How to earn money with machine learning and sports betting
Read this article to see how to achieve up to over 20% ROI per bet
In my previous article about sports betting and machine learning I already described how I technically approached the problem of predicting the outcomes of sports events. After one year of hard work, I’m proud to present https://sports-ai.dev to you. On this website, you can find predictions not only for football matches, but also for basketball, handball, volleyball, hockey and tennis events.
Intro
I won’t go too much into detail about the technical stuff at this point, for that, please refer to my previous article about that topic. That’s because I want to focus on the potential earnings from my machine learning models and secondly, I don’t want to share too much information about my models in order to keep them my “business secret”.
In addition to my previous article I can tell, that I have developed own, separate models for each market and each sport. Additionally, I managed to successfully combine several models to one accurate voting classifier. So far, I have developed predictive models for the moneyline, asian handicap and totals market.
In this blog post, I want to present the results you could have achieved with the predictions of https://sports-ai.dev in the football market from February 1st until April 9th.
How to make profits with sports-ai.dev
In the end, it’s all about the prices that are offered to you. If you approximately know the probability of a certain outcome of a sport event, you can also determine a fair price for that outcome. A classic example for that is a coin toss with fair odds of 2.00 (2.00 = 1/0.5 with 0.5 being the probability of the outcome). If you now get a price above these fair odds of 2.00, you have successfully found a thing called value bet. The value results from the fact hat odds above 2.00 are overpriced. Applied to the coin toss example it means that if you for instance get odds of 2.10 for a betting on heads, you will in the long run pretty sure make some decent profit.
With the algorithms I have developed, you can approximately get the probability of the outcomes of sports events. Of course, you cannot always be right, but if you find opportunities for value bets often enough, you will make some decent profit in the long run.
Get up to 27% ROI per bet
In the underlying dataset 17.800 football matches between the time from February 1st until April 9th are included. Here’s a little example of how you would take a betting decision. Let’s take the match Tigres vs. Patriotas from the Copa Colombia played on 6th of April. Sports AIs algorithms predicted a probability of 40,48% Tigres winning the match. If you transform the probability to an odds format (1/0,4048) you get fair odds of 2,47. William Hill for example offered a price of 2,75 for a home win of the Tigres, which leads to an edge of approximately 11%. In such a case, it is advisable to bet on this outcome at that price of 2,75 offered by William Hill.
If you apply this strategy to the period under consideration, then following results would have been achieved. Notice, that in this example only bets on the home team winning in a football match were placed.
So the chart above shows the profit development for bets with a stake of 100€ per bet. In that particular scenario bets werde placed, when an edge of at least 3,00% was found. That resulted in a ROI of 13% per bet. The higher the edge needs to be in order to place a bet, the higher the ROI will probably be. Take a look at the charts below.
You can see that there is a positive correlation between the minimum edge of a bet (called threshold on the right side) and the ROI. The downside though, is that the amount of matches meeting the criteria get lower the higher the threshold is set. With a threshold of 14% we achieved a ROI 27,62%!
Conclusion
If you are curious about the predictions and want to see more content about this topic, then feel free to visit Sports AI at https://www.sports-ai.dev.
