Five things a bar chart hides from a scalper that tick data shows
Posted: Sat Oct 31, 2026 8:15 am
A one-minute bar says four things: open, high, low, close. A scalping strategy lives and dies on what happened inside it, and the bar throws most of that away. A list of the details that matter most:
- The order of events. If a bar touched both the target and the stop, backtest software on bar data has to guess which came first. Different tools guess differently, and each guess can flip a result from profitable to ruined.
- The real spread. Most historical bars are bid-only, with a fixed or average spread added. Actual spread balloons at exactly the times a scalper is tempted to trade.
- The gaps in time. Thirty ticks in one second followed by forty seconds of silence both end up as one candle.
- The speed of the move. A ten pip bar formed over fifty seconds of grinding is a different animal from the same bar formed in two seconds, and any signal built on the bar treats them identically.
- The quote quality. Flicker, stale prices and brief spikes get folded into highs and lows without any trace of where they came from.
Tick data solves some of this, but not everything. Tick data from one broker isn't the tick data from another, and a lot of "tick" files are actually interpolated from minute bars. The file looks granular but contains nothing new. A quick check is looking at the number of ticks per minute in a quiet hour. If every minute has exactly the same count, the ticks were generated, not recorded.
There's also the question of which side is recorded. Bid-only tick data cannot say when a long would have been stopped, since long stops trigger on the bid and short stops trigger on the ask, and the ask side is missing.
Testing the same strategy on bar data and on genuine tick data usually produces two very different equity curves. The gap between them is the part of the backtest that can't be trusted.
Has anyone checked how much a strategy's results shift when switching from bar data to genuine tick data?
- The order of events. If a bar touched both the target and the stop, backtest software on bar data has to guess which came first. Different tools guess differently, and each guess can flip a result from profitable to ruined.
- The real spread. Most historical bars are bid-only, with a fixed or average spread added. Actual spread balloons at exactly the times a scalper is tempted to trade.
- The gaps in time. Thirty ticks in one second followed by forty seconds of silence both end up as one candle.
- The speed of the move. A ten pip bar formed over fifty seconds of grinding is a different animal from the same bar formed in two seconds, and any signal built on the bar treats them identically.
- The quote quality. Flicker, stale prices and brief spikes get folded into highs and lows without any trace of where they came from.
Tick data solves some of this, but not everything. Tick data from one broker isn't the tick data from another, and a lot of "tick" files are actually interpolated from minute bars. The file looks granular but contains nothing new. A quick check is looking at the number of ticks per minute in a quiet hour. If every minute has exactly the same count, the ticks were generated, not recorded.
There's also the question of which side is recorded. Bid-only tick data cannot say when a long would have been stopped, since long stops trigger on the bid and short stops trigger on the ask, and the ask side is missing.
Testing the same strategy on bar data and on genuine tick data usually produces two very different equity curves. The gap between them is the part of the backtest that can't be trusted.
Has anyone checked how much a strategy's results shift when switching from bar data to genuine tick data?