I Built an AI Sports Betting Bot with ChatGPT
Вставка
- Опубліковано 9 чер 2024
- I gave WagerGPT $2000 to make sports bets with and in this video i'll explain how I built this AI sports betting bot as well as whether it lost it all or made a potential total of $6,850.00 after 24 hours. It's available starting today at wagerGPT.co . First, a disclaimer - Do NOT invest any money in any type of sports betting bot or algorithmic engine that you are not willing to lose. I gave this sports betting bot money because I was willing to lose money to make a great video for my AI Wizards out there. Each week i'm learning more on how to optimally prompt ChatGPT to help me build MVPs, and it's a skill i'd like us all to learn, which is why i made this video. The tools used in this tutorial are the OpenAI API for sentiment analysis, Python for Programming, BeautifulSoup for Web Scraping, PyTube for UA-cam Videos, Requests for API requests, Flask for Web Development, and Fly.io for deployment. Let me know what you think in the comments and I hope you have a great day today. Enjoy!
Sign up at WagerGPT (Only 500 Spots available)
wagergpt.co/
P&L report:
wagergpt.co/reports
Want more AI/ML education? Connect with me here:
Twitter: / sirajraval
TikTok: / sirajraval
Facebook: / sirajology
Linkedin: / sirajraval
I Built a Sports Betting Bot with ChatGPT:
• I Built a Sports Betti...
I Built a Trading Bot with ChatGPT:
• I Built a Trading Bot ...
Watch ChatGPT Build an AI Startup:
• Watch ChatGPT Build an...
Watch ChatGPT Build a Finance Startup:
• Watch ChatGPT Build a ...
Watch Me Build a Startup Playlist:
• Watch Me Build a Finan...
Join my AI Community:
gptschool.app/ Sign up for my AI Sports betting Bot, WagerGPT! (500 spots available):
www.wagergpt.co
Is it possible for Siraj to do anything other than slick and half-assed?
Why is your video still up???
Aaaaaannnnndddddd we’re back
How accurate is this when the bot has a prediction win chance of 99.1%?
how do you cancel before the free trail is up
you can't call your bank remove his token from the merchan and get new card remove all your cards and remove him from all the token merchant
Hi, I'm trying the trial version, and it has a lot of bugs when you try to open the reasons for the bet, for example if you look at a basketball bet it opens up a detailed football bet, I hope you will solve the problem. Thanks
Thank you, yeah several other users mentioned this, I’ll have it fixed in 24 hours
this is fixed now!
@@SirajRaval its still hapening
Noticed this too with some NHL pick when I chose Bet suggestion it suggested a NCAA bet :D@@Djr2699
scammmmmmmmmmmmmm@@SirajRaval
How do you use sportsdataIO if the data is scrambled
Why I am getting different code block with same questions as u, on step when I ask: "Ok, please create a logistic regression model to predict the winning team using this dataset".. Should I use gpt 4?
Ahhhhh ...You started by showing us how code it then scrapped it and now you're trying to charge us for it. Man I'd pay you for an indepth course on how you did it ....i'm not gonna pay a monthly subscription just to use your bot; its shame your first video on this subject was really interesting and informative; but now you're like everyone trying to buck of openai pffffffffffftttt! Its gambling you can win 95% of the time and still be in the red if you kept it open source we could of all been contributing with money management strategies to come up with a solid system ...im sorry it was great idea but now its just lame!
I have tried it for the last 2 days. Made 5 bets and lost 4 of them all giving 95% or more accuracy. Statistically the strength of its predictions must be off to lose this much.
So its now working
just pick the opposites then :D
@@ronizilla286i laughed at first but I’ve gone 3 for 3
Do you have an android phone app for this or any link to the full tutorial?
i have a question, does this offer consistent predictions? i mean lets say it told me to bet on the celtics to beat GSW then i ask it an hour later it says bet on gsw, because it leverages GPT which is prone to hallucinations/inconsistency at times. Also what would be nice to see an expected value calculation, not only telling us which team will win but if the current odds make it valuable to bet on (based on a positive expected return) given the probability wagergpt calculates
Yes it does offer consistent predictions. Those predictions can and do change as a game nears due to new injuries and lineup data, but it stays consistent. I'm adding +EV this week
awesome, please let me know when +EV i available and ill subscribe for a month to test it out@@SirajRaval
Hey Siraj,
I signed up for a trial, do you place bets based on the ones that have the highest probability?
Trying to understand your bet selection process.
Hey Sam, awesome welcome! That's the idea, the higher the probability, the more AI believes a given team will win.
I can't signup? the link is dead.@@SirajRaval
How about you just put your money into a GIC rather than buying this snake oil
Haha, you got scammed by an obvious scammer who's been called out repeatedly.
I'm trying to do something similar to a game called crash na stake, I've already created an api that returns the data in json but predicting it in a basic way is still difficult, any tips?
Yea, ask an actual expert in comsci not this scammer charging people $200 for chatgpt. This man can’t stop stealing code
Hi Siraj, how can I end the subscription in the first 7 days?
Hey Siraj, tried signing up for the wagergpt but seems like it's not available anymore, can we please know when will you roll out the beta again and if there is any way I can subscribe now?
If you’re just willing throwing away $200 a month, you can throw some my way and I’ll send ya a link to chatgpt
I thought you get the API key from the free trial?
Hi I just sing up for the wage chat gpt There is something wrong. With the home/away team. It’s giving wrong information. When you take explanation. Gives wrong info the away team presented as home team. Please doble check this.
hey i just pushed a fix for this, check again and post back here.
I don't really understand how you use the data from the video analysis
Also, I would recommend charging only a one-time fee for lifetime use of your product as opposed to a monthly fee, OR at least make your monthly fee much smaller.
So you offer affiliate programs? Would love to get this out to your broader audience ?
How can I cancel? pay without paying attention to the price :(
Interesting, I just created one today although I didn’t go this deep and broad. Just selected NBA and loaded historical data for 2 teams that face off late. Fingers crossed!
LFG!!!! lmk if you win
did you hit
He died
Siraj - Second Chance Scamm3r 😂
Also, can you add spreads/total predictions to your bot?
yes, just made a note of this. this will be in an updated version
at 10:24 you say player season stats but your screen shows league feeds and all teams
Tried the telegram bot for tomorrows nba games and there’s an error with home/away teams . Its opposite. Explanations also make mistakes .
noted, fixing this asap tomorrow. expect it fixed in 24-48 hours
Also, how does it deal with players who are Game time decision?
@@thefruituse a third party debit like Koho ;)
Hey some information you should include to make your bot more accurate is , whether overall for that day and also how teams react to certain whether conditions, teams who are fighting for playoffs spots , head to head overall win/lose ratio , opposing skill player match ups (example: how a good wide receiver has matched up against better vs average cornerback defenders, and also players statistical information, like if a player averages a certain amount of points a game , that would determine a lot for the bot to estimate a amount of points scored that game instead of looking at a teams overall performance.
Great points. thanks, i'm constantly updating the data points
Woah. Interesting concept
I'm interested in using this for mines and geology active and non active blm land claim management? With past assays of mineral for states. All low sulfide epithermal zones and hypogene deposits.
i have now lost 4 bet they were my first 4 bets wtf is happening
which bets did you take? DM on telegram as well, i want to see how i can help
Lost all my bets so far. I had only chosen NBA games with over 93% win estimate. Your bot is not working.
the script in your google colab is different from the one chatgpt gave me, are you leaving out any edits as a way of possibly gatekeeping?
Yea he is
Yea he is
Yea he is
Yea he is
Yea he is
when you added in the pre-game data what endpoint did you actually use? As it wasn't any player data nor team (all)
I'm stuck there as well, not sure which endpoint he used.
He's a scammer. Can't give you any useful info.
OMG are you really convinced by a UA-camr that has been found to be a scammer? A fool and his money are easily parted
pretty interesting.. where are you based out of?
There are 15 NBA games from 5 Dec to 7 Dec. You have picked the away team for all of them. 9 of them at >90%. There might be something wrong with your model.
There are still a ton of bugs in the bot and it doesn't appear to be using GPT 4. Also, is your betting strategy to bet a unit or to win a unit? I'm testing this out with straight bets during the trial window and will post overall results - both for betting a unit and betting to win a unit. Just using $2 units for the test and will report back.
i just pushed a major fix this morning, that should fix a lot. it's using GPT-4-Turbo. I bet a unit, please do share your results! also i just added a link to our private telegram community inside the information button of the app
@@SirajRaval Sounds good, I'll check out the fixes. I was still having issues at the time of my post this morning. I'm betting one unit and taking all picks for NBA, NFL, NHL, and EPL. Will report back with results and of betting a unit or betting to win a unit generates better results. Record was 5-5 last night and up 0.875 units with one unit bets. Would have been up 1.25 units betting to win a unit.
@@pmach2039 any update on the performance?
how'd it go?@@pmach2039
@@pmach2039how did it end up performing?
Great video!
Looks like you might have a problem with overfitting, though. Some of those 95%+ predictions don’t pass the smell test - specially the EPL 95% draw predictions.
thanks. yeah Luton in particular keeps having issues, pushing a fix for that later today.
He’s a scammer
ive tried this but chatgpt is always saying it cant help me due to restrictions
wagerGPT is a great alternative that won't say no
I'm growing more and more untrusting of chatgpt everyday, not because its terrible, its extremely good for this type of work; however, I feel more comfortable trying to localize. Have you thought about this? What would be your approach to localize all the chatgpt parts, if possible? Thanks
Good point. the models are all custom, ChatGPT is used to summarize it, and for now its the best. if that changes i'll have the whole thing be DIY. it's all about profitability for the end user at the end of the day
After observing 50 games, its basically 50/50
Basically
50/50 would be nice at plus odds like in the video. Imagine flipping a coin where you either win 25 or lose 10.
Aside from increasing the amount of social media sentiment analysis in your WagerGPT's prediction model, is there anything you can do or you are doing to increase the consistency of a 70+% win rate? Because at this time it seems that some people are experiencing 70% or more win rate and some are getting only 50-60% win rate. With a 50% win rate, you might as well take a random uneducated guess when making the bet and it will be just as effective in terms of probability of winning/losing.
Your youtube comment response bot is interesting, made with open ai gpt as well I bet
I appreciate it, but it's me. UA-cam would ban a comment-bot coming from a verified account.
Your results spreadsheets seem to have some errors e.g. 24/12 you recommended Houston Texans over Cleveland Browns. The bet lost but is recorded as a win. There are lots of other losing bets recorded as wins.
Cleveland Browns were predicted to win with 88.2% probability but Houston Texans were recommended which looks strange
Also, you seem to be quoting ml odds when recommending spread bets. Hopefully these are not used to calculate p/l for spread bets
It would be helpful if you could add a column that shows the p/l for each bet
Thanks for the feedback! Yeah the prediction likelihood was just one data point wagergpt uses to make the prediction. The bet recommendation is the final output. Sometimes the bet recommendation is the opposite of the prediction likelihood, additional real-time data effects the final bet recommendation. I can add that to the FAQ
just enjoying my cup daily coffee
Siraj- you seem sincere, I really want to try this but I literally don’t have $199 to give you right this moment. Wish there was a promo code for 30 days
Thanks Frank! Use the promo code ORIGINALWIZARD to get 30 days free
@@SirajRavaldude thank you so much for this. It’s about time someone created something to level the playing field with the sportbooks. They’ve taken thousands from me! Thanks again for letting me try this!
Hi Siraj, I noticed that you have done similar videos in the past. Have you thought of maybe doing a video series whereby you incrementally improve on the software rather than every time it seems you’re building a new one?
great idea, a whole sports betting playlist
You can ask chatgpt questions and percentages and place the bets urself its not hard I just hit ...ask percise questions and question it if it's not using the recent trends
Salute bro I respect the vision. This is a game-changer. I remember the first video u made about it 10 months ago. I knew u was on to something special. I will stay updated on this
200$ per month ? Are you nuts ???
Man's charging $200/month without even showing any backtesting or P/L proof 😂
😂
What a joker 😂
@@MvniBuiltxI got it for free, I duplicated the bot 😂😂😂
@@MvniBuiltx
I give mine away for zero dollars because it’s obviously in beta still, and they only offtrade i ask is that you join our small community of degenerates so we can compare our bet prompts and earn/loss ratio
I don't get the significance of the reddit ish component. Also, planning to soon make 100K a month from participants in your project takes the fun vibe out of it. Maybe a back end reward for co-development? Not being sarcastic, just thoughts from a pleased subscriber.
If u take the opposition suggestion of the bot most of the time u will win with the cover spread+. Don’t use to win match.
Hey I bought the bot but I still haven’t received the gmail to get access to the bot
this has happened 3-4 times. i'll get you fixed asap. first, check your spam folder. if its not there, email me at hello@sirajraval.com and i'll get you the access code same day
@@SirajRaval I messaged you
Wow.. I'll watch later
I don’t pay for picks. I just need to figure out how to code my own ai bot
I really like, and I definitely see the value in this video, but it’s a little hard to follow you step-by-step when the screen is heavily cropped. I feel as if showing the whole screen and maybe a smaller WebCam would definitely fix this issue
If you believe in this, I have a fantastic bottle of snake oil you might be interested in
Wow we have idiots on the chat, that’s how these people get their money from
it can be done with chatgpt 3.5?
yes but i just updated it to gpt-4-turbo
The thing is terrible. It has me reading that the sharks are at home tonight 12/05. Not sure what that’s all about. Also one of data points was like “team has better record so they are better team” the thing is not good, putting it nicely.
24 hours from now, the update is going to make the explanations a lot better.
Dude, now you tried to explain causality related. So thumbs up.
Always improving, thanks for your consistent feedback rodrigo
doesnt work platform doesn't update
How about trying your skill on horse race prediction?
I keep hearing about this. I will add horse race prediction to the bot in a future update. im starting to realize how popular it is.
Yes, horse racing. The recent gpt web and upload capabilities that you use is perfect for immediate data interpretation! @@SirajRaval
to cancel you have to call your bank remove all your cards each single acount also you need to remove all the tokens merchant linked to you'r acount so he won't be able to charge you make sure they remove all the tokens merchan linked to you'r acount
Have you thought about making a poker bot, It would make money while people are sleeping.
Interesting, but unless you post the profit and loss. Its kinda hard to a partake with our $$$. I understand the cost involved in making this. You dont learn this skills overnight. But for us to believe you that this is not just another service that other people are already providing. You need to provide evidence that this works in the long term. In any case, keep up the good work. Watching with keen interest on the side line.
P&L available as of today, i'm planning on more data and better performance every week wagergpt.co/reports
explanations are still off and it doesnt even show the NBA games for tomorrow
This sounds amazing but where is the long-term P/L of the bot? have you even backtested it?
Why would people sign up to an unproven BOT to burn their own money?
I’ve been using it myself since February and it has had at least a 70 percent win rate for me. The idea is that with the right bets you win much more than the subscription, like what I did in the video. I agree that I should map out the P/L even further but I think user testimonials will speak for itself. I’m confident that people will love this bot as I do. Plus it’s a 7 day free trial so just cancel it on day 6 if you don’t like it.
70% win rate at $1.20 odds wouldn't be profitable. P&L is the only relevant data point, not win rate.
@@SirajRaval The win rate is irrelevant. You could be winning 90% of bets at 1.000001 odds.
This would be very dangerous to post to a huge following with no real proof of it being profitable.
Please post P/L since February showing stake, profit loss, ROI etc
Alright I’ll gather those stats and make a report tomorrow. I’ll post it in the video description and the website. Thanks for letting me know.
I just tried this for the Green Bay game tonight. It thinks Aaron Rodgers is in the team with the explanation. How then is this bot accurate ?
Gonna ask Coffee Zilla to take out the new video too?
Unbelievable! ChatGPT takes away the pain of mapping data.
When I look at the PL, then there were 11 wins, 9 losses so far, and that seems to be the most important fact to me. Now one needed so many bets over a longer timeframe, that it is statistically relevant. Its a superinteresting experiment! And then there might be additional factors to classify a teams strength, which you might add over time. Also, you might announce your bets pre-game, until the bot is proven to be successful.
Any updates?
Yea I don’t want to public post but would like to chat more if you can. A handful of explanations are not relevant I don’t know if that is affecting the bot
alright sounds good please message me privately
@@SirajRavalhow can we private chat ?
what camera do you use? Amazing quality
I turned $43 to $114 in a few hours and winning 4/5 bets using this ai got so far. I would’ve won more but I was skeptical since it was picking towards the underdogs more frequently. So I waited until halftime to see whether it was a smart choice. I did this with 3 of the bets and still made a decent amount. Just wanted to say thanks and if I make over 500 dollars profit durning my free trail, I’ll probably be paying the 200 dollars.
Did u end up paying the amount?
Hi! Would you mind to tell me - did you earn more money from this bot?
hey what you earned in 3 months
If the AI were any good, it should not matter whether the picks were underdogs. Don't bet on long term success.
I want to like this I really do, but the bot makes very little sense most of the time. It changes its mind on predictions, will say one team is likely to win for prediction at 80%, and then when you press Ai bet suggestion it’ll say to pick the opposite team. Or you click it and it starts giving you a pick on a completely different game. When I can actually get it to make sense, it makes good picks
Great feedback and i agree with you. Hallucinations must be solved, i am actively working on this, it's quite difficult but i know its solvable
@@SirajRaval completely understandable. At the moment I am just trying to tread lightly, work out what the bot likes as some of the Ai picks like Top 5 will contradict the manual prediction, so just seeing slowly what works currently. I do think a discord server would be great for this kind of thing also for feedback, chatting etc. And to probably lower the price until common bugs get fixed and it starts working more reliably, just some thoughts. Other than that I’m a huge fan of this kind of thing
Where is the Gambling Addiction Notice... This is advertising a gambling product and the notice should be provided.
At least 1000 bets is needed to confim the ststistical reliability. I am very experienced sports bettor.
Here before comments are disabled.
Chatbot generates a 70% win rate, but wanted to “keep it close to the chest”
…..(creates a TY video to sell a $200/mo product)
Lol.
I bet 10,000 but took it out due to some of the people in the community. im going to start low and put my findings on here tomorrow
What findings did you get?
@@PartialGangthe bot doesn’t work as well as advertised. I’d avoid it. I went 5-25
@@Djr2699 I see, I’m testing the trial currently and testing it rn with the games on, I’ve avoided the AI top 5 picks etc and just selecting individual games on their win % and betting off that
@@PartialGangquestion how have your bets gone? Since you where betting on there % I was curious to know how you did?
@@Jaycookie-fq4eq mixed tbh, it’s design is very good, straight forward and nice, but lacks consistency within its own model. I know sports very well anyway and do my own research and use this to back up my own claims mostly, with it I’ve made 10U so far this week, which is nice, but with just selected picks, biggest problem is it changing it’s mind, I’ll click a game like today, for NCAAB, and it will tell me the home team will win 87% chance, then later I check again and it’s changed to the away team at 66%, which makes it hard to follow, when it works it does very well though, just needs improvement which ik he is doing, idk how much time he’s putting into it though
Been using this bot a fair amount past week on trial. Liking it a lot, some flaws but on the most part good. If improving this is something you plan on doing I’d love to get in contact to give you daily detailed feedback on bugs and other stuff and happily be a beta tester, definitely has great potential
hey PartialGang, thanks for your feedback. Let's definitely chat, you should join our private member only telegram group. the link is in your email, DM me there!
$199 a month? LOL theres hundreds of people on twitter offering free parlay picks using ai.
Sorry, but I see several RED FLAGS here:
- We are missing any kind of long term validation over more than 2 bets. Why didnt you just split the money into 20+ bets to get a more sound sample to evaluate its performance?
- Apparently you have 0 knowledge about the NBA, as you mention "Penalties" when talking about pre-game stats
- in you social-crawling a text withe the Houston Texans was displayed, which is a NFL and not a NBA team, even though you typed in that you would search for the Lakers. Generally all the social contents seem to be from the NFL. looking at your snippets. => The code snippets that you show right after these NFL posts are then including "celtics" again, which is a team name from the NBA.
- charging 200$ for this without proper long-term evaluation and the other mentioned red flags is a big no no for me, I am sure you are great with chatgpt, but charging 200$ after a video like this.
1. The profit and loss statement is available as of today, i'm planning on more data and better performance every week wagergpt.co/reports. 2. You're right, i am going to educate myself a lot more about each sport, and i'm now incentivized too. i never claimed to be a sports fan. 3. $200 will make sense once more people comment on their success, everything depends on the performance of the model
You are a genius my friend!!! I made the payment, just waiting for the email and info.
Coffeezilla already owned you
Requesting a football prediction plz
So I tried it, and here is the truth: It was a 7-day trial, which means I bet every day, and it was awful. Out of about 10 bets, only 3-4 were correct. The bot suggests bets that are a joke, and you think something went wrong with it, but no, the bot just gambles away all your money. The $200 per month fee has to be a joke.
shit im paying this no problem. hes basically giving this tool out for free
Hello conman
gotta make those quick bucks before getting exposed again ☕️
Siraj proved himself to silence his critics. I offered him free advices to improve his bot accurancy. Since the bot is totally out of reality with claiming 99.9% confidence to one team wins in its predictions multiple times. When I messaged to his group, Siraj kicked me out in a second. So thanks man, I was your fan. I did not believe videos about you scamming people. Now I see that it has some realistic basement. Guys safe yourself money. 200 usd per month is making someone rich for no proven statistical record. Maybe he will block me again here:)
i'm not silencing anyone and i haven't "blocked" you. See the many comments that have both positive and negative things to say. if someone cancels their subscription, they will be removed from our private group chat, its unfair to paying customers. Lastly, the bot is constantly improving and no longer makes predictions over 97%.
It’s 50/50 most of the time. It’s not 70% winning rate.
Hi, are you originally an indian?
yes, i was born in houston texas USA but my parents are indian. i have spent 6 months traveling in india as well and love the culture.
also, why not use GPT 4
Great point, simply the cost of token generation but v2 will use GPT-4
it now uses gpt-4 , just pushed a fix
In some games it gives explanations for different games. for example:
Wolverhampton Wanderers / Burnley
Dec 05, 2023, 19:30 UTC
Prediction:
Burnley are going to win: 95.0% likelihood
Explanations:
1. The Penguins will have a higher win probability due to their strong performance this season.
2. Statistical analysis favors the Penguins based on their recent head-to-head matchups with the Flyers.
3. The Penguins' odds reflect their status as the slightly more favored team, despite the Flyers having home advantage.
I just pushed a fix for the explanation bug. It was hallucinating, but we can't be letting that happen.
Is the prediction correct, even if the explanation is wrong?@@SirajRaval
yes exactly. sometimes it hallucinates but i set the temperature to a better threshold to avoid that. @@user-hu9bf2ie4z
Sorry, but I see several RED FLAGS here:
- We are missing any kind of long term validation over more than 2 bets. Why didnt you just split the money into 20+ bets to get a more sound sample to evaluate its performance?
- Apparently you have 0 knowledge about the NBA, as you mention "Penalties" when talking about pre-game stats
- in you social-crawling a text withe the Houston Texans was displayed, which is a NFL and not a NBA team, even though you typed in that you would search for the Lakers. Generally all the social contents seem to be from the NFL. looking at your snippets. => The code snippets that you show right after these NFL posts are then including "celtics" again, which is a team name from the NBA.
- charging 200$ for this without proper long-term evaluation and the other mentioned red flags is a big no no for me, I am sure you are great with chatgpt, but charging 200$ after a video like this.
Great way to make people fool
hope i see you one day. coffee will be on you, s c u m b a g
Did he notice that even spelling certain words incorrectly, chatGPT understood that it was a typo and gave the correct answer anyway.
yeah i've started to be less stringent on typing accuracy, since chatGPT is really good at understanding prompts despite typos.
once a scammer, always a scammer.
Do you know how fucking unreasonable a 70% correct rate is in any sports gambling situation.
Such a scam.