PALMTREE STATISTICS
Digitalization: From manual tasks to Bots
Palmtree Statistics is a Sport Trading company specialized in Football. In their early days, every trader was helped by an assistant. While the trader was focus on the matches and placing bets, the assistant was providing statistics of the matches using some python scripts or excel sheets. This approach limited their scalability because:
-
Assistant could run one algorithm per match at a time
-
The assistant response was not as fast as they need
-
Every trader could track ~5 matches
AUTOMATION
The first step was to create a platform to automate tasks of assistant.
The platform, called Pelota, was able to track every professional football match around the globe with a very low latency, mixing data from different external sources. That was a really big achievement, but the key part was running their machine learning algorithms in real time per football match. This fact improved the productivity of the traders and assistants promoted to junior traders with the help of Pelota.
Every trader had the possibility of calculating matches odds customizing parameters in the algorithms. This feature made the difference between Senior traders inside and outside the organization.
Finally, we also added a tool to compare how good were the bets of every trader, in that way they could learn how to improve according to the situation or from other traders with more success. After all the improvements, senior traders could follow up to 20 matches at the same time.
BOTS
Pelota was really powerful in terms of providing odds of football matches and helped traders to make decisions. Next step was really clear in the agenda, building bots. Platform provided very important data for human traders therefore, we should develop bots that were able to make decisions like the human traders did. We developed a farm of bots where every bot was focused on different kind of bets, hence, different strategies. They were trained using an offline approach.
ARCHITECTURE
