How to read TradingView Deep Backtesting results (2026)
Deep Backtesting is TradingView's mode for running a strategy over far more history than the chart itself loads — and the results panel it produces looks identical to the standard Strategy Tester while being computed differently enough to mislead. This guide covers what actually changes in deep mode, which numbers to read first, and the three places deep-mode results most often lie: fill assumptions, date-range edges, and sample composition.
What does Deep Backtesting change versus the standard tester?
Two things, and both matter for reading the output. First, range: deep mode runs the strategy over the full selected date range on TradingView's servers, not just the bars your chart has loaded — which can take a test from a few thousand bars to a decade of data. Second, feedback: deep mode returns the summary report and equity curve, but you don't get the on-chart trade markers for the full range, so you lose the ability to eyeball individual entries against the bars that produced them.
That second change is the important one. The standard tester lets you audit trades visually; deep mode asks you to trust the aggregates. Reading deep results well means compensating for that lost audit with the checks below.
Which numbers should I read first?
Read in this order, and stop at the first failure.
1. Total trades. Below roughly 100 trades, most of the other statistics are noise wearing a suit. A decade-long deep test with 60 trades is a small sample with a long calendar, not a big test. 2. Profit factor after costs. Confirm commission and slippage are set in the strategy properties before believing anything — deep mode inherits whatever the properties say, and the default is often zero. A profit factor that drops from 1.6 to 1.1 when you add realistic costs wasn't an edge, it was a rounding error. 3. Max drawdown against your intended account, not against the test's starting capital. A $9,000 drawdown is survivable on the test's $100,000 notional and fatal on a 50K funded account with a $2,500 trailing floor. 4. Percent profitable and average win/loss together, never separately. 70% winners with 0.4 average win/loss and 35% winners with 2.8 are both viable shapes — but they fail completely different prop firm rule sets, which is a separate question from expectancy. 5. The equity curve's composition. If one cluster of trades contributes most of the profit, the test is a story about one regime, whatever the date range says.
The bar magnifier catch
Deep Backtesting supports the bar magnifier (fills modeled on lower-timeframe data), but only if it's enabled — and whether your limit and stop orders fill realistically depends on it. Without magnification, an intrabar sequence where price touches both your stop and your target inside one bar gets resolved by assumption, and the assumption flatters some strategies and buries others.
The practical check: run the same date range with the magnifier on and off. If results move materially, your strategy's economics live inside the bars, and the less favorable reading is the honest one. Strategies with tight stops relative to the bar size of the test timeframe are the most exposed.
The date-range edge problem
Deep mode invites decade-long ranges, and long ranges quietly average across regimes. Three edge checks before trusting a long test. Split the range into halves and run each: a strategy that made everything in one half is a bet on that half's conditions returning. Check the instrument's history for contract or specification changes across the range — a decade of an index future spans multiple volatility regimes and at least one repricing of what "one point" costs you. And confirm the range's start isn't inside a period the strategy's indicators need for warm-up; early trades computed on partial lookback are artifacts, not signals.
What Deep Backtesting still can't tell you
The report is one historical sequence with all its luck embedded. Two properties of your strategy stay invisible no matter how long the range: how the same trades would have fared in a different order — which is what decides survival against a trailing drawdown — and how wide the range of plausible outcomes is around the one path you got. Both are distribution questions, and the way to read them is to push the trade list through the Monte Carlo simulator and read percentiles instead of a single line — we've written a full guide to reading Monte Carlo results. Before that, the trade list itself should survive the backtest quality checker, and if you're unsure whether your range is long enough to mean anything, the sample-size math is in how long you should backtest.
FAQ
Why are my Deep Backtesting results different from the Strategy Tester on the same strategy?
Different data range is the usual answer — the standard tester only sees the bars your chart loaded, deep mode sees the full selected range. If the range is identical and results still differ, check bar magnifier state and whether the chart timeframe matches.
Does Deep Backtesting include commission and slippage?
Only if they're set in the strategy's properties — deep mode inherits them, it doesn't add them. A deep test with zero costs on a high-frequency intraday strategy can overstate results by the entire edge.
How much history should a Deep Backtest cover?
Enough to produce a statistically meaningful trade count across more than one regime — for most intraday futures strategies that's measured in hundreds of trades, not years of calendar. Long calendar with few trades is still a small sample.
Can I see individual trades in Deep Backtesting mode?
Not as on-chart markers across the full range — that's the mode's main reading limitation. Export the trade list and audit a random sample of entries against historical bars instead of trusting the aggregates blind.
---
Interface behaviour described as of August 2026; TradingView updates its tester regularly — verify details against current documentation. Not financial advice.