TSCI.pngReward to Pain Scaling


A tradescape graph can use one of three different reward to pain Z-Axis scaling options. Each has their own special function when evaluating and comparing tradescapes.

Reward>Pain

This is the default option for a tradescape scaling. When selected, the lower limit of the gradient used for the contour or surface will be 1.0. This means the surface or contour appears on the plot only where more reward than pain is found.

For entities that have a strong measure of long term growth, or which evidence a good measure of order in their price movements, or both, most of the tradescape surface will be plotted. For those entities with little to no long term growth, or even net loss, and those that are difficult to signal for the order in the price movements, it is possible that none of the tradescape surface will be plotted.

This is the strictest of the three criterion for reward to pain scaling. The threshold for this >P plotting depends completely on the definition of pain inherent in the reward-to-pain metric chosen. All of the metrics in the Trading Sciences platform use a robust trend or CAGR for the reward. It is expressed as the annual growth rate in %. The difference rests in the definition of pain. If that metric is RRt, the pain will be the average retracement from the all-time high. If the long term traded robust CAGR is 15%, only those backtests where the retracement in the equity curve is less than 15% is shown. If the metric is R, the pain is a worst-case measure of the depth and length of some count of the worst drawdowns. If the metric is this robust Sharpe ratio, the pain is defined as the average volatility, computed from bar to bar variations extrapolated to an annualized % volatility.

In all three instances, there will be profitable regions that are not plotted because the trading pain exceeds the reward.

Keep in mind that this option does not isolate the contribution of the active trading from the performance of the underlying. There are entities where the growth in a buy and hold scenario is so strong that the surface may be drawn everywhere from that alone, even when the trading system has an adverse impact on the reward-to-pain.

Reward>0 (Win)

This is a less stringent scaling where all profitable zones are rendered in the contour or surface. While the gradient reflects the pain associated with the trading, there is no attempt to differentiate where reward exceeds pain. If any profit at all is observed, the surface or contour is drawn. The lower limit on the reward to pain metric is thus zero.

In this instance, it does not matter if the RRT, R, or Sharpe is used. The lower limit will represent the threshold for a profitable return. The regions that are not plotted will be those where a loss occurred in the backtest.

This option likewise does nothing to isolate the contribution of the active trading from the performance of the underlying.

Reward>Underlying

This is the scaling to use to visualize the impact of the signaling in the active trading. The lower limit of the gradient is set to the value of the reward-to-pain metric for the underlying buy and hold for this same analysis period. This is a relative scaling that may result in very different lower limit thresholds when comparing tradescapes side by side.

In this instance, all zones where the backtests underperform the buy and hold are not rendered. The threshold will be the exact value of the underlying, and as such the gradient scale will not have friendly rounded values. If no surface is plotted at a given time horizon and lag, it means that the trading signals failed to offer any improvement in reward-to-pain, and most likely produced an inferior one. Where a surface is drawn, the actively managed trading, the signaler's entries and exits, added value, outperforming the underlying.

Full Gradient

8031.png  If you would like to see all of the tradescape with an adjustable scaling that accommodates all of the Z-values in the contour or gradient, you can click the Reset Default Scaling button in the graph toolbar. In this case, a scaling is made that accommodates the full range of reward-to-pain values. To restore any of the predefined scaling options, simply select on one of the three Z-Axis Lower Limit buttons.