Your browser version is outdated. We recommend that you update your browser to the latest version.

Like Our Page

Backtesting for Probabilities

Posted 2/20/2016

To determine the probabilities for winning trades on the EURUSD, the Retracement Strategy is optimized over a period of a few weeks to a few months to determine the set of parameters with the highest probabilities of winning.  In other words, to determine the current winning streak! Once the optimal

parameters are determined, the strategy is backtested to look for unexpected losing streaks within the time frame. Here are the results from 11/1/15 to 2/12/16, which show a slow period over the holidays when trading was thin, but no significant losing streaks. The strategy appears to have consistent performance during consolidation, down and up trending markets. 


Backtest ResultsBacktest Results


These are the optimal parameters for this time frame. They represent the current winning streak. So far, similar parameters have been working fairly consistently for many months, but no set of parameters works indefinitely. My strategy is to optimize the parameters a few times each month to find the current winning streak and avoid the current losing streak.

The trading parameters look like this:



These parameters show that the strategy is trading on a 15m chart for determining Fractals and Market Flow. It is using a 1H chart to determine the MACD signal crossing, and it is using a 4H chart to determine the Trend Bias. These time frames have been optimized to provide a certain number of trades over a reasonable time frame. You can set them to whatever you like, for any currency pair. 

A swing high Fractal in this instance, requires the 4 candles to the left and the 4 candles to the right of the central candle, to "close" lower or equal to the close of the central candle. The opposite must be true for a swing low. We call this a Fractal of strength 4 on the Close. Normal Fractals would be referred to as Fractals of strength 2 on the High & Low. Why the different fractals? To better approximate Supply and Demand levels on a higher time frame.

The minimum swing from the close of the previous swing low candle to the close of the previous swing high candle, must be within the limits of 27 to 55 pips for a qualified Up Swing. 

The optimal retracement level following an Up Swing is 70% or 0.7. For a Down Swing, the optimal retracement level is 50% or 0.5. When a qualified Up Swing (Down Swing) is measured, the Retracement Strategy will place 2 Limit Entry orders at this retracement level, in advance of price reaching this level. 

The orders will be blocked if the 15m Market Flow, the 1H MACD or the 4H Trend Bias are not in alignment with the trade. 

The Limit Entry orders are input with a 10 pip Stop Loss, and 2 different price targets. The first target is 10 pips, or 1:1 reward to risk. The second target is set to 60 pips, or 6:1 reward to risk. Why so high? Because the optimization process has shown that over the past 24 trades, there has been a 40% probability of reaching anywhere from 3:1 to 6:1. In other words, if it goes to 3:1, it will probably go all the way to 6:1. Preferably, I will be awake to manage the trade in case it turns around and goes against me. 

The strategy will automatically move the Stop Loss to breakeven once price has moved about 10 pips in my favor. 

Trading is interrupted automatically during news events by pre-programming the 6 timers for the week's upcoming events. The timers are set in NYT for my own convenience, but follow the setting of your own TS2 charts.

Timer SettingsTimer Settings


The statistics and probabilities are predicted by the TS2 backtesting software, and by an Excel spreadsheet I created. Which I use to distinguish between winning trades and breakeven trades. Since the Stop Loss is moved to breakeven, a breakeven trade is considered a winning trade in TS2. I prefer to track them separately.

Backtest Statistics
Backtest Statistics


The Trade Statistics show that there is nearly 78% chance of a winning or breakeven trade, and only a 22% chance of a losing trade, where the average profit per trade is over 2:1, above the average loss per trade. 

Predicted probabilities at optimal settings.Predicted probabilities at optimal settings.

These statistics calculated in Excel, show that the current winning streak is offering a 74% probability of hitting a 1:1 on the first half of the trade, and a 40% probability of hitting a 6:1, with a 36% chance of it turning around and breaking even on the second half of the trade. 

The prediction here is that there is about a 25% chance of being stopped out and a 75% chance of a winning trade, with an average of nearly 3:1 reward to risk. I'll take those odds! 

Over the next few weeks, the real results will be posted here to keep a journal of the performance. See MyFXBook certified results for complete details.