Your EA passes the Strategy Tester and fails live

Almost always it is one of four things, and all four are visible before you risk anything.

It is usually the modelling, not the market

The Strategy Tester offers several ways to invent what happened inside a bar. Open prices only assumes a single price per bar. Every tick builds a plausible path between the open, high, low and close using an interpolation, and plausible is not the same as real. Real ticks replays what the broker actually recorded.

The gap between them is not cosmetic. A strategy whose stop and target can both be reached inside one bar gets a verdict from the interpolation rather than from the market, and the interpolation resolves the tie the same way every time. Run the same rules under two modelling modes: if the equity curves disagree materially, the strategy is reading the model, not the price.

Spread in the tester is not the spread you trade

Backtests frequently run on a fixed or current spread while the account trades a floating one that widens at the open, at news, and in the last minutes of the session. A scalping strategy is priced almost entirely by spread, which means the tester is measuring an instrument that does not exist.

Test with the widest spread you have actually observed on your account, not the average. If the edge survives the bad case it is an edge; if it only survives the average it is a hope about liquidity.

Fills are assumed, not simulated

The tester fills your order at the price you asked for. A live broker fills at the price available, requotes, applies a minimum stop distance, and may not let you place a stop as close as your rules do. Strategies that depend on tight stops are the ones most exposed to this, and they are also the ones that look best in a test.

Check the symbol's stop level and freeze level against the distances your rules use. If your stop is inside the broker's minimum, the tester ran a strategy your account cannot place.

The history you tested may not be the history that happened

Downloaded history is a broker's reconstruction, and different brokers disagree, particularly on the extremes of a bar, which is precisely where stops and targets live. Symbols also change: contract sizes, swap rates and trading hours are edited and the tester applies today's settings to years of old bars.

This is the least fixable of the four and the most important to know about, because it puts a floor under how precisely any backtest can describe the future.

What to do about it, in order

Re-run under a second modelling mode and compare. Re-run at your worst observed spread. Compare your stop distances against the symbol's minimums. Then forward test on a demo account fed by the same broker, which is the only test that uses the same data, the same spread and the same fill logic as the account you intend to trade.

A divergence that survives all four is information about the strategy. Anything you skip is a divergence you will meet later with money on it.

Questions

Is 'every tick based on real ticks' always the right modelling mode?

It is the most faithful mode the tester offers, and it is slow. The practical approach is to develop under a fast mode and to confirm anything you intend to trade under the most faithful one available. A result that only exists under the fast mode is a result about the interpolation.

Does a high modelling quality percentage mean the backtest is accurate?

No. It describes how completely the tester was able to reconstruct price movement inside bars from the data it had. It says nothing about whether your spread, commission and slippage assumptions match your account, which are the assumptions that usually decide the outcome.

Should I trust a demo forward test?

More than a backtest and less than live. A demo uses the broker's real feed and real spread, which removes the two biggest backtest errors, but it does not queue your order against other people's, so it still flatters fills.

Sonar Sciences is a research and validation platform, not an investment adviser. Nothing on this page is investment advice or a recommendation to trade any instrument or strategy. Platform names are trademarks of their respective owners and we are not affiliated with them. Trading leveraged instruments involves substantial risk of loss.