Round 1 has begun. The algo trading portion is nearly identical to the tutorial round, just using different day’s data. One inconvenient change is that Round 1 only provides log information without PnL reporting, so I have to calculate my profits manually from the logs – quite troublesome.
As a result, I didn’t make any major modifications, just did some parameter tweaking based on my log-estimated PnL. The only interesting observation was that the organizers provided three days of trading data along with their bots’ PnL. I noticed that across all three days:
⦁ The bots consistently made/lost just a few hundred dollars on pearls
⦁ But earned 10,000-20,000 on bananas
This suggests:
⦁ The pearls bots might be trading randomly without additional information
⦁ The bananas bots appear to have some price prediction capability
If this hypothesis holds, I should consider how to leverage this information in my trading strategy – though I haven’t figured out how yet.
On another front, the manual trades presented an interesting forex arbitrage problem:
⦁ 5 total transactions involving 6 currencies
⦁ Both start and end with “shell” currency
⦁ Only the middle 4 currencies can be exchanged
⦁ Each currency has just 4 possible exchange options
⦁ Total possibilities: 4^4 = 256 combinations
⦁ The optimal sequence was 412314, yielding 1.0436 profit
