Describe a strategy in words, get a backtest

From a paragraph to a strategy you can read

Write 10 to 2,000 characters describing what you do: the instrument, when you get in, where the stop and target sit, what you avoid. That text goes to DeepSeek V4 as data, never as instructions, and comes back as one strategy in the same rule format the builder uses, plus a list of anything your description said that the rules cannot express. A draft that fails validation is sent back once for a second attempt; two failures and you get the validator's own errors instead of a broken draft. The strategy lands in the builder with a banner marking it as a draft. Nothing is saved, and nothing runs, until you look at the rules and choose to run it yourself.

28 indicator lines

A rule can compare price, a number, or an indicator line against another, at a bar you choose (up to 50 bars back). The original ten were SMA, EMA, WMA, RSI, ATR, MACD, Bollinger Bands, Stochastic, ADX and Donchian channels. Eighteen more followed: ROC, Momentum, standard deviation, Envelopes, TRIX, Aroon, HMA, DEMA, TEMA, Awesome Oscillator, Ultimate Oscillator, Chandelier Exit, Fractals, Alligator, a linear-regression channel, and three that read volume: OBV, MFI and CMF. Every line uses the same parameters and defaults as the desk's own chart, so a rule reads a value the same way your eye does when you look at the chart.

What tick volume means on forex

There is no single forex exchange, so there is no honest count of contracts traded the way a stock market has one. What FxMind's feed records instead is tick volume: how many price updates a bar saw, which is the volume figure OBV, MFI and CMF actually read. A bar nobody ever recorded volume for reads zero, never a guessed number. And a zero is treated as a missing reading, not a neutral one: a rule checking one of these three indicators over a window with no recorded volume at all sees nothing and cannot fire on it, and the run's own report says how many bars were affected in plain language, next to the other reasons a run might hold back.

The fill rules that stop a backtest from flattering you

1. 1A market order fills at the close of the bar it decided on. No peeking at the next bar's open to be kind to the result.
2. 2When a stop and a target both sit inside one bar, the stop is taken. A bar cannot say which price came first, so guessing the kinder one is not a backtest.
3. 3A price that gaps past a stop fills at the open, not at the stop. A stop is an instruction, never a promise.

Runs, bars and drafts, by plan

Every plan, Free included, gets automated runs; what scales is how many a month, how much history one run can cover, and how many word-described drafts a day.

Manual practice has its own limit: Strategy Lab keeps 5 sessions at once on Free, 25 on Trader, 200 on Pro, 500 on Elite and 1,000 on Desk. Describing a strategy in words is the quick way to get rules to test; replaying a setup bar by bar is how you learn to trade it. See /features.

What a finished run shows you

Net result after costs, gross before them, a profit factor once there is a win and a loss to compare, drawdown as a share of the peak, Sharpe and Sortino on daily returns, a breakdown by year and by month, and why entries were held back (a filter, the trade cap, a size below the minimum, both directions firing on the same bar). The in-sample and out-of-sample split sits next to the whole result, so you can see whether the rules still hold on the part of the range they were not built around. A run also stores a hash of every bar it read, so running the same rules again on the same data gives you the same trades back.

Monte Carlo: a different tool, for real trades

Once you are placing real trades, a related but separate tool asks a different question: not "did this rule work on the past" but "what range of outcomes could my own results produce next." Monte Carlo resamples your own closed trades, thousands of times, and shows the spread of equity paths and drawdowns that history could still produce. It never touches a backtest run and never forecasts the market; it reshuffles trades you already made. It is on every paid plan, Free excluded.

What is not built yet

An automated strategy only places market orders at a bar's close; pending orders, partial closes and adding to a position are not supported, so the language refuses them rather than approximating them. And the data underneath is 1-minute candles, never ticks or seconds: a strategy is tested at the resolution FxMind actually stores, never finer than that.