The answer first

What TradingView backtesting really is (the answer first)

TradingView backtesting is the process of mechanically reproducing a set of trading rules on historical chart data to see how a strategy behaved in the past. The first thing to internalize is that the object of the review is not the final net profit. Net profit moves dramatically with how you set your assumptions, so deciding whether a strategy is good or bad from that figure alone leads to bad judgment.

A sound backtest checks for fragility in this order. First data quality (no look-ahead or repainting), then repeatability (does the same result appear under the same conditions), then loss behavior (how deep does it sink, how long are the losing streaks), then costs (have spread, commission and slippage been included), and finally out-of-sample performance (does it hold up on periods not used to build it). Only after working through these five layers step by step can you say you have actually validated a result.

This article is the parent page for ten specialist articles. It points to the entrance of each stage and links to the individual articles for deeper procedures. Every number and figure shown here is fictional educational data; none of it is a recommendation of a specific strategy or a suggestion of future results.

Clarifying terms

Strategy Tester, Strategy Report, Deep Backtesting and forward testing

Sorting out similar-sounding terms before you start saves confusion later. A strategy is the script itself, written in strategy() form, which defines the trading rules. The Strategy Tester is the panel that runs that strategy and lets you control and view the results. Inside the panel are the Strategy Report (Performance Summary), which collects the aggregate figures, and the List of Trades, which shows each trade individually.

Deep Backtesting is a mode that runs over a longer period or finer data; it is heavier than a regular backtest but lets you increase the sample. Forward testing is a separate check that looks forward in time, closer to live operation, rather than at the past. Because reproducing the past (backtesting) and observing the future (forward testing) have different aims, avoid using the result of one to speak for the other.

Button names and panel layouts can change. Confirm the latest UI names in the official documentation, and do not assert specifications you cannot verify. If you want to move on to the exact steps for pulling out a CSV, the guide on exporting Strategy Tester data to CSV is the concrete next stop.

The full picture

The full picture: seven steps from backtest to robustness testing

Do not treat validation as a one-off. Start by locking your assumptions and move in one direction until you are looking at out-of-sample stability. The SVG below lays out those seven stages from left to right (it scrolls horizontally).

Steps 1 to 3 are about pulling out the raw material correctly, steps 4 to 5 are about reading that material correctly, and steps 6 to 7 are about testing whether it stays intact when the assumptions break. The sections below walk through each stage in order.

Pulling out the material

What to export: Performance Summary vs List of Trades

Staring at the numbers on screen is not validation. The starting point is to export the results to CSV or XLSX so you can re-read them in your own hands. There are broadly two kinds of information you can export.

• Performance Summary: aggregate values such as net profit, win rate, Profit Factor, maximum drawdown and trade count. Good for grasping the whole picture of a strategy on a single sheet.
• List of Trades: the trade-by-trade detail of each entry and exit, profit and loss, and holding period. Good for examining the sequence of losing streaks, the source of drawdowns and exit efficiency in fine detail.

Aggregate values are convenient, but they are also rounded results. The order of losing streaks and the depth of a drawdown can only be tracked accurately from the detail (the List of Trades). Because CSV column layouts vary by environment, being able to map columns on the import side keeps things stable. The concrete export procedure is collected in the CSV export guide.

Reading correctly

Reading KPIs: never judge on a single number

Net profit, win rate, average win and loss, expectancy, Profit Factor, Sharpe, maximum drawdown and trade count — pull out any one of these on its own to evaluate a strategy and you will misread it easily. The following is an illustration using fictional educational data. The numbers themselves carry no meaning; they exist only to show how the metrics should be cross-checked against one another.

In this example the win rate is 47%, below half, but because the average win (+1.56R) is larger than the average loss (−1.00R), expectancy comes out on the positive side at +0.20R per trade. Dismiss it as a “weak strategy” from the win rate alone and you miss that relationship. Conversely, a Profit Factor of 1.38 tells you nothing about whether the risk is bearable until you read it alongside the −15.3% maximum drawdown and the sequence of losing streaks. The definitions and reading of each individual metric are detailed in the guide on reading backtest results.

Testing for fragility

Risk and robustness: drawdown, Monte Carlo and stress tests

Two strategies with the same net profit can have completely different characters depending on how deeply they sank on the way to that profit. The figure below pairs an equity curve on top with an underwater curve below it (the decline from the running peak), aligned vertically. Both use fictional educational data.

Looking at the top panel alone, it appears to climb steadily; overlay the bottom panel and you can see there was a stretch where it sank −15.3% from its running peak. Whether you could withstand that decline in live trading is best thought about not in currency but in terms of the percentage and the time it took to recover. How to calculate drawdown, recovery periods, and metrics such as Calmar and Ulcer are covered in the maximum drawdown analysis guide.

Beyond that, Monte Carlo analysis tests whether this result simply happened to occur in that particular order. By reshuffling or resampling the sequence of trades, it shows the depth of losing streaks, the distribution of drawdowns and an indication of the risk of ruin. How to read sequence risk and the risk of ruin is explained in detail in the Monte Carlo and risk-of-ruin guide.

Testing out of sample

Out-of-sample testing: OOS, walk-forward and parameter stability

The most commonly overlooked pitfall in validation is only ever checking a result on the period you built it on. Of course it looks good on the period used to build it (in-sample, IS); the question is whether it also holds up on the period you did not use (out-of-sample, OOS). A design that splits the data, decides conditions on one part, and confirms them on the other is called walk-forward. The thinking behind the split is made concrete in the walk-forward and out-of-sample testing guide.

Another thing that matters is parameter stability. When only one particular parameter setting produces a standout result, that may not be the strategy’s true ability but overfitting to that data. Checking whether nearby values continue smoothly (a “plateau”) or collapse when you nudge them slightly (a “cliff”) makes it easier to judge whether the choice is sound. How to spot overfitting is covered in the overfitting and parameter stability guide.

The idea of not deciding on a single KPI can be drawn as a wheel. You look at five perspectives at once, and wherever one is low, that is the weak spot (the values are fictional educational data).

Advanced diagnostics

Advanced diagnostics: MFE/MAE, cost sensitivity and multi-strategy portfolios

Once the basic validation is in place, the next step is exit efficiency, cost assumptions and the perspective of the whole portfolio. These are not mandatory, but they start to matter when you are refining a strategy.

• MFE / MAE: look at how much unrealized profit (MFE) and unrealized loss (MAE) each trade carried while it was open, to visualize whether your take-profit and stop-loss placements are reasonable. Details are in the MFE and MAE exit analysis guide.
• Cost sensitivity and stress tests: progressively tighten spread, commission and slippage to see how far the assumptions have to worsen before performance breaks down. Details are in the cost sensitivity and stress test guide.
• Multi-strategy portfolios: taking correlation and allocation between strategies into account, look at the combined drawdown to see the whole picture. Details are in the multi-strategy portfolio guide.

Avoiding pitfalls

Common mistakes and how to avoid them

Most validation failures collapse into the same handful of patterns. If any of these ring a bell, that is your entry point to the next article to read.

• Look-ahead bias (repainting): using information from a candle that has not yet closed. Catch it early with a data-quality check.
• Missing costs: running without spread, commission or slippage, which makes results look better than live trading.
• Too few trades: a small sample where the result is swayed by chance.
• Period or symbol bias: depending on a single instrument or a single market condition and failing elsewhere.
• Picking a single best parameter: choosing one standout point without checking the stability of its neighbors.
• Post-hoc storytelling: inventing reasons to fit the numbers that came out. Fix your hypothesis in advance.

Start free

Start free: using the TradingView Backtest & Robustness Lab

Everything covered so far can be tried directly on your own data with the TradingView Backtest & Robustness Lab. It is designed in stages by depth of analysis: confirm the basics free first, and consider higher-tier features once you need them. Because pricing can change, this article does not fix a figure; check the current scope on the plan page.

This tool mechanically aggregates, diagnoses and visualizes the backtest results you load; it is not investment advice, a buy/sell signal, a forecast, a performance guarantee, automated trading, order execution or a source of real-time market data. Depending on your needs, you can also make use of the free TradingView indicator collection you can add to charts, and a cross-comparison of SG Group’s financial tools.

Self-check

Backtest health self-check

The check below is an educational self-assessment. It does not issue a pass/fail or a trading decision; it only indicates which article might be worth reading next. Your input stays in the browser and is not sent anywhere. First, here is a static reference guide by number of checks.

Learning roadmap

Learning roadmap (10 articles)

Starting from this parent guide and working through the cluster in beginner, intermediate and advanced order builds up the full picture of validation.

Summary

Summary: the answer to the main question and your next step

What you should really check in a TradingView backtest is not whether it made a profit. It is the work of confirming data quality, repeatability, loss behavior, costs and out-of-sample performance in one direction across seven steps. Net profit and win rate are only the entrance; the character of a strategy only becomes visible once you bundle trade count, drawdown, expectancy and robustness together and read them as a whole.

As a practical checklist: (1) lock your assumptions (period, symbol, timeframe, costs) first, (2) export the results to CSV and re-read them, (3) do not judge KPIs in isolation, (4) look at how it sinks with drawdown and Monte Carlo, and (5) confirm resistance to breaking with OOS and parameter neighborhoods. Hold to these five points and you will not go far wrong.

Related articles

Related reading