This is a great thread with some quality posts.
I have a few comments as both a system developer (Drunk Uncle) and an investor who has been trying to diversify across multiple systems in addition to my own. Beating the market with a diversity of systems is a difficult problem. As evidence of that here’s a link to the performance of Collective2’s own combined system product, ScoutAlpha, which apparently has a negative return since inception in Aug 2016:
https://trade.collective2.com/scout-alpha/performance/
My own experience with a group of systems from last year was good but I did not do as well as I’d hoped. I made a few mistakes I’m trying to learn from and I’m always looking for good advice so I appreciate this thread.
Here are a few thoughts I think are key in building a portfolio of systems:
0) Understand Survivorship Bias. Fundamentally C2 highlights leading systems culled from a large population of trading systems. There will always be winning systems to put on the leaderboard, but will those systems continue to perform going forward? Most systems will not. This is the core problem we face.
en.wikipedia.org/wiki/Survivorship_bias
1) Related to #0, avoid hot systems without much time behind them. I violated this concept with two systems late last year by subscribing to them when they had only a few months of very good/steady returns and both tanked the moment I subscribed. I had reasons for both subscriptions… one system had a developer with other successful systems and the other had a large number of daytrades I thought made the data statistically significant. The market changed gears and both tanked; one system was killed by its dev and the other dev announced a change (see #4 for my view on that). I’ve had better luck with systems that have a longer history of success even if returns aren’t as sexy. Number of trades is not enough, you need time in the market. Waiting for 3 quarters of data is a sweet spot IMO.
2) Avoid large drawdown systems even if the return appears worth it. Fill your stable with systems that don’t fail big when they fail and you’ll do less damage to the returns of your profitable systems when things go wrong. It’s also easier to determine when a system might be having a problem before you lose a lot of capital if the system has reasonable drawdown.
Also see posts above by LeslieGray and her ideas on drawdown and correlation.
You can get Correlation data on C2 here (using C2 explorer):
forums.collective2.com/t/correlations-table/8455
3) Related to #2, avoid Martingale systems, systems with extreme leverage, and other systems with large hidden risks. Some systems on C2 have great looking equity curves or high returns but have significant hidden risks.
Here’s a post I wrote on how to detect martingale systems on C2:
forums.collective2.com/t/dangerous-strategy-with-a-martingale-example/8633/5
4) Be worried when a developer changes their strategy. Creating a system that is profitable into the future is difficult even for developers that know what they are doing. Usually changes to a system need to be forward tested to prove out the change and you don’t want real money at risk on someone’s test. An exception to this might be systems that are regularly updated/changed and have a history showing the success of that process. Otherwise the announcement of a significant change to a system (usually due to underperformance) is often the warning of worse to come IMO.
5) I am personally wary of discretionary systems and prefer systematic traders. Enough of a trading history might convince me take a chance on a discretionary system but it’s not my preference. Last year my worst performer was a discretionary futures system that stopped performing.
As a final thought it’s good to Understand how a system works as much as possible, where it’s risks are, what its skew is, etc. Unfortunately these things take experience and can be difficult to get on some systems. One theme I see right now is VIX systems are all over the place as we’ve been in a falling volatility environment. I would be careful not overweight my portfolio with VIX systems. VIX systems can fail very big very fast and I think there will be a tough time for those systems at some point ahead where many will not survive. JMO.
Here’s a post I wrote on skew:
forums.collective2.com/t/histograms-examples/8350/3
