forex-scalping.com

Take profit on a buy closes on the bid, which stretches the real target

A common arithmetic slip: a buy is entered at the ask and its take profit executes at the bid. The target is five pips above entry. But entry happened at the ask, and exit happens when the bid reaches the target. So price, measured at the bid, has to travel five pips plus the spread.

With a 1.5 pip spread, the market must move 6.5 pips from the bid at entry to hit a five pip target. For sells, the situation mirrors, with the ask needing to fall to the target.

Scalpers working with small numbers feel this strongly. A 3 pip target on a 1 pip spread requires a 4 pip move by the bid, which is a third more distance than the nominal target implies.

Backtests that use mid prices or bid-only data hide this. The effect shows up only in live results as a persistent lag behind the testing numbers.

The cure is simple: model both sides of the quote in testing, or add the average spread to every target.

Has a bid-only backtest ever made a strategy look better than live trading proved?

I only noticed this effect properly when I split my live results by direction. My sells matched my backtest closely, but my buys were consistently a little worse, and it took me a while to see that my test data was bid only. On a 4 pip target with a typical 1.2 pip spread at my broker, the buy side needed the bid to move 5.2 pips, which is nearly a third more distance.

I fixed it in two ways. In testing, I use a tool that models the spread on every bar and I add the average spread of the session to my buy targets. In live trading, I place the take profit one spread further than the number I had in mind, then judge whether the setup still offers enough room.

Some setups stopped being worth taking after that adjustment, especially the quick scalps around the end of the New York session when spreads widen. Dropping those improved my live win rate by about four points over a quarter.