Free forex backtesting is not a myth. MetaTrader ships a strategy tester at no cost, TradingView includes Bar Replay on its free plan for daily and higher timeframes, and at least two open-source Python engines can be installed today without paying anyone. The catch is not price. It is that the free tier decides which questions you are allowed to ask, and several of the well-known free tools have quietly stopped receiving updates.
What the free tier removes
Every free backtesting product withholds something specific, and the something is usually data depth or execution realism. A tester can only simulate what it can see, and free plans tend to cap history, restrict timeframes, or exclude the tick-level detail that separates a realistic fill from an optimistic one.
Before you run anything, write down which of these your tool actually provides: historical depth in years, the finest available interval, whether spreads are fixed or modelled from actual quotes, and whether overnight swap is applied. If you cannot answer all four, the backtest output is a graph rather than a measurement.
MetaTrader's tester and the broker data problem
Both MetaTrader 4 and MetaTrader 5 include a strategy tester, and MetaTrader 5 added real tick modelling and multi-symbol testing on top of the older engine. The limitation is not the software. It is the history file, which comes from your broker and varies enormously between brokers for the same pair and the same period. Feed a tester thin data and it will still produce a smooth equity curve, because it interpolates over the gaps without warning you. Before trusting a result, download the history, count the bars, and compare the missing days against a second source. If your broker's EUR/USD history is missing three weeks of 2020, you did not test 2020.
TradingView replay, by plan
TradingView publishes its own plan matrix, which makes this easy to state precisely. Bar Replay covers historical data by day and higher timeframes on every plan including the free Basic tier. Replay of minute-based data starts at the Essential tier with 180 days of one-minute history, extends to 365 days at Plus, and becomes all available intraday history from Premium upward. Deep Backtesting, which runs a Pine strategy over the full history rather than the bars loaded on your chart, is a Premium feature. The free Strategy Tester is real and useful, but it tests the chart's loaded data, and the free plan carries a limited historical bar allowance.
That makes free TradingView a reasonable choice for daily-chart swing ideas and a poor one for anything you intend to trade on a 5-minute chart.
Python libraries and the maintenance question
The open-source route has a maintenance problem that no tutorial mentions. Backtrader was for years the default Python backtesting framework for retail traders, and its last significant development activity was around 2020; its built-in Yahoo Finance data feed stopped working when the upstream service changed. Zipline was the engine behind Quantopian, which shut down on 3 November 2020 and sold its intellectual property to Robinhood, and the original project did not continue. The community fork zipline-reloaded, maintained to keep the library running on current versions of pandas and numpy, is the practical descendant.
Active alternatives exist. QuantConnect's LEAN engine is open source and can be self-hosted at no cost, works in C# and Python, and supports forex among other asset classes. The trade-off is setup effort and a steeper API.
| Tool | Cost to start | Data comes from | Practical status |
|---|---|---|---|
| MetaTrader 4 / 5 | Free with a broker account | Your broker's history files | Maintained, quality varies by broker |
| TradingView | Free plan, paid tiers above | TradingView feeds | Maintained, depth tied to your plan |
| Backtrader | Free | You supply the feed | Last significant activity around 2020 |
| Zipline | Free | Custom data bundles | Original project stopped after 2020 |
| QuantConnect LEAN | Free if self-hosted | Your own or licensed feeds | Actively developed |
Numbers worth reading, and one worth ignoring
Win rate is the number everyone quotes and the least informative of the set. A 70% win rate with an average winner of 4 pips and an average loser of 40 pips is a losing system, and it will look wonderful for two months. Read profit factor, which is gross profit divided by gross loss, alongside maximum drawdown, the largest peak-to-trough decline in equity. Then read the number of trades. A result built on 24 trades tells you almost nothing, because the confidence interval around the win rate is wider than the difference between a good system and a coin flip.
The number to ignore is the ending balance.
It is a single number produced by one path through history, and history only happened once. Two strategies with identical rules can produce wildly different equity curves depending on which year you started. If a vendor leads with a final balance and no drawdown figure, that is a description of a marketing asset rather than a tested method.
Testing across regimes, not windows
Pick periods that behave differently from each other, and pick them before you see the results. Currency markets in a trending interest-rate cycle look nothing like the same market during a range-bound policy pause, and a strategy tuned on one will usually fail on the other in a way that a single five-year test hides. Run the same rules on at least three distinct periods, keep the parameters fixed between runs, and treat any period that requires a parameter change as evidence against the rule rather than for it.
For regulatory context, the Commodity Futures Trading Commission and the National Futures Association both publish retail forex education material, and the NFA's BASIC database is the register for checking whether a firm you intend to trade with is actually a registered retail foreign exchange dealer. Backtesting tells you about a strategy. It never tells you about a counterparty.
General information only, not investment advice. Tool features and pricing change, so confirm current plan details with the provider and current registration status with the relevant authority.