Pick a date in 2023, click replay, and the chart stops there. Everything to the right of the cursor is hidden. Press the forward button and one candle appears. That is the whole feature, and it is the closest a discretionary trader gets to a laboratory.
The value is not the button. It is that you have to decide with the same information you would have had live: no right-hand side of the chart, no knowledge of where the day closed, no hindsight about the news release that moved it. Most strategies that look excellent on a static chart fall apart in the first twenty replayed setups, because on a static chart the eye finds the entries that worked and skips the fifteen that did not.
How the mechanics actually behave
Replay runs on the timeframe you have open. Start it on the daily and each click advances a full day. Start it on the 5 minute chart and each click advances five minutes, which is where the tool gets slow and where most people quit. The practical setup is to open replay on the timeframe you take entries on and add higher timeframe context in a second pane, because the replay cursor is shared across the layout.
Two limits matter. First, intraday history on lower timeframes is shorter than daily history, so a 1 minute replay of a 2019 session is usually not available at all. Second, a replayed candle is drawn as a finished candle, so you never see the wick forming, the retest that failed by two pips, or the spread widening around a release. That last point is not a small detail. If your rules depend on where price was inside the bar, replay is measuring a friendlier market than the one you trade. Candle close times and the broker's session boundaries shift bar shapes too, so a replay run on one data source will not reproduce exactly on another.
The discipline that makes the sample honest
Manual backtesting fails for a reason that has nothing to do with software. Nobody records the losers. The fix is mechanical: before you advance the chart, write the entry, the stop, the target and the reason in a sheet or in a trading journal. Then advance. Then record the outcome and move on without touching the entry you already wrote.
Two rules that raise the quality of the sample sharply:
Never rewind after seeing the result, and never change the rule mid-sample. If the rule needs changing, stop, change it, and start a fresh sample from a new date. A sample where the rules drifted halfway through measures nothing.
Sample size is the other trap. Fifteen trades tells you almost nothing about a setup, and a run of nine winners in a row inside those fifteen will convince you otherwise. A few hundred occurrences across different market conditions is where the numbers start to settle. That is a lot of clicking, which is why serious manual testers pick one setup and one pair rather than trying to validate an entire plan in a weekend.
Where replay beats a coded backtest, and where it does not
A coded backtest in Pine or in MetaTrader's tester will run ten thousand trades while you make coffee, and it will do it identically every time. What it cannot test is a rule you have not managed to write down. Most discretionary methods contain judgement: whether a level counts as clean, whether the structure looks impulsive, whether a session is worth trading at all. Replay tests that judgement directly.
It also trains something a coded test cannot. Sitting through forty replayed bars in a losing position is unpleasant in roughly the way the live version is unpleasant, minus the money. Traders who have done a few hundred replayed trades hesitate less at the entry, because the pattern is familiar rather than theoretical.
What replay is bad at: costs and execution. It will not deduct spread, it will not model slippage on a stop, and it will not requote you. Take the number your replay run produced and subtract realistic transaction costs by hand before you decide anything. A setup with an average win of eight pips does not survive contact with a real bid ask spread.
A workable routine
The version that produces usable data in a week of evenings: choose one instrument, one entry rule and one exit rule. Choose a start date at least a year back so the sample crosses a change in volatility. Advance in fixed increments and log every occurrence of the setup, taken or skipped, with the reason. Stop at a pre-decided number of occurrences rather than when the results look good.
Then run the same rules forward on a demo account for a month. Replay measures the rule; forward testing measures the rule plus you plus your broker's execution, and the gap between the two numbers is the most informative thing you will learn. If the forward result is far worse, the difference is usually costs, hesitation, or a rule that was quietly bending during the replay.
Fitting it into a plan
Replay results belong in the same document as your risk rules, not in a separate mental note. A setup that produced a 1.4 average reward to risk over 200 replayed occurrences tells you what position size and what reward to risk assumptions the plan can carry. It does not tell you the setup will repeat. Markets change regime, and a sample drawn entirely from a trending year describes a trending year.
Trading is high risk and a favourable backtest is not a forecast. What replay buys you is the ability to discard bad ideas cheaply, before they cost real money. That is a smaller claim than most backtesting content makes, and it is the one that holds up.
"Replay is only useful if you write the trade down before the next candle prints. The moment you rewind because you did not like the outcome, you are not testing anything."
— Roman Onta, Executive Director, SINGUARD
Key Takeaways
- Replay hides the right side of the chart, which is the only way to test a discretionary rule without hindsight.
- Log every occurrence before advancing the chart, including the ones you skip, or the sample is worthless.
- Replay draws finished candles, so it hides intrabar behaviour, spread and slippage. Subtract costs by hand.
- Validate a replay result with a month of forward testing before it influences real position sizing.
Frequently Asked Questions
Does TradingView replay work on all timeframes?
It runs on whatever timeframe the chart is set to, but intraday history is shorter than daily history, so very old 1 minute or 5 minute replays are often unavailable. Higher timeframes go back much further.
Is manual replay backtesting better than a coded backtest?
They answer different questions. A coded backtest runs huge samples identically but only tests rules you can express in code. Replay tests discretionary judgement, at the cost of speed and of the discipline required to stay honest.
How many trades does a replay sample need?
Far more than most people run. A few dozen occurrences is noise. Aim for a few hundred across different volatility conditions, and treat any result drawn from a single market regime as untested elsewhere.
About the Author
Roman Onta is an Executive Director at SINGUARD. He builds the Prop Firm CRM, the Broker CRM, Scalegram and CopySignals side by side with his brother Alex Onta, and he helped on the design of eTrader, the division Alex built and leads. His ground is worldwide payment processing, AML compliance and the corporate structures brokers are built on, work the two of them carry together, shaped by executive roles in the UAE and international corporates. He lives and works in Dubai for most of the year. Meet the executive duo leading Singuard's five divisions.