Advertisement IC Markets

Tick Data Versus Bar Data: Why the Same Test Gives Different Results

Share, develop, and backtest custom MQL4/MQL5 Expert Advisors, Python data-scraping scripts, trading bots, and automated market alert systems.
Post Reply
Fairman
Posts: 6955
Joined: Tue Jul 21, 2026 7:11 am
Location: Abuja

Tick Data Versus Bar Data: Why the Same Test Gives Different Results

Post by Fairman »

Tick Data Versus Bar Data: Why the Same Test Gives Different Results

When you test a strategy, the quality of price data changes the answer. Two main types exist: bar data and tick data.

Bar data, also called OHLC data, records open, high, low, and close for each period. It is compact and widely available. But it hides what happened inside each bar. If a candle's range includes both your stop and your target, a test cannot know which was hit first. Software often assumes a rule, and that rule can be optimistic.

Tick data records every price change. It allows tests to replay the actual sequence of movement, resolving the order of events inside bars. It also shows spread variation and, with good quality, realistic stop behavior.

The difference matters most for short-term strategies with tight stops and targets. For daily strategies with 100-pip stops, bar data gives reasonable results.

Tick data has its own problems. Files are large, and quality varies. Brokers' tick feeds differ, and some historical data is modeled, meaning interpolated from bars and not recorded, which can be misleading. Platforms usually label the modeling quality of a test.

If you test a scalping strategy on one-minute bars, treat the results with suspicion.

Practical step: check the data quality percentage reported in your testing platform and what kind of data it used.
It’s Fairman :geek:
Recommended broker for automated trading & scalping IC Markets
Fairman
Posts: 6955
Joined: Tue Jul 21, 2026 7:11 am
Location: Abuja

Re: Tick Data Versus Bar Data: Why the Same Test Gives Different Results

Post by Fairman »

Very common story with scalping EAs: great curve on M1 bar modeling, then flat or worse once you rerun the same rules over the same period with real ticks and variable spread. Nothing changed except the data.
It’s Fairman :geek:
Fairman
Posts: 6955
Joined: Tue Jul 21, 2026 7:11 am
Location: Abuja

Re: Tick Data Versus Bar Data: Why the Same Test Gives Different Results

Post by Fairman »

If your test shows a low modeling quality number, don't argue with it. Get better data or treat the result as a rough sketch only.
It’s Fairman :geek:
LondonNewsTrader
Posts: 421
Joined: Mon Sep 21, 2026 9:30 am

Re: Tick Data Versus Bar Data: Why the Same Test Gives Different Results

Post by LondonNewsTrader »

The news minutes are where the two data types disagree most. Real tick data from a release often shows gaps, a frozen quote for a second or two, then a jump several pips away with a much wider spread. Modeled ticks interpolated from M1 bars draw a smooth path through that minute instead, so a stop inside the jump gets filled at its level in the test when live it would have filled at the far side. If a strategy holds positions through releases, that's one more reason to only trust real tick tests.
FXS Prop Desk
Posts: 1689
Joined: Wed Oct 07, 2026 9:52 am

Re: Tick Data Versus Bar Data: Why the Same Test Gives Different Results

Post by FXS Prop Desk »

For scalping strategies, the difference can be large. Bar data does not show the order of prices within a bar, so a test may assume a target was hit before the stop when the reverse happened.

Tick data with variable spreads is the most realistic option. Even then, the test does not include slippage, so live results are usually somewhat worse.
FXS Official Team | The Idea Lab & Strategy
Shadow Trader
Posts: 5594
Joined: Wed Sep 09, 2026 5:19 pm

Re: Tick Data Versus Bar Data: Why the Same Test Gives Different Results

Post by Shadow Trader »

Trades where the stop and the target both sit inside one bar can be counted before trusting a test. That gives a quick bound on how much the bar data flatters the result.

Say a scalping system uses a 5 pip stop and a 5 pip target, and in 400 trades on M1 bars, 48 have both levels inside the same candle. If the tester resolves those in the trade's favour, it books 48 wins worth 240 pips.

If the real split is closer to even, those trades are 24 wins and 24 losses, which nets zero. The difference is 240 pips over 400 trades, or 0.6 pips a trade, and that is more than many scalping edges are worth after spread.

The count is easy to produce from the trade list and the M1 bars. Anything above a few percent of trades means the test needs real ticks before the equity curve says anything.

The slippage point above is the other half. Real ticks fix the order of events but still fill every stop at its exact price, so slippage on fast moves is missing from even a perfect test.

Has anyone compared the ambiguous bar count across pairs? It seems likely to be much higher on GBPJPY than on EURGBP for the same pip distances.
Shadow Trader
Posts: 5594
Joined: Wed Sep 09, 2026 5:19 pm

Re: Tick Data Versus Bar Data: Why the Same Test Gives Different Results

Post by Shadow Trader »

Feeds disagree at the fraction of a pip level, and scalping tests are often sensitive right there. Tick data from one source can print a high a few tenths above or below the broker's own price.

A quick stress test is to count winners whose target was reached by less than half a pip. Those are the trades most likely to change result on a different feed or on the live account.

Say a system with 3 pip targets and 3 pip stops runs 800 trades, and 40 winners touched the target by under half a pip. If half of those would miss live and go on to the stop, 20 trades swing by 6 pips each, which is 120 pips or 0.15 pips a trade.

That looks small, but on a 3 pip target it is a big slice of whatever edge the test shows. Has anyone run the same rules on two tick sources and compared the trade lists?

Time stamps are the other quiet difference. Many tick files are in UTC while the broker server sits two or three hours ahead, so a session filter written in server time can test the wrong hour unless the data is shifted first.
Shadow Trader
Posts: 5594
Joined: Wed Sep 09, 2026 5:19 pm

Re: Tick Data Versus Bar Data: Why the Same Test Gives Different Results

Post by Shadow Trader »

When a bar's high touches the target and its low touches the stop, the backtester must decide which came first. This same-bar conflict occurs often in scalping tests, where distances are small relative to bar range.

Counting how many trades in a backtest involve such conflicts reveals the degree of uncertainty. If a large share of trades is affected, the headline result carries wide error bars. Some platforms report this number, while others require custom analysis.

Resolving the conflict properly needs lower timeframe or tick data. A common workaround uses a smaller timeframe, such as seconds, to determine sequence inside a bar. Another is to assume the worst case for every conflict, which gives a conservative lower bound on performance. Comparing worst-case results with best-case results shows how much of the strategy's apparent profit depends on ambiguous trades. A robust system remains attractive even when every conflict resolves against it, which is a demanding but useful test.
Post Reply