Back to Community
It's all about that META: An idea to crowd-source the contest scoring system

Quantopian currently defines a scoring system based on a set of metrics and requirements. Next, algo writers optimize their code to maximize their score based on this criteria. The goal is that Q will "guide" the algo creators into creating a return stream that matches what Q's investors desire.

What if a second layer, a meta layer, was added to the contest? The community would submit custom scoring systems that rank the submitted algorithms. Then these meta systems would paper trade the top 10% (or similar) of algos, during the same contest period that the existing contest uses. Then the meta portfolio with the highest return at the end of the contest would win the meta contest.

Would this be helpful? The positives are that the community can unleash our creativity in another avenue, and perhaps the best will rise to the top. The downsides are that the return streams may not match what Q is looking for (i.e. low volatility).

Any thoughts?

5 responses

+1.5 * 2 / (1 + 2)

Hmm, interesting idea, but I don't think you can score the scoring systems by highest return. We need some sort of balance of a variety of metrics representing return, risk and risk-adjusted return, to judge the out-of-sample performance of the scoring system on the algos.

@Gary: Not sure what you mean by that.

@Simon: You are saying that we would need Q to create a scoring system to rank the community scoring systems? ;)

That was my +1 for your idea. Was so inspired that the scoring circuits in my mind were activated and it turned out to be a somewhat involved +1.   :)     (have to do the math)

Assuming the data has been kept, you could backtest scoring systems also.