Hi everyone! I would like to discuss the TC score in the numerai tournament with the community.
I started the numerai tournament from this January. This tournament was exciting to test my machine learning skills. It’s been very successful in the last four months and I’ve been able to win some medals. However, after the introduction of v4 data, the TC score has deteriorated significantly.
Specifically, the percentile of TC scores from rounds 303 to 308 was 72 to 92, which was very good. However, from rounds 309 to 314, it has deteriorated significantly from 0.6 to 26.
During this period, the model has been trained on v3 data and has not changed. However, why is the tc score so fluctuating so far?
(1) Is it a problem with the versatility of the model which I trained?
(2) Is it due to this recent market downturn?
(3) Is it because I am using a model trained with v3 data? Need a model trained with v4 data?
I’ve doubled the TC in score multipliers because I’ve had good the TC scores so far (rounds 303 to 308). Of course, I think it’s my responsibility that I haven’t managed my risk, but is it good for tournament participants to face such a fluctuating TC score?
Even if you think alone, there are limits, so could you please let us know what you think? Thank you.