Reading a record

Backtest, out of sample, forward-tracked: what's the difference?

The three kinds of trade on our strategy pages, how each one was produced, and how much weight each deserves.

16 October 2026 · 3 min read · All articles

Every trade on our strategy pages carries a label: Backtest, Out of sample or Forward. In a trade list they look alike, with the same columns and the same points, but they were produced in different ways and deserve different amounts of trust. Many strategy pages give one headline number without saying which kind of trade it comes from. This is how to tell them apart, and why we keep them separate.

1. Backtest

A backtest runs the strategy over past prices and records the trades it would have taken. Ours come from TradingView's strategy tester, with 1 tick of slippage on market and stop orders.

The catch is how the settings were chosen. A strategy's settings (a stop size, a filter, a time window) are picked by looking at results on past data, usually the same data the backtest then shows. So a backtest tells you how well the settings fit the history they were chosen on. That still matters: a strategy that cannot even fit its own history is finished. But it says little about the next few months.

What we do: each strategy page says which data the settings were chosen on. Midas Model's settings were chosen on the same backtest period its page shows, and the page says so.

2. Out of sample

Out-of-sample trades come from the same backtest, but from after the last day of data used to choose the settings. Those trades played no part in the choice, so they are a fairer test of whether the rules carry over to prices they have not seen.

ICT22 Model's settings were chosen using data up to 27 July 2026. Its backtest runs to the start of October, so the 16 trades from 4 August onwards are out of sample. On its page they are marked Out of sample in the trade list and shaded on the chart.

Two limits. Out of sample is still a backtest: it can be re-run, and if the code changes, so can the trades. And it is only as clean as the process behind it. Had we looked at those weeks and gone back to adjust the settings, they would no longer be out of sample. Ours have not changed since they were chosen, but you have to take that on trust, which is why the third kind matters most.

3. Forward-tracked

Forward-tracked trades are recorded one by one as they close, from the strategy running in real time on TradingView. Both our strategies have been forward-tracked since 1 October 2026. Each trade is stored when it closes and is not re-run or revised afterwards. Nobody, us included, knew the outcome when the strategy took the trade.

There is one exception, and we name it. For the first days our server rejected the tracking alerts because of a key error, so the first forward trades, both on 5 October, were restored by hand that day from TradingView's alert log. The forward record explains it.

Forward-tracked is the hardest evidence of the three, and for now it covers the shortest period. That is why the strategy pages hold back the forward win rate and profit factor until there are 30 forward trades: from a handful of trades they are mostly luck.

What none of them is

All three use TradingView's simulated fills. None is a brokerage statement and none involves real money. Real fills can differ: a limit order TradingView counts as filled may not fill at your broker, and market and stop orders can slip more than the 1 tick we allow for. Forward tracking removes hindsight. It does not remove the gap between a simulation and an account.

How to read our strategy pages

When you compare strategies, here or anywhere else, ask which kind of record a number comes from. A headline that blends all three is mostly the first kind.