FX Conversion in Multi-Currency Backtests

A global equity strategy holding a Japanese stock and a UK stock has to report its combined performance in one base currency — say, US dollars — but the moment you convert local returns to a base currency, you're mixing in the return of the yen and the pound against the dollar, whether you intended to or not. A backtest that quietly does this conversion without separating the two effects can attribute currency moves to the strategy's stock-picking skill, or vice versa, distorting the reported Sharpe ratio in either direction depending on which way currencies happened to move during the sample period.

The standard fix is to decompose return into a local-currency component (the stock's performance in its own currency) and an FX component (the currency's move against the base currency) and report both separately, so a reviewer can see whether performance came from picking good stocks or from an unhedged currency bet the manager may not have intended to take. A second, easy-to-miss detail is which exchange rate a backtest uses at the moment of a trade — using a single end-of-day rate for every trade during the day, rather than the rate actually prevailing at each trade's execution time, can introduce a small but systematic bias, especially in a strategy with intraday international trades.

Whether a strategy should hedge its currency exposure or leave it unhedged is a separate design decision — but a backtest must at least report FX and local return separately, so that decision can be made deliberately rather than being made implicitly by whatever the base-currency conversion happened to do to results.