Rolling EV Ratio: A Trend Indicator or Performance Measurement Statistic
Here is a super-simple but useful metric for strategy evaluation or just for mathematically defining the reward/risk for holding a stock or index at a given point in time.
Rolling EV Ratio= cumulative W% (up periods/total periods) x W/L ratio (sum of wins/sum of losses)
The period length should be anywhere from 5-252 days, with shorter term periods showing mean-reversion and longer periods reflecting more stable estimates of reward/risk. Effectively the Rolling EV Ratio must be greater than .5 to have a positive edge, for example a 50% winning percentage with 1 to 1 odds is an even money proposition (50% x1=.5). Our goal is to have an EV that is as high as possible, but it is also worthwhile looking at changes in the distribution of the Rolling EV Ratio. In either case, the 252 day Rolling EV Ratio as applied to the Dow Jones Industrial Average tested over (gulp!) 20,000 bars produced the following results when traded above .5 and exiting below .5:
Using the same indicator on the SPY produced strong results as well. In recent years the Rolling EV Ratio has been superior to the the 200ma, especially when used in a more “normalized” framework. There many ways to apply this metric–it can be used first of all on a weekly, monthly or yearly time frame over a variety of period lengths to obtain a more complete picture. Secondly, it can be used to evaluate strategies as a means over turning systems on or off, or more importantly-to vary the allocation to a system. Finally, it can be used as a relative strength or momentum measure that has the benefit of overlooking volatile stocks in lieu of preferring those that have the best combination of consistency and reward to risk.
CSSA 
Hi David,
For the Dow test you show, what was the lookback period? (5 or 252 days)
Do you think that going short (instead of exiting below 0.5) would improve the CAGR even more?
Hi David,
Interesting post. I am most interested in using this metric as a rolling performance measure for turning systems on or off (or most likely, increasing or decreasing allocations). Specifically, I have been pondering a “meta-indicator” system that tracks a wide variety of indicators and combines their predictions into a single exposure (positive or negative) for the next day. These weighting of each indicator’s signal in the aggregate exposure would be based on a rolling measure of accuracy/profitability over various timeframes. Seems like the using EV for this makes sense.