“Mean-Dev Trio” Systems

I am now offering a “signals service” for some of my systems that I personally trade. They¬†can be purchased as a subscription here. I am willing to sell the actual methodology of these systems as well, if you have backtesting software to handle the calculations. Contact me here for more information about a flat fee for the system details, or if you have questions.

Requirements for actually trading this system are at the end of this page.

At the beginning of 2016, I developed the core subsystem of a trading system I like to call “Mean-Dev Trio”. I have been trading the core subsystem (“Mean-Dev SPY”) for over a year now, with good results.

At the beginning of 2017, I developed the other two systems, because I wanted additional systems that had a low correlation to each other, while trading the same underlying ETF. The core subsystem is a mean-reversion system, while “Mean-Dev Momo Long” (aka “MDML”) is a momentum trade. “Mean-Dev Short” is, as you might have guessed, a short system. Low correlation means less drawdown, which in turn allows me to leverage the results.

Hypothetical profit from a starting account of $100,000, compounded, no commissions. *Live trading portion does not reflect commissions.

I used proper in-sample/out-of-sample testing to develop these systems. Generally, one ignores the in-sample period and focuses on the results of the “out-of-sample” data (which is not tested until the system is finalized). This reduces the likelihood of curve-fitting, which can be disastrous to trading models. The OOS period¬†performs very well.

Even better, the core subsystem has 15 months of truly out-of-sample live trading experience behind it. The proof is in the actual money earned. The other two subsystems have been traded live since 1/1/17.

All of these systems are short-term “swing trade” systems. They have a low overlap, and are only in the market about 40% of the time. This reduces exposure to catastrophic market events.

Gray line is buy-and-hold for SPY. Orange is combined Mean-Dev Trio systems. Other colors are individual subsystem performance.

Win rates are very high, in the range of 60-70%. Wins tend to be many and smaller, and losses tend to be fewer and larger. Even so, the maximum drawdown of these combined systems is such that you can comfortably use 3x leveraged ETFs. You maintain a better risk profile than buy-and-hold, with a greatly multiplied profit.

Drawdowns are dwarfed by buy-and-hold (gray). Orange is the combined drawdown for the Mean-Dev Trio system.
Here you can see the drawdown profile of the combined Mean-Dev Trio system by itself. It’s relatively consistent over the years, with biggest drawdowns in 2001, 2008/09 and 2011. Not surprising!

Because the drawdowns of this system are relatively small – even during bear markets – it is entirely practical to use leveraged ETFs. I use UPRO and SPXU, a pair of 3x long/short ETFs that have been around since 2009. Here’s how they would perform.

Even during this period of extended bull market activity, the leveraged version makes about double what buy-and-hold made. The drawdown profile very closely matches buy-and-hold, except for a slightly higher maximum drawdown during 2011. One could always reduce ones position size a tiny bit to match the MaxDD profile exactly.

Want to nerd out on the stats for this set of systems? Here’s a PDF.

Trading is of course very risky. Most traders lose money. You are responsible for any trades you make, including any losses you incur. I am providing signals for trades I am making myself, so I have actual money at risk. None the less, the future is unknowable, and you might lose money. Trade at your own risk, and don’t come crying to me if things go poorly.

Requirements:

  • You need to have sufficient capital that 1% profits per trade aren’t eaten away by commissions. If you’re trading a smaller account, consider a zero-commission brokerage.
  • Signals are only sent when they appear, in the evening after the closing bell. There are an average of 3.5 trades a month, although there will be a few more signals that aren’t acted upon.
  • One system has you buy at the next day’s open, and holds for a set period. The other two will have a signal sent, but will be bought at the close of the next day ONLY if the day is up or down (depending on the subsystem’s requirements).
  • The only stops are “time stops”. Mean reversion systems tend to suffer by adding stop-loss requirements. Your protection against too much drawdown is simply to get out after the maximum holding period have passed.
  • One system has no exit criteria other than the hold period. The other two exit at the first profitable close.