My value betting experience: 100,000 euros for 4 months

[777games]

Friends, hello everyone!
Sorry for my English

We are the Almanacc team, I create this topic in order to share my experience in Value. I would also like to find interested people with whom I can exchange experiences and some interesting developments in the field of Value Betting.

All our work is based on work with the Bet365 bookmaker. I will start with the history of working with this bookmaker. In February 2019, our team collected a 101,000-match dataset from the forks of the Positivebet scanner. Further, this dataset was analyzed in detail, analysis was performed in Python using the standard libraries pandas, numpy, scipy, plotly.

In the course of our work, we wrote our own basic functions that facilitated the further analysis of such a large dataset. (A little later I will show you an example from our dataset)

After a preliminary analysis, we began to build a mathematical model using a genetic algorithm to optimize and select parameters. One of the parameters was selected using the xgboost algorithm. (I can tell in more detail if this is interesting to someone)

From the results, the algorithm for 4 months earned 100,000 euros. The financial strategy assumed a 2 percent flat (2 percent were derived experimentally, show the best combination of ROI, maximum drawdown)

The average ROI for the period from April to July was 7 percent. I can give a list of bookmakers with whom Value was in an individual communication. Below I will attach the MeanValue chart for the period from April to July 2019

I would like to talk with teams or professionals who also work deeply with Value on the topic of exchange of experience.
I wish you success and victories!

 

[grandvictoria]

he weapon used to beat your data based adversaries is using even more data and it seems obvious to use a data source that actually depicts indiscrepancies such as an arb/value-scanner. So congratulations with your results and this very quantitative approach!

I am unable to understand your use of bet365 limited accounts. You state that all your work is based on the bet365 bookmaker and in the answer you confirm that these are limited accounts. If you were to make 100.000 euros out of that, I assume you must be a large team have many accounts, not just one? Or is it true that bet365 would let you do that on only few accounts?
I programmed a bot which I ran last year on my limited bet365 account. It just made bets on anything, but wasn't very succesful though (ROI 0.25%, 10k bets), probably because I did not use any kinds of filtering. I found the event useful though, as bet365 now acts as reference in some occasions. Even if I had had 7% ROI, I hardly believe that bet365 would allow me to reach such a profit from small limited stakes - if I was a bookie I wouldn't. And how can you use 2 percent flat strategy on a limited account?

But maybe I have just misunderstood something. Maybe it is all your work that is based on the bet365 bookmaker being your reference and bets are placed at other bookmakers, softs, sharps, exchanges?

Anyways, thank you for this post - it is really admirable to see such a data based dedication. I might send you a message for exchange of experiences. My approach is also quite data based, but not comparable with the scale of your project. I have decided to move the execution part exclusively into shop betting, so that might differ much from your procedures.