[Journey] Predicting Sports Results with Artificial Intelligence

Cool one , I've been in edge betting for a while. http://tradematesports.com/ (I'm not affiliated in any way), I used this one for a while, but on what is your program based on?

Unreliable Witness said:

Brilliant use of software.

Be aware that betting companies will ban you if you are too successful. Their businesses are out to take money from you, not to let you make money.

Click to expand...

Asian bookies don't, they only care about revenue.

Interesting but you should keep this to yourself until you are 100% sure it works.
If it works then there are numerous opportunities for you.

Asif WILSON Khan said:

Interesting but you should keep this to yourself until you are 100% sure it works.
If it works then there are numerous opportunities for you.

Click to expand...

If it works, he shouldn't sell it, he got to keep it for himself then.

Quite different than the other journies. best of luck

Only$$$ said:

If it works, he shouldn't sell it, he got to keep it for himself then.

Click to expand...

No, he shouldn't sell it but he will need lots of people to lay on bets.

LOVE this kind of stuff, please keep this thread alive!

rancidricky said:

Very interesting project! I know zilch about AI, so does it like learn as it bets, and change its strategy? btw have you checked out OpenAI gym?

Click to expand...

Thanks! Yes it predicts a score, calculates the difference to the real score and adjusts accordingly. Never heard of that, will have a look at it tomorrow, thanks

LordTony said:

Work smarter not harder! Seems like an amazing idea, good luck. Will be following.

Click to expand...

Thank you!

hokuspokus said:

If you get 7.5 in return per game you would had to have 134 games correct to make even.

Let's say you instead was playing over/under 2.5 goals and that the rerurn is 2.0 per game, then you would had to have 523 games correct, would this be more profitable for your merhod?

Click to expand...

I'll try out different betting strategies/methods in the coming days/weeks and will report about my results, thanks for the input anyway

Carterm said:

This journey will be different and unique. really will follow it.

Click to expand...

Thanks and hope you'll be entertained on the way

Only$$$ said:

Cool one , I've been in edge betting for a while. http://tradematesports.com/ (I'm not affiliated in any way), I used this one for a while, but on what is your program based on?


Asian bookies don't, they only care about revenue.

Click to expand...

Which results did you have so far with this site?

My tool is currently based on a teams previous scores and performances.

Asif WILSON Khan said:

Interesting but you should keep this to yourself until you are 100% sure it works.
If it works then there are numerous opportunities for you.

Click to expand...

Thanks but why do you think I should keep this to myself?

DentalMidget said:

Quite different than the other journies. best of luck

Click to expand...

Thank you!

malvor said:

LOVE this kind of stuff, please keep this thread alive!

Click to expand...

Thanks man I'll try to do so

issorc said:

Based on the predictions I placed another two bets for today:

1. Bundesliga - Hertha BSC 2:1 Bayer Leverkusen
Serie A - Cagliari 3:1 Sassuolo

Please note, I would never bet some of the games I'm currently betting, but it is what the program is predicting. My main focus was to remove feelings, emotions and other stuff out of the process and make predictions based on data and stats only. Maybe in a few weeks I'll change how I place bets but currently I'm just doing what my tool is telling me to do.

Click to expand...

Damn you got the first one exactly :O

iSmexy said:

Damn you got the first one exactly :O

Click to expand...

Saved me a small profit for today, will update tomorrow with all bets from today and detailed stats

Best thread on BHW, without a question, wish to see the day that lots of threads on this forum are like this

issorc said:

Saved me a small profit for today, will update tomorrow with all bets from today and detailed stats

Click to expand...

Would predicting which player's to score or assist be difficult to implement ?

Cititechno said:

Best thread on BHW, without a question, wish to see the day that lots of threads on this forum are like this

Click to expand...

Thank you for your kind words!

iSmexy said:

Would predicting which player's to score or assist be difficult to implement ?

Click to expand...

In the current version I guess that would be actually pretty hard. I currently do not work with any data regarding players.


Yesterday's prediction and the real result:
2. Bundesliga - Heidenheim : Darmstadt 0:0 | 2:2 -> Lost - -0,2€
Serie A - Lazio : Neapel 1:0 | 1:4 -> Lost -0,2€
1. Bundesliga - Hertha BSC : Bayer Leverkusen 2:1 | 2:1 -> Won +1,89€
1. Bundesliga - 1. FC Köln : Eintracht Frankfurt 0:2 | 0:1 (Forgot that one yesterday) -> Lost -0,2€
Serie A - Cagliari : Sassuolo 3:1 | 0:1 -> Lost -0,2€

So 1/5 yesterday. 2 games were off by 4 goals, 1 game by 3 goals and second game in the bundesliga was pretty close.

Overall I'm left with 1,09€, but I had another system bet I actually nearly placed accidentally and later cashed out to decrease loss but still lost 0,59€.
That means that in total I made 0,4€ yesterday.

Interesting journey.

A few questions:

• Which parameters did you select as the input?
  
• Do you takes into account the sequences of the matches? I mean, do you implemented a RNN like LSTM to include bad or good series of teams in the predictions?

Thanks a lot

I've never seen a bot succeed in this area but I wish you luck

You said that your tool has around 25% accuracy, what percentage are you currently at in practice?
Also, from the odds that the tool has predicted correctly, in comparison to the average odd payout (so if you have 4/1, 6/1, 5/1 the avg would be 5/1), how close or wide do they defer?
That would be interesting to track and I would love your input on it

This is interesting, but you will have to make at least 1000 bets before judging if your ANN is predicting correctly.

What I understood that your bot only predicts the winner and the score. I think it would be wiser to you and the bot to predict the percentages of winnings like 23% 40% and 37% (home-draw-away). Then you compare those odds to all the bookmakers and make a bet if it's profitable. That's what the professionals do.

Anyway interesting thread and journey! Goodluck!

Never seen this before. Brilliant !

You should be more careful when you have higher accuracy. These betting companies are not gonna accept losing money.

ju111 said:

Interesting journey.

A few questions:

• Which parameters did you select as the input?
  
• Do you takes into account the sequences of the matches? I mean, do you implemented a RNN like LSTM to include bad or good series of teams in the predictions?

Thanks a lot

Click to expand...

Input until now mainly wasd the goals the teams shot in the previous games and their position in the table. I'm currently working on a new version with more features (right now the total numer of features is 22, it will be 80+ afterwards). I'll update about this once I'm done.
Regarding the sequences, I'm feeding the tool the last results so yes it knows (or should know) whether a team is on a good or bad streak

everythingred said:

I've never seen a bot succeed in this area but I wish you luck

Click to expand...

Thanks We'll see

ScribScribScrib said:

You said that your tool has around 25% accuracy, what percentage are you currently at in practice?
Also, from the odds that the tool has predicted correctly, in comparison to the average odd payout (so if you have 4/1, 6/1, 5/1 the avg would be 5/1), how close or wide do they defer?
That would be interesting to track and I would love your input on it

Click to expand...

So far 9 games have been played and I won 4. So can't really say anything about practice because 9 games are way too few for any conclusions.
And I do not really understand your second question completely sorry. My tool does only predict a score but does not tell me how likely it is that this result occurs. So I do not have any odds for that, I just bet what it predicts. No matter of the odds at the bookmaker are 8.00 or 17.00. If that's not what you meant I'm sorry and maybe you could explain again what you mean and I'm happy to answer it.

MacRichard said:

This is interesting, but you will have to make at least 1000 bets before judging if your ANN is predicting correctly.

What I understood that your bot only predicts the winner and the score. I think it would be wiser to you and the bot to predict the percentages of winnings like 23% 40% and 37% (home-draw-away). Then you compare those odds to all the bookmakers and make a bet if it's profitable. That's what the professionals do.

Anyway interesting thread and journey! Goodluck!

Click to expand...

Yes it's way too early for any judgements about the prediction quality but I'm working on it and I'm trying to place as much bets as possible to identify how good it works.
I'll take that into consideration but for these predictions I'd need to change how the tool works so I'll implement it in a few weeks maybe and see how good it works. Thanks

LightningKing said:

Never seen this before. Brilliant !

You should be more careful when you have higher accuracy. These betting companies are not gonna accept losing money.

Click to expand...

I know, currently I'm only burning one betting account (in case it should work out great the next weeks). As long as I'm only limited on one I still can come up with a tactic to make the most out of it but it's too early for that. I won't spend time thinking about this as long as I don't have a proven money making straategy

Thanks for all the inputs so far guys! I'll most likely update with my bets for the weekend later!

As promised, the bets for the weekend My new model is working so far (still need to implement a BTTS prediction) but I'm not able to use it yet. It needs for each team either the last 3 home or away results and most leagues are not that far into the season. So nearly all predictions are still from the old model. If it continues to bring good results I might use it to crosscheck results with the new model.

Ligue 1:
Marseille 0:2 Toulouse

Danish Superliga:
AC Horsens 0:1 FC Helsingor
Hobro IK 1:0 Brondby
AGF Aarhus 0:1 Midtjylland

Chinese super league:
Yanbian Fude FC : Liaoning Hongyun - O2.5

Croatian First Football League:
HNK Cibalia 3:1 Inter Zapresic
Istra 1961 0:0 Dinamo Zagreb

Irish Airtricity League:
Bohemians Dublin 2:0 St Patricks

Czech First League:
Slovan Liberec 1:0 Slovacko

Russian First League:
Rostow 1:0 Lokomotive Moskau

Spanish Second League:
Granada 1:0 Cordoba

English Premier League:
Arsenal 2:0 West Bromwich
Leicester 0:2 Liverpool


Here a few more predictions I did with my new model:

English Championship:
Aston Villa 2:0 Nottingham Forest

English League 1:
Oxford United 1:0 Walsall

English League 2:
Rotherham 2:0 Oldham
Barnet 3-1 Crawley Town

I'm not sure if I'll update every day, maybe only once after the weekend. If you have any questions, feel free to ask

Which tool/framework do you use?