forex-scalping.com

There's a gap between "the backtest looks good" and "this runs with real money". A formal promotion process makes that step less of a leap.

A staged approach that some EA builders use:

Stage 1, research: backtest with realistic costs, walk-forward checks, parameter sensitivity. Exit criteria written in advance (e.g. profit factor above 1.2 after costs on out-of-sample data).

Stage 2, demo forward test: 4–8 weeks on a demo with the same broker and server as the live account. Compare each trade with what the backtest would have done over the same period.

Stage 3, live minimum size: the smallest lot the broker allows, for at least 100 trades. The goal isn't profit; it's to measure real slippage, rejects and behaviour around news.

Stage 4, scaled live: increase size in steps only if stage 3 results are within the expected range from stages 1–2.

At each stage, define what would send the EA back a stage: e.g. live slippage more than double the backtest assumption, or drawdown beyond the 95th percentile of the Monte Carlo estimate.

It feels slow. It's much cheaper than finding out at full size.

How do you decide when an EA is ready for real money?

Demo accounts are a fine way to learn a platform. As a test of a scalping strategy, they're misleading in ways that favor the trader.

Demo servers typically fill at the quoted price instantly. There's no last look, no rejection, no slippage. Orders never compete for liquidity, since they don't reach a real market. Spreads on demo are often tighter than on live, or at least more stable.

Live accounts pass through real execution. Orders travel to liquidity providers, sit in queues, and sometimes bounce. Slippage appears. Spreads widen when the market is stressed, and in a way a demo never reproduces.

The gap is largest for the strategies that depend on speed and precision: tiny targets, tight stops, entries on breakouts. A demo might show a 70 percent win rate that falls to 52 on live, simply because of the costs and execution frictions the demo ignores.

Another difference is psychological. Nothing hurts on a demo. Hesitation, flinching at the close button, and holding losers too long all show up only when real money is at stake.

The sensible step between demo and full size is a micro live account. The goal isn't profit. It's to measure the real fill quality, with all its flaws, at a cost small enough to ignore.

A few specific numbers are worth capturing during that stage: average slippage on entries, average slippage on stops, rejection rate, spread at the time of entry compared with the quoted average, and execution time.

Those five numbers explain most of the difference between a demo result and a live one. If the live numbers are close to the demo assumptions, the strategy has a chance. If not, the strategy needs adjusting before size goes up.

At the end, rerun the backtest with the measured slippage and delay plugged in. If the adjusted result still holds a decent edge, size can increase gradually. If the edge collapses, the strategy was relying on conditions that the live environment doesn't provide, and that is a far cheaper lesson to learn at micro size than at full size.

How big was the gap between demo and live results the first time they were compared here?