Quantopian's community platform is shutting down. Please read this post for more information and download your code.
Back to Community
How do you know if you have a good algorithm?

So I have my VIX bot written and it's feature complete. It relies on VIX etfs, so it can't really backtest further back than late 2012.

Back testing from Jan 1 2013 to yesterday, it would have been up 465.4% with 100k capital.

Alpha 1.22
Beta 0.54
Sharpe 5.45
Sortino 7.98
Information Ratio 4.78
Volatility 0.24
Max drawdown 13.6%

When comparing returns to drawdown, I think this has beaten anything else I've seen on these forums, but I'm wondering if maybe people have algos that do quite a bit better with less risk and they just haven't shared them since they realized they have the real deal. I'm also concerned if that that timeframe is insufficient, since we really haven't seen the VIX go up to anything that would threaten a short-the-vix strategy that badly.

How long should I let this thing paper trade for before moving to real money? I feel like the only thing it is untested with is true market turmoil, or black swans, which we may not see anytime soon.

1 response

In a bull market, going long SPY, or conversely, short the VIX, is a no brainer, right?

Depending on what instruments your algo "trades", you may find that those are hard to borrow, so your stategy might not be implementable. If you are good at options, you can try to do a synthetic short, but it will cost you in transactions, slippage, liquidity risk, and time to manage.

"better with less risk and they just haven't shared them since they realized they have the real deal"

I am sure some exist, given the law of large numbers. There are 80K "developers" here, and at one per developer, that is 80K algos, of which 1% should be very profitable on a consistent basis, either due to skill or luck. There is no "real deal" though, since backtesting plus out of sample performance for a couple of years does not guarantee future performance whatsoever.