Interpolated ticks are not ticks
Some platforms generate ticks from bar data. The process creates artificial ticks between the open, high, low, and close, with spacing and sequence based on a formula.
These synthetic ticks look real. They fill the strategy tester, and the results look plausible. They have no relationship to actual market microstructure.
A strategy that depends on tick behaviour, such as one that exploits quick reversals after spikes, cannot be tested with synthetic ticks. The reversal pattern in the data came from the generator, not the market.
Another issue is spread. Historical bar data often has a fixed or average spread, which does not reflect the real variable spread. Backtests using it underestimate costs in the exact hours when costs are highest.
Honest testing needs real tick data with real spreads, ideally from the same broker used for trading. Sources include the broker's own history, third-party providers, and personal recording.
The quality of the data decides the quality of the conclusion. A fast scalper using bar data is working with a drawing of a landscape, not the landscape.
What type of data is the backtest running on, and does it include real variable spreads?
Interpolated ticks are not ticks
Interpolated ticks are not ticks
It’s Fairman 
-
FXS London Scalper
- Posts: 976
- Joined: Sun Oct 18, 2026 12:25 pm
Re: Interpolated ticks are not ticks
Interpolated ticks inside a bar can never reproduce the real order of highs and lows, so tight-stop scalps look smoother in the tester than live. Real tick data from the same broker is the only test worth trusting for that kind of system.
FXS Official Team | London Session Desk
-
Amanhasnoname
- Posts: 2060
- Joined: Fri Oct 30, 2026 9:45 am
Re: Interpolated ticks are not ticks
Mine was running on interpolated data for almost a year before I noticed, and I only caught it by accident. I had a spike-fade system on GBPUSD that looked beautiful in the tester, with a smooth equity line and very few losers. When I finally exported a slice of the test ticks and laid them over a chart of real recorded ticks from the same hour, the shapes had nothing in common. The generated ones walked neatly from open to low to high to close, while the real ones jittered back and forth dozens of times inside the same minute.
The spread issue hit me just as hard. My history file used a flat spread of about one pip, but my own recordings showed it widening to three or four in the first minutes after the rollover and around news. Most of my "best" trades in the test were taken exactly in those windows.
These days I record ticks from my live account with a small script that writes bid and ask to a file. It took a few months to build up enough data, but the tests I run on it actually match what I see live.
The spread issue hit me just as hard. My history file used a flat spread of about one pip, but my own recordings showed it widening to three or four in the first minutes after the rollover and around news. Most of my "best" trades in the test were taken exactly in those windows.
These days I record ticks from my live account with a small script that writes bid and ask to a file. It took a few months to build up enough data, but the tests I run on it actually match what I see live.
Re: Interpolated ticks are not ticks
The order of highs and lows is where the damage is biggest for anything with a stop under about 5 pips. Picture an M1 bar on EURUSD with a 4-pip range: open 1.10800, high 1.10830, low 1.10790, close 1.10820. A long at 1.10800 with a 2-pip target and a 1-pip stop is a winner if the generator walks to the high first and a loser if it visits the low first. The tester has to guess, and most generators pick the same path every time, so the outcome is set by a formula rather than the market. Repeat that over 400 trades and a coin-flip ambiguity turns into a smooth curve.FXS London Scalper wrote: Fri Nov 13, 2026 7:25 am Interpolated ticks inside a bar can never reproduce the real order of highs and lows, so tight-stop scalps look smoother in the tester than live. Real tick data from the same broker is the only test worth trusting for that kind of system.
One cheap check before paying for data: rerun the test with every ambiguous bar forced to hit the stop first. If the profit factor drops from, say, 1.4 to under 1.0, the edge lived inside the guess. Same-broker ticks are still the proper fix, but that worst-case pass tells you quickly whether the effort is worth it.
It’s Fairman 