#1 Goal – predict when bookies get their odds wrong. If we can do that, we can take advantage of “miss pricing” in football betting, as well as any sport of your choosing. In this video, we explore getting some football data with the odds from other bookies and running that through a supervised Machine Learning model.
You will notice that we get a 6% edge when doing so. This is very exciting and warrants further investigation for sure.
Can we profit with the same method in any other areas? For example, horse racing? If so, where would you recommend we pull the data from (or perhaps we need to collect our own)?
website:
Link do Vídeo
Doesn't machine learning need to include the game stats for a comprehensive mathematical and up-to-date version of data on which to eventually provide a decision as to the likely outcome of future matches?
no point in doing this, bookies will kick you out if they see that you have an edge. You must beat Pinnacle or Betfair Exchange's odds if you want to make money long-term
Please add the ml tools that you use
What are the softwares and cites you use in video?
You are assuming odds of fav are 2.0 or less. They are not. Often, the fav is as high as 2.6+. Poof, there goes your edge
The problem with bookies is that if you start winning ,you’ll get your account gubbed. Good news is that you might beat exchange odds.
THANK YOU.
There are three possible outcomes so this is useless
Awesome video, would love to be able to automate this ML and pull from multiple sites and compare expert picks and then dump into a Power BI report. Keep up the great videos and if you have any information on which sites are the best to scrape expert picks or predictions from I’d love to know! Thanks!
I wonder if the process is related to neural network?
head pain
There are usually 3 outcomes, but you narrowed it down to two – Which means its home win vs away win or draw. The terminology for picking a team to win or draw is called double chance, the odds for double chance are much lower. If the predictions are 54% accurate, I guess that means you need the bookies odds for double chance to be greater than a 54% chance, ie 1 / 0.54 or 1.85 for it to be worth betting on. So if the bookie offers 2.00 for away win or draw, and the model predicts not a home win, its worth the gamble!
That's how I'm seeing it anyway. Great video!
This no longer appears to be available so don't join up thinking it is.
Hello, great video! I'd like to know, if the FTR column wasn't given how can I possibly make excel compute it for me automatically?
great video! I'm a 3rd year cse student and wanted to know is any part of your platform free to use for educational purposes. for example I would just like to some simple ml models using your interface.
This is the first time I come across this channel and I really liked the video, which is greatly structured and there is a lot to learn from. However, the results are wrong and meaningless, because of one small error you've made at the beginning. The idea to narrow down the prediction to just Home and Not Home is great, but you are wrong about the bookmaker's prediction. You are checking if the bookies put the home odds to be the lowest. So, let's say the odds were the following: Home – 2.50, Draw – 3.00, Away – 3.00. Then you assume that the bookies are favouring the home for the win (but you forget to turn the bookmaker's odds to only two outcomes). However, if you convert the odds to percentages, to bookmakers give around 40% chance for Home and around 60% chance for Not Home. If the game finishes Not Home, you would have assumed that the bookmaker is wrong, but it's actually correct. In this particular case, if you have made a bet for Not Home, you would have bet on odds around 1.60 (this should be the combined odds for Draw and Away). With such odds you would be at break even with success rate of 62%, which means 56% accuracy doesn't give any edge. So, the results from this experiment are a bit meaningless, but congrats on the effort 🙂
What you have done seems pretty correct. I have tested it with RandomForest and got about 53% precision (sure, there might be something better).
What my issue is the mean odds, when the bookie was wrong, sit at 2.04.
Which means the odds for our bet will be below 2.00 after their spread. Meaning we are about break-even or so, which is still pretty decent result though.
I love your videos.. certainly planing on subscribing to crypto wizard😘.. Can one actually make money from the crypto wizard cos that’s the major aim at least one should be able to meet up with the monthly subscription on the crypto wizard.. secondly, I don’t understand how you did the calculations on the target cos you used the final score to determine the Target( how do we now place bet?).. I’m not too clear about it.. sorry if I bother you😢🙏🏻🙏🏻
Hello, is it an EPL? what about the draws? Did they have those? In the description they have FTR = Full Time Result (H=Home Win, D=Draw, A=Away Win). Try to open 2020-2021 sheet, they have draws as well. This might affect logics dramatically when you compare actual results with betting odds
Amazing. But. You are not getting the edge if you are right 0.56 times out of 1 unless the average odd you bet on is greater than 1/0.56 which is ~1,78 for this percentage. However the video is great