The Emotional Quant Curve

Very timely post, considering at this very minute I’m testing my “favorite” system with different time periods and parameters because it’s still “bullish”, and perhaps sub-consciously I’m trying to find settings that will make it flip “bearish” to be more in-sync with my (wrong-more-often-than-right) gut feeling about the market. One thing I struggle with as a system trader is When to Begin the System Test (perhaps it could be post idea!). Generally 12/31/1999 seems to be a common date, as it includes both the dot.com crash and financial crisis, however, if we step back a little bit further into 1998 we can include the Asia-Contagion mini-bear market. Thanks for the post!

The concept of using out of sample prior to in sample seems reasonable. I floated the idea with Howard Bandy while back. Let’s say he was against it but I cannot find exactly what he said. This topic is worthy of discussion. Do you have any comments or justification for using out of sample prior to in sample? Is there some risk or bias associated with doing it this way?

I trade two models for proprietary trading firm with my own account. One of my models is an opening range breakout system, and I found that in bear markets (defined as SPY below its 200 day moving average), it has a higher average trade, but also a much higher variance. I made a change where I cut trading leverage in half during bear markets, and its overall risk adjusted returns improved. You really have to look at everything!

Cesar, this is one of the best trading “curves” I’ve ever seen as it has accurately portrayed my journey with many different systems over my life. I’ve found one of the best ways to combat this, is to have multiple systems firing at the same time with no system over 30% of the portfolio. Along with this, I have a big chunk of money in a very conservative asset allocation strategy. Combining a solid asset base with the diversity and timing of multiple systems seems to solve a lot of problems for me, along with keeping my drawdowns emotionally reasonable and my returns above average. You can then use leverage to up the risk/return levels or to lower the volatility, to taste. Thanks for a great post.

Cesar, you have been doing an amazing amount of research. I’d go to the extent that every quant who thought about doing CT systems has seen your research. In situations like these, have you ever considered the go-with-the-flow method of trend following? In stocks it’s rather relative momentum. However, there’s no stock that makes its way to the sky without breaking out. The advantage is that you have a stop, but there s no upside potential. the way it can move is unlimited. Every day you look at your screen, you keep the shining green ones. The red ones can’t go any further. The backtests are less stellar. 15% over 2 decades. CAR/MDD 0.5. However, the game doesn’t change, it’s the same. Whatever makes a move has to breakout. You’re a 100% equities. How about bonds, currencies, commodities in addition. Buy what’s moving up, sell what’s heading down. It’s risk allocation. No parameter setting, because it just doesn’t matter. If stuff trends, any parameter will do. It’s Med Faber stuff, how the endowments do things. What do you think? You were the pioneer on MR in the 1990-2014 phase were big money was made. Will it ever be that way? How could I convince you to look at some trend following stuff in combination with asset class rotation?