forex-scalping.com

Why one-minute bars lie about what happened inside them

A one-minute candle has four numbers: open, high, low, close. Between those four numbers sits a whole path that the candle cannot describe.

Two bars can look identical and have completely different stories. One opened at 1.0850, ran to 1.0860 in the first ten seconds, fell to 1.0845, and closed at 1.0852. Another opened at the same price, drifted down to 1.0845 first, ran to 1.0860 in the last ten seconds, and closed at 1.0852. Same OHLC, opposite sequences.

For a scalper with a 4-pip target and a 4-pip stop, the sequence decides the outcome. In the first bar, the target hit before the stop. In the second, the stop hit first.

Backtests on bar data have to guess. Most engines assume a default order, such as open, then low, then high, then close, or the reverse, depending on the candle colour. The guess flatters some strategies and punishes others.

The danger is greatest for tight targets and stops, since both sit inside the typical bar range. The wider the targets compared with the bar, the less the sequence matters.

Tick data removes the guessing. It costs storage and time. For short-term strategies it is the only data that tells the truth.

A fast diagnostic is to count what share of the strategy's trades have a stop and target both smaller than the average one-minute range. If that share is high, the strategy lives inside the zone where bar data cannot be trusted. Some traders move the test to ticks only for that subset of trades, which saves time while still removing the worst source of error. The comparison between bar-based and tick-based results on the same subset is often the most persuasive evidence available.

Is the current backtest using ticks, or is it relying on an assumption about what happened inside each bar?

A one-minute chart prints a bar every minute whether the market did anything or not. A quiet minute shows a tiny bar, a busy one a big bar. The clock sets the sampling.

Tick bars sample by activity: a new bar every 100 ticks, say. Busy periods produce many bars in a short time, quiet periods produce few. Range bars close when price has moved a fixed distance, ignoring time altogether.

Each has strengths. Time bars match how the world schedules itself: sessions, news, and rollover. Tick bars adapt to volatility and often show cleaner patterns during active moments. Range bars filter noise and make support and resistance easier to see.

There are costs. Indicators calibrated on time bars behave differently on tick bars. Backtesting tools may not support them well. And tick counts depend on the data feed, so a tick chart isn't portable between brokers.

Statistically, activity-based bars tend to give more normally distributed returns, which suits some mathematical models.

For scalping, trying all three on the same strategy and comparing results often shows which one fits the logic.

Which bar type feels most natural for scalping, and what drove that choice?

A one minute bar shows the open, high, low and close of price in that minute, but not the order in which the high and low occurred. A bar with a high three pips above the open and a low three pips below could have gone up first, down first or back and forth several times. For strategies with stops and targets inside the bar's range, the order matters: it decides whether the stop or the target was hit first. Backtests using one minute bars must make an assumption, and different testing tools make different ones. That can make the same strategy look much better or worse depending on the tool and the assumption it uses.