OHLC, one-minute or tick: which data should you backtest on?

Every backtest makes two decisions before it runs a single trade: how finely it sees each bar, and how far back it looks. Get either wrong and the result is not a pessimistic or optimistic estimate of your strategy — it is an estimate of something else entirely.

Resolution: what each kind of bar can and cannot tell you

The classic trap is the intrabar ambiguity. A daily bar tells you the high and the low but not which came first. If your stop and your target both sit inside one bar, the backtest has to guess — and most platforms guess optimistically. One-minute data shrinks that window; tick data removes it.

How much history: the arithmetic of proving an edge

The honest question is not “how many years” but “how many trades”. A strategy’s average result per trade has a statistical noise that shrinks with the square root of the number of trades. That gives a simple rule: if your average trade divided by the spread of trades (a per-trade Sharpe ratio) is s, you need roughly (2 / s)² trades before the edge clears a two-standard-error hurdle.

Two consequences follow. First, a strategy trading twice a week needs years to produce 400 trades; a year of data proves almost nothing. Second, weak edges — which is what most genuine edges are — need thousands of trades, and therefore deep history at fine resolution.

Depth also buys regimes, not just trades

Markets change character. A history that contains only a calm uptrend will pass strategies that fail the first time volatility doubles. Your data should include at least one sharp sell-off, one high-volatility recovery and one long range-bound stretch. Ten years of daily data covers that easily; for intraday work it is the reason deep one-minute and tick archives matter.

Three data traps that inflate every backtest

• Unadjusted corporate actions. A 1:1 bonus halves the price overnight and looks like a crash. Your intraday and daily series must be adjusted consistently — see splits, bonuses and dividends.
• Survivorship bias. Testing only today’s index members excludes the stocks that were dropped or delisted, which flatters results. Use point-in-time membership.
• Look-ahead. Using a bar’s close to decide a trade inside that same bar. It is the single most common bug in home-built backtesters.

Once a strategy survives the numbers, size it properly before it touches capital: the free position size and risk-reward calculators turn a backtest into a plan you can actually trade.

Choosing resolution and depth by strategy type

The pattern is consistent: the shorter the holding period, the finer the resolution you need — and, because short-term edges are usually thin, the more trades you need to prove them. That is why scalping research is the most data-hungry work in trading, not the least.

How to check a vendor’s history before you buy

1. Pick a stock that split or issued a bonus. The daily and the five-minute chart should both be smooth across the date.
2. Count the one-minute bars in a full session. Missing minutes in a liquid stock mean gaps in the archive.
3. Compare the daily close with the exchange’s official close for a few random days.
4. Check an expired futures contract. Good archives keep expired contracts; poor ones only keep continuous series.
5. Ask where the history came from. Recorded live, or reconstructed later? Reconstructed data rarely has true tick sequence.

Questions traders ask