Stationarity and Trading Regimes

[Do Or Die 10]

Previous article in this series are: Trend Following Vs Mean Reversion: Trading Regimes, Introduction to Understanding Volatility, Trading Regime Analysis Using RSI, Trading Regime Analysis Using RWI, Trading Regime Analysis Using Chart Patterns- Part 1, and Trading Regime Analysis Using Chart Patterns- Part 2.

 

People spend most of the time trying to ‘find’ a better indicator but it is ironic that very little time is spent to analyze phases in which a particular indicator works well and how these phases change.

 

Stationarity is the statistician’s term for regimes in price behavior. It is defined as a lookback period for which the averages and other statistics (e.g. flux densities, variances) do not change over time. For example, range-bound markets where tops and bottom tend to appear after predictable drift. Non-stationarity implies that the characteristics of sample data change from time to time. Now the fact is that markets move in cycles of low-stationarity to high-stationarity.

 

To expect an indicator (for example, moving average or ‘double top’) to perform consistently over time is a capital mistake because it assumes stationarity in market prices. If you study a pattern’s forecasting ability in 2010 and expect that ability to remain same in 2011 irrespective of market characteristics in 2011, it simply does not make sense. There is a fundamental uncertainty in price behavior. You can never tell if the trading regime will shift with your next trade (except ofcourse, if you are a vendor who needs to prove his accuracy). This is also why trade/risk management becomes very important- ability to handle losing trades. From the technical analysis side, it also implies the ability to quickly ‘adjust’ to the market.

 

So if I’m using a strategy- say the ubiquitous moving average crossover, the consistency of its profitability depends upon the stationarity cycle of the market. The ‘amount’ of stationarity over a lookback period depends over time frames and instruments. For example, stationarity on intraday time frames in stocks is far less than stationarity on intraday time frame in DJI. In general, stationarity on EOD time frames tends to more than the stationarity on intraday timeframes. It makes sense to chose a time frame on which you are quick to identify the phases of high stationarity.

 

Measuring Stationarity and Identifying Stable Stationary Cycles

 

For day-to-day trading, the focus is to find longest lookback period for which the market exhibits high amount of stationarity. Now the first important question is, how you measure stationarity. Dr. Steenbarger suggests the ‘dirty’ t-test which can be achieved in MS Excel via TTEST function. T-test is a poor approximation for stationarity, but a good start. You can divide the time series into two halves and see how closer the t-test is to zero (the distributions are similar during that lookback period). The process will be repetitive to find the optimal window for a phase of high stationarity.

 

Once you have the window, confirm the behavior of a trading indicator/strategy during that time. Check if they respond well to price swing highs and lows and the optimal stops/drawdown for such strategy. Finally, start trading with this strategy until the profitability starts coming down (which means the trading regime is shifting).

 

As previously questioned, some may think this is curve-fitting. To quote Dr. Steenbarger:

My response is that optimization is only a problem when you fail to take stationarity into account. If you know you are trading within a stable regime, it makes sense to do your best to capture the rules the market is following over that period. My swing trading methodology might best be described as serial optimization: continually hunting for periods of stationary market behavior and trading optimized models derived from those periods.

…Perhaps this is why we see so few traders incorporating stationarity into their analyses: It is time-consuming to assess market windows, operative trading rules, and test strategies for exploiting those rules. It is easier—and far more beguiling—to assume that a single system or indicator will produce consistent profits. More than one person has encouraged me to make my writing, research, and trading strategies less complex so that they can be more readily understood and accepted by the bulk of traders who attend seminars, buy trading books, and hire gurus for advice.

 

Using historical backtests you can determine regimes in which you strategy works best. For example, test over 10,000 bars of history to test how long and how frequent were the windows in which the strategy trades profitability. This increases your confidence in strategy and your task remains to identify (future) regime phases in which your strategy works good.

The next important question is to identify short-term non-stationarities which indicate a regime change. This depends on how you measure stationarity, as well the particular regimes you want to identify. Compare this chart to the third chart in Trading Regime Analysis Using RSI

 

 

[sf99167 10]

Would this be similar to stocks which are in their "Dormant" periods??

[Do Or Die 10]

There isn't any stock 'universally' dormant.... and every stock may be dormant for a group of traders who perceive an uninteresting risk/reward in a certain phase.

 

Trading regimes can be put in two major categories: Trend Following Vs Mean Reversion

[zdo 20]

Do or Die,

 

If you have time, please explain how you see

>going away from a 'mean'

>coming back to a 'mean'

as regimes?

 

thx

 

zdo

[Do Or Die 10]

Hi Zdo, I'm not sure if I understand you correctly.

 

The way I define a 'trading regime' is- a phase when particular strategy consistently outperforms/underperforms. Trading regimes can be identified in certain time intervals and certain instruments.

 

In this market approach- indicators, patterns, and methods of analysis are just 'windows' to see the market. For example take the 20 period MA- if it's sloping upwards you can say market is trending upwards; if prices are too far from the MA you can say they are due for correction and so on. A trading strategy can formed with this window with buy-sell rules.

 

So a trading regime is a phase when a particular window captures most of price behavior, and its associated strategy is significantly profitable. The strategy itself could be trend following, mean reversion, momentum buying, breakouts and so on. For example, some stocks will show more profitability in breakout trading than others. This can be backtested per 100 breakouts on 15 min time frame (say).

 

Gong away from a mean or coming back to mean should solely depend upon how the mean is defined. A trading regime can be defined and backtested based on this analysis method.

[Do Or Die 10]

I will recommend the book "Trading Regime Analysis: The Probability of Volatility" for discretionary traders. The good part is that it clarifies the concepts and provides many ideas. The other side is that the book is verbose.