Given an automated system which can run backtests for a given market over multiple assets (eg. forex pairs), what is the best or most standardised way to output your results? I'm most interested in how people are compositing and evaluating their results. If I run a backtest over a single pair eg. AUDUSD, I would calculate the following data points (example data only): - dates: 01/01/19-30/06/19 (6 months) - total trades: 100 - win/loss trades: 40/60 - avg win: +1% - avg loss: -0.5% - starting balance: 10000 - final balance: 11000 - profit amount: 1000 - profit %: +10% - total R (accum risk/return): 10R - sharpe ratio: 0.6 If you run another backtest, but instead over multiple assets? eg. AUDUSD + EURNZD + USDCAD. How do you composite the results to get an overall picture of performance? 1. Would the starting balance be divided up among each asset? eg. 10000/3 = 3333 for AUDUSD, 3333 for EURNZD, 3333 for USDCAD. 2. How would you accumulate the profit % value? eg. 10% for AUDUSD + 8% EURNZD + 4% USDCAD = +22% total. Or, if the starting balance is subdivided as above, do you accumulate and divide? eg. +10% +8% + 4% / 3 = +7.33% total. Apologies for the long thread. Any advice or links would be greatly appreciated. Thanks, Jim
Two quick thoughts: 1) "Composite" and "output" are nouns. If you can re-write your sentence with verbs, you'll attract more knowledgeable responses. 2) Yes: studies across different sets of assets/underlyings must be unitless for meaningful (proportionate) comparison. You're on the right track.