A Volatility Filter

Traders thrive on volatility. If the market doesn't move, we can't profit. It seems logical, then, that volatility could be used as a filter in a trading system. But how? If high volatility is good, should we wait for signs of high volatility before entering a trade, or should we take a counter-trend approach and wait for low volatility in expectation of higher volatility to follow?

Consider the chart below, which consists of daily bars of the E-mini S&P 500. Below the price series I've plotted the true range. This is NOT the average true range but simply the true range of the daily bar. To me, the striking aspect of this plot is the cyclicality of the true range. It displays a surprising amount of regularity, with peaks following troughs every day or two. This tells me that if we want higher volatility, we need only wait for a low volatility day. The following day, or perhaps the one after that, will very likely have higher volatility, as defined by the true range.

Figure 1. Daily bars of the E-mini S&P 500 with the true range plotted below.

One application where we might want to use volatility is with a breakout-type system. If we want to enter the market on a breakout, we may benefit from restricting our entries to days that have higher volatility (e.g., higher true range). To see if this is true, I created the following simple system, shown below in EasyLanguage:

Input: StopPts  (30);  { Size of money management stop, points }

Var:   TR       (0);   { True Range }

 TR = TrueRange;

 Condition1 = TR < TR[1] and TR < TR[2];
 {Condition1 = true; }
 Condition2 = C > C[14] and C < C[2];
 {Condition2 = C < C[2];}
 Condition3 = C < C[14] and C > C[2];
 {Condition3 = C > C[2];}

 If Condition1 and Condition2 then
    Buy next bar at Highest(H, 2) stop;

 If Condition1 and Condition3 then
    Sell short next bar at Lowest(L, 2) stop;

 If MarketPosition = 1 and C > EntryPrice and BarsSinceEntry >= 1 then
    Sell this bar at close;

 If MarketPosition = -1 and C < EntryPrice and BarsSinceEntry >= 1 then
    Buy to cover this bar at close;

 If MarketPosition = 1 then
    Sell next bar at EntryPrice - StopPts stop;
 
 If MarketPosition = -1 then
    Buy to Cover next bar at EntryPrice + StopPts stop;

This system buys at the highest high of the last two bars and sells at the lowest low of the last two bars following two setup conditions. The first condition is our volatility filter. In order to determine that we're at a low point in the true range cycle, we require that the true range is less than it was on the previous bar and less than it was two bars ago. The two-bar-ago condition roughly matches the cycle length of true range, as seen from the plot in Fig. 1.

The second condition is a simple price pattern. The system wants to buy dips in an uptrend and sell rallies in a down trend, so for an up trend, we require the close to be greater than it was 14 bars ago. The dip is defined by C < C[2]; i.e., the close is less than it was two bars ago. We use the reverse logic for a short trade setup. If these two conditions are met, the system tries to enter on the next bar using the highest high/lowest low. The system exits at the first profitable close after at least one day and uses a fixed money management stop defined by the input StopPts.

Now for the results. Using TradeStation 6, I tested the system over the last 750 days on the continuous contract E-mini S&P (symbol @ES in TS 6) with $75 per contract deducted for trading costs. The lookback length (maximum number of bars the strategy references) was set to 15, and the input StopPts was set to 30.

Using the volatility filter, as in the code above, the results were as follows:

   Net Profit:      $10,863

   Percent Winners:    71% (21 trades total)

   Ave Trade:      $517

   Max Drawdown:  -$3,013

   Profit Factor:       3.82

If we remove the volatility filter by commenting out Condition1 and setting Condition1 = true, we get the following results:

   Net Profit:      $13,713

Although the net profit is higher, it takes twice as many trades and produces twice the drawdown. The profit factor -- a measure of robustness -- is much lower without the volatility filter. The volatility filter screens out half the trades but less than a quarter of the profits, indicating that it works as intended.

To confirm these results, I also tested the volatility filter with the simpler price pattern shown by the commented-out Condition2 and Condition3 variables above. These simpler conditions are C < C[2] for a long entry and C > C[2] for a short entry. Without the volatility filter, the results are marginal at best:

   Net Profit:      $3,525

While I wouldn't recommend trading this simple system as written, I think it demonstrates that a simple volatility filter based on the true range can improve the performance of a breakout system.