I use market data + the output of my models to give me
an idea which side i should bet. Yes , if the line move against me , it
either means that there are some new information came out such as last
minute lineup changes / injury.
Please dont discredit the effort Kyle put into this. It is a pretty good
starter model and with tuning with more advanced data , it can probably get
maybe .5-.1% more accurate.
—
Reply to this email directly, view it on GitHub
<#376 (reply in thread)>,
or unsubscribe
<https://github.com/notifications/unsubscribe-auth/AY7NRWXJZCFLOJFVA4TNFOLYHYV5NAVCNFSM6AAAAABAEEW4H2VHI2DSMVQWIX3LMV43SRDJONRXK43TNFXW4Q3PNVWWK3TUHM3TONJXGI3TS>
.
You are receiving this because you were mentioned.Message ID:
<kyleskom/NBA-Machine-Learning-Sports-Betting/repo-discussions/376/comments/7757279
@github.com>
|
Beta Was this translation helpful? Give feedback.
-
|
good luck with your 58% model and get your money down to bet. |
Beta Was this translation helpful? Give feedback.
-
|
Buddy said “beat sports betting” 😂 You a sharp or something?
…On Mon, Dec 4, 2023 at 3:24 PM aaaclones ***@***.***> wrote:
stfu you don't understand how it works , at 1% edge you think it's good
enough to keep putting your money on 1% and riding it ? go ahead and do it
then you dummy don't try and start a conversation being condescending you
obviously have no clue on how to beat sports otherwise you would know this
repository is trash , if the line moves against you chances are your on
the
wrong side and you can't middle . Go fan out with Kyle you dummy . Get
lost
and dont talk to me anymore your not worthy
On Mon, Dec 4, 2023 at 12:19 PM albeng8 ***@***.***> wrote:
> I dont think you understand how efficient the nba market is . Getting
1-2%
> edge over closing with vig is crushing . If you can come up with 58%
model
> and you are able to put down money to the book without getting limited ,
> props to you.
>
> Your example with the line movement clearly show you dont understand how
> sport betting work. I use market data + the output of my models to give
me
> an idea which side i should bet. Yes , if the line move against me , it
> either means that there are some new information came out such as last
> minute lineup changes / injury.
>
> Please dont discredit the effort Kyle put into this. It is a pretty good
> starter model and with tuning with more advanced data , it can probably
get
> maybe .5-.1% more accurate.
>
> —
> Reply to this email directly, view it on GitHub
> <
#376 (reply in thread)>,
> or unsubscribe
> <
https://github.com/notifications/unsubscribe-auth/AY7NRWXJZCFLOJFVA4TNFOLYHYV5NAVCNFSM6AAAAABAEEW4H2VHI2DSMVQWIX3LMV43SRDJONRXK43TNFXW4Q3PNVWWK3TUHM3TONJXGI3TS>
> .
> You are receiving this because you were mentioned.Message ID:
>
<kyleskom/NBA-Machine-Learning-Sports-Betting/repo-discussions/376/comments/7757279
> @github.com>
>
—
Reply to this email directly, view it on GitHub
<#376 (comment)>,
or unsubscribe
<https://github.com/notifications/unsubscribe-auth/AQHPWRO2BRAWC2DE3Y25KWLYHYWRNAVCNFSM6AAAAABAEEW4H2VHI2DSMVQWIX3LMV43SRDJONRXK43TNFXW4Q3PNVWWK3TUHM3TONJXGMZDS>
.
You are receiving this because you are subscribed to this thread.Message
ID:
<kyleskom/NBA-Machine-Learning-Sports-Betting/repo-discussions/376/comments/7757329
@github.com>
|
Beta Was this translation helpful? Give feedback.
-
|
your dumbass said 1% is crushing lmao stfu get out of here that's not how it works go push marginal edges and think your beating it |
Beta Was this translation helpful? Give feedback.
-
|
good luck with your 5% edge model. |
Beta Was this translation helpful? Give feedback.
-
Is this possible to run this entirely in google collab? I can get it to run and produce the initial outputs but none of the commands to update the data or train the data further seem to be working (they return syntax errors).
I have tried repeatedly to get this to run natively on my machine but no matter what I do I get the "OSError: Cannot parse keras.." error, even after trying multiple different versions of python and pandas as suggested in other discussions.
Does anybody have any insight on how to get this properly running on google collab or have any suggestions on other things I can try? Any and all help is greatly appreciated!
Beta Was this translation helpful? Give feedback.
All reactions