team-045

Our approach

We did not implement the most complex mathematical formula, as the market is quite unpredictable, and our knowledge was too low to find a working complex algorithm in that short time. In addition, the computational effort was not worth it (at least for us), in the sense of performance, as one has to react quite fast and make a decision on what to do. Our basic principle is similar to regular stock markets (sell baskets expensive and purchase stocks cheap). In every iteration, we firstly evaluate the price history surplus stocks and baskets, then we search for a "good trade" (a trade that makes profit). Once we find such a trade, we execute it. This reduces our risk as we only execute definitely positive trades. After this execution, we perform the hedging to ensure our ratio is correct (not only after each trade). The iteration speed was adapted to approximate the maximum requests of 25 per second, which makes our algorithm quite fast.

Challenges

We first tried to solve the issue with a very complex mathematical formula. That went a little too far and lead to a lot of problems. Then we decided to kind of restart with a simpler approach that could lay a base. This went a lot better. It was though still difficult to test if our approach worked (especially at night when not that many other traders were online). As other teams often did weird transactions, so you could not assure if an increase of PnL was source by a change from us or if it was the result of another team making bizarre things.

Built With

• cloud9
• python