Strategy Tester: What It Does, the Main Platforms and What to Check

A strategy tester is the part of a trading platform that runs a rule, written as code or set up in a form, over historical prices and shows the trades it would have produced: entries, exits, profit and loss, drawdown and statistics such as the win rate. It answers one question, what would this rule have done in the past, and only under the assumptions it was given. This page covers the tool itself: how the main testers work and which settings matter. The method, from writing the rule to the biases that inflate a result, is in our guide to backtesting.

What a strategy tester simulates

The tester reads the bars one at a time. At each bar, the rule decides whether to place, change or cancel orders, and a simulated broker decides whether and at what price they fill: a market order at the next available price, a limit order at its price or better, a stop order once the price trades through it. The fill is where the simulation departs from reality. A bar records four prices, the open, high, low and close, but not the order in which they came. On a daily bar where both a stop and a target were within reach, a tester that only knows the four prices has to guess which one came first. A finer mode, minute bars or real ticks, removes much of that guess.

TradingView documents its default guess: when the open is closer to the high, its broker emulator assumes the price went open, high, low, close; when the open is closer to the low, open, low, high, close. It also assumes no gaps inside a bar. On a wide bar that touched both a stop and a target, that rule alone decides whether the test books a win or a loss, which is why the bar detail setting below matters most for strategies with tight stops and targets.

The main strategy testers

TradingView keeps detailed data for up to the latest 9,000 trades, and its Deep Backtesting, on Premium and higher plans, runs the strategy on the symbol's entire available dataset. MetaTrader 5 adds optimization over parameter sets, forward testing on a part of the period kept apart, multi-currency tests, visual testing, remote agents and the MQL5 Cloud Network. On TrendSpider, long and short strategies cannot be tested at the same time, and an AI Explain button comments on the results (more in our TrendSpider review).

Setting up the TradingView strategy tester

These settings are in the strategy's "Settings/Properties" tab, and a script can set its own defaults in strategy(). The defaults are those of TradingView's Pine Script documentation.

In the strategy() declaration, the matching parameters are commission_type and commission_value, slippage, process_orders_on_close, use_bar_magnifier, backtest_fill_limits_assumption and fill_orders_on_standard_ohlc.

Setting up the MetaTrader 5 strategy tester

The MetaTrader 5 tester runs Expert Advisors, the platform's trading programs; its tick generation mode decides how much of the path inside each bar the test sees. The documentation lists five modes; four of them replay prices:

Three other settings change the result. The execution delay can be set to none, which the documentation calls "ideal" conditions, to a fixed delay such as the measured ping to the trade server, or to a random delay, 0 to 8 seconds 90% of the time and 9 to 18 seconds otherwise. The forward option splits the period into a back test, used for tuning, and a forward test, a half, a third, a quarter or a custom part of the period, always taken at its end, on which the chosen parameters are checked. Profit calculated in pips speeds up a test but leaves out swaps, commissions and margin control, and the documentation advises using it only for a rough estimate. The tester also loads at least 100 bars of history before the start date, so a test that starts close to the beginning of the data is shifted later.

What to check in any strategy tester

• When orders fill. An order triggered by a bar's close should fill on the next bar at the earliest; a test that fills at the signal bar's close knows that close in advance.
• What it assumes inside a bar. Find out how the tester orders the high and the low, and whether a finer mode, lower timeframe bars or real ticks, is available.
• Costs. Commissions, spreads and slippage are often zero by default; set them, then rerun with higher costs than you expect.
• Limit order fills. A limit touched by one tick is not always filled in real trading.
• Chart type. Heikin Ashi, Renko and other non-standard charts show computed prices; run the test on standard candles.
• Data. How far back the history goes, and whether it includes delisted securities. A test on today's list of stocks leaves out the companies that failed: in a study of US common stock funds, published in 1996, Elton, Gruber and Blake measured the gap between the surviving funds and all funds at about 0.9% a year (working paper version).
• Out-of-sample testing. A forward period, or a part of the data kept out of the tuning, is the only part of a report that was not fitted.
• The report. The number of trades, the largest drawdown and the list of trades, not just the net profit.

The pitfalls behind these checks, look-ahead bias, overfitting, selection bias and survivorship bias, are explained with a worked example in our backtesting guide. TradingView's documentation adds that an overfit strategy often fails on new, unseen data, and that slippage is dynamic and unpredictable, impossible to simulate precisely.

The chart pattern figures on this site come from our own test bench, built with these checks in mind: a pattern is recorded on the bar where a trader could have seen it, and everything measured afterward starts on the next bar. One limit remains: on stocks, most companies that were delisted after a bankruptcy or a merger are missing from our data, a serious survivorship bias that we have not measured; stock results lean bullish, and the methodology says so. The full method is in our methodology.

More guides are in technical analysis.