About two years ago, Vic and Laurel discussed control charts. I use a similar concept by tracking the drawdowns of various trading strategies. If a strategy is working well, there will be drawdowns, but the drawdown amount will repeatedly return to zero as profits exceed losses. Conversely, if a strategy is losing, the drawdowns will get further and further away from zero. Thus, a drawdown that exceeds a preset threshold or fails to return to zero for an extended number of trades might indicate that the cycle has changed.

Andrew Moe adds:

Improvements to control charts can be made by upgrading the fixed window look-backs to exponential or DSP style windows (see Ehlers, Jurik, et al). Information theorists will find additional gain via the use of fast-responding windows on the entropy. These methods are used to turn various strategies, traders, and funds on and off. I particularly like Mr. Ellison's idea of return to even from the max vis-à-vis survival statistics as traders tend to get paid on making all-time highs, and the distance and journey between tell much about the method.

Vincent Andres writes:

Cycle changes are not directly viewable. Hence we're often trapped.

I believe cycle changes are only viewable in some slightly more indirect universe of parameters. Don't look at prices or first differences of prices.

Cycle changes are of course written in them, but a bit too deeply to be seen just on the surface (price). We must go one (or more) levels downstairs. We have to search/compute indirect/deep parameters from the prices.

Plotting those deeper parameters will of course show deeper things. Just as using a microscope or X-Rays. Some parameters we may consider:

1. Correlations (in many ways)
2. Regression coefficients
3. Lags
4. Price distribution parameters

Scatter-plots may also be of interest. Points clusters may correspond to regimes.

Vic and Laurel recently had an interesting post about ordering/ranking. In a general sense, rank measures can be used in many ways — what's the cluster of the leading horses, and is it changing?

A great part of physics is simply building measurement tools. Then using them. Is market physics so different from other physics?

Adam Robinson explains:

Vincent's insights answer his own query, for the difference is fundamental and ineradicable.

In the physical world, phenomena consist of relations among unvarying "observables", and even at the quantum level, observing those phenomena offer us at worst a position vs. momentum tradeoff. Light quanta might shift a subatomic particle we have in our sights, but they don't usually change the particle.

With markets, once one or more traders observes a phenomenon (e.g., equity markets are rising as the yen falls, whatever — even spurious correlations) they begin to trade/arbitrage away the phenomenon so that it diminishes or even disappears entirely.

Hence ever-changing cycles that vex and humble us.





Speak your mind


Resources & Links