Long short statistical arbitrage on Cryptocurrencies - Algo for sale

I ran my model on 20 cryptocurrencies from Feb 2016 to Feb 2018 and here are the results:

I am ready to sell this algorithm if there are people interested. Min bid: 100K USD.

You can adapt this algorithm to equities as well.

Some details of the algorithm:

1. Long Short (using margin)
2. Dollar neutral (your upfront cost is only the margin)
18 responses

Hi Pravin -

Just curious - if the algorithm can be adapted to equities, then why have you not done so and offered it to Quantopian?

Because Quantopian does not scale.

Scale in what respect? Are you talking about their computational platform?

Yes. Very slow and times out.

So are you getting any support (see https://www.quantopian.com/posts/speed-please-2)? Or at least an explanation?

Hi Pravin,

Can you provide more detailed metrics such as CAGR, maximum drawdown, Sharpe, etc.? Thanks.

@Grant - They are supporting me.

@James - I just downloaded minute data. Will provide other metrics soon.

@ Pravin -

Do you have any out-of-sample, real-money results (that would be audit-able)? I might pay $50 for an algo, but not$100K, unless there were some way to validate it (in which case, I'd pay \$100).

Thanks Grant for your support. If I had audited results I would start my own fund by now. This is a chicken and egg problem.

Frankly I think you’ll have a hard time as a freelancer. If it can be made to work on equities maybe you can get an audience with Quantopian. At least with them you can run it for 6 months out of sample on their platform and they can see that you haven’t overfit.

@ Pravin - Any updates on the metrics...kinda hard to analyze performance just based on your above graph, if you are seriously trying to auction it out for bid. I have a 5 crypto long only portfolio that is currently being evaluated by a VC. If your algo is something that will pass their scrunity, I might throw it in the deal. You do know the complexities in trading cryptos, using broker price data/timezone, 30% collateral on borrowed, difficulties in shorting, etc.

Here's the stats on my system:

Statistics from  1/2/2016 - 2/28/2018
Initial capital 1000000.00
Ending capital 29992510.33
Net Profit 28992510.33
Net Profit % 2899.25 %
Exposure % 39.71 %
Net Risk Adjusted Return % 7300.18 %
Annual Return % 399.26 %
Risk Adjusted Return % 1005.32 %

--------------------------------------------------------------------------------

Max. trade % drawdown -38.16 %
Max. system drawdown -9640584.80
Max. system % drawdown -27.22 %
Recovery Factor 3.01
CAR/MaxDD 14.67
RAR/MaxDD 36.94
Profit Factor 2.20
Payoff Ratio 3.02
Standard Error 4976688.50
Risk-Reward Ratio 2.22
Ulcer Index 6.51
Ulcer Performance Index 60.50
Sharpe Ratio of trades 1.37 (uses 5.04% risk free rate)
K-Ratio 0.0489

--------------------------------------------------------------------------------
Percent Profit for ETH-USD 664.50 %
Percent Profit for XMR-USD 77.86 %
Percent Profit for BTC-USD 363.43%
Percent Profit for LTC-USD 280.95%
Percent Profit for DASH-USD 351.30%



James. Apologies for the delay. I use zipline fork for backtesting (enigma-catalyst) I am trying to use pyfolio on this but without success. Please allow some time because I might have to calculate everything myself.

By the way, I am also trying this on crypto futures on bitmex (easy to short and more liquid), although universe is smaller. Downloading data and struggling to align future maturities.

@Pravin,

No problem, I understand and will wait. Thanks.

@James,

Pardon me my ignorace, but what is a VC?

Hi Tim,

VC stands for Venture Capitalist or also referred to as angel investor.

Thanks for the explanation, James. You guys have created some awesome algorithms. Good luck with their monetization! :-)

P.S. I realize I am pushing the boundaries of your kindenss and patience, James, but do you think that your VC might be interested in something like this:

Backtest period Jan 2007 -- Dec 2017
Sharpe ratio 1.08
Annual return 5.7 %
Annual volatility 5.3 %
Max drawdown -5.4 %
Daily value at risk 0.6%

Can therefore be leveraged three- of four-fold.

Average Alpha 0.05, Average Beta 0.07
Positive return every year, except 2015 (-1.0%).

Hi Tim,

While a long backtest is always a good thing in showcasing the systems' generalization capabilities, consistency and adaptability over a long term, the stats the you provided above, does not give me enough information to fairly evaluate the system. The cyrpto market is really just in its infant stage and much of the action just happened in the last two years. The last two years of data is more relevant than the last ten years (I don't even know that futures existed 10 years ago!) because they have more meaningful data in terms of volume and liquidity.

Can you give me more specific performance data/metrics on just last two years starting January 2016, futures instrument you are trading, exchange, contracts, continuous pricing or rollover methods, etc?

PS- Tim, I might be misunderstanding you when your say " Trading futures, long only", I mistook it as crypto futures. Now just realized that you might have meant Futures in general. If so, disregard my comments above. My VC is only evaluating crypto currency systems.

Hi James,

Many thanks for looking at the performance paramerters trat I provided on my algo and for your valuable comments. As you have correctly inferred yourself, the algo does not trade cryptocurrencies, but rather futures in general. In this regard I owe you an apology, as I now realize that I should have mentioned this from the very beginning. I suppose I got carried away by the possibility of having the algo evaluated by a VC.