forex-scalping.com

The strategy tester's visual mode: what to look at besides the final result

Most people run the strategy tester, glance at the profit and the equity curve, and move on. Visual mode, where you watch the EA trade on a chart in real time or sped up, is slow and a bit tedious. But it shows things the summary never will.

The first thing I look for is whether trades happen where I'd expect. If the strategy is supposed to buy pullbacks in an uptrend, are the entries actually on pullbacks? It's surprising how often an EA's logic does something slightly different from what you intended.

Second, how stops and targets sit relative to structure. A stop that looks reasonable in pips might be right on an obvious level, where you'd expect it to get hit.

Third, what happens around news and session changes. Does the EA trade into a spike? Does it hold through rollover when spreads widen?

Fourth, the losing streaks. Watching a run of losses play out shows whether they come from one kind of market condition, which might be filterable, or are just random.

I don't watch the whole test. A handful of days in different conditions, a trend day, a range day, a news day, is enough to spot most problems.

Drawing your own marks on the visual tester chart helps. Pause, mark where you would have entered by hand, then resume and compare with what the EA does. If they differ a lot, either the code doesn't match your idea or your idea isn't as clear as you thought. Both are useful to know before trusting the results. It's also a good way to find the edge cases that the rules don't cover properly. If you can't describe the difference between your marks and the EA's entries, the rules probably need tightening.

Speed control makes this bearable. Run quickly through quiet periods and slow down around entries and exits. In MT5 you can also start the visual test from a specific date, which helps when you want to look at a particular day from the results, like the worst losing trade. Going straight to the biggest losses and watching them play out is probably the most valuable fifteen minutes you can spend with the tester. Losing trades usually teach more than the winners, and in visual mode you can see exactly what the EA saw at each step.

Watching visual mode for even a few hours often reveals trades the summary hides: entries just before news, stops hit on spread spikes, or positions opened and closed on the same bar. Those patterns rarely show in the profit figure.

It also helps to compare visual mode with the trade list. If the EA opens trades at prices that look strange on the chart, there may be a problem with how the tester models ticks, which is important to know before trusting the result.

Watching how the EA handles open trades during news is a good use of the visual tester. Find a few big data releases in the test period, slow the tester down around them, and see what the EA does. Does it enter just before? Does the stop get hit by the spike? Does it close early or hold? The final report won't show these details, but the visual mode does. For EAs that are meant to avoid news, it's also a quick check that the news filter actually works in the test.

The name suggests it. The setting is called every tick, the quality bar after the run shows 90 percent, and the natural assumption is that the tester replayed what the market did. It did not.

In the classic MT4 tester, history is stored as one-minute bars. Every tick mode takes each of those bars and generates a series of synthetic price points inside it, following a fixed interpolation routine. The ticks look plausible, they hit the open, high, low and close, but the path between them is manufactured. A real minute might have had 140 quotes with a sharp spike and an immediate reversal. The simulated minute might have a dozen evenly walked points.

The famous modelling quality percentage adds to the confusion. A high number mostly reflects how much of the history is covered by minute data of reasonable quality. It does not certify that the tick path is real. A test can show a very high percentage and still be built entirely from interpolation.

What this affects most is anything that depends on intrabar order: tight stops, tight targets, breakout entries on the first touch of a level, and trailing logic. A strategy built on small moves can look very different once the real quote sequence is available. Strategies that act on bar close and use wide stops are much less sensitive.

The fix is to feed the tester actual tick history, which many people do through imported data files or by using platforms that support real ticks natively. The results often get less flattering, and that is the point.

Has anyone seen a strategy that held its numbers when moved from interpolated ticks to real imported ticks?

What do you look for in the tester's equity line versus balance line? I find the gap between them during open trades tells me a lot. A big gap means the EA lets trades go deep into drawdown before closing them, even if the balance curve looks smooth. That can be a warning sign for strategies with averaging or wide stops. The balance curve can hide risk that the equity curve shows clearly. I'd always look at both in the visual tester, not just the final balance. A big hidden drawdown is easy to miss otherwise.