forex-scalping.com

A tick chart for a major pair looks like a continuous line. A tick chart for something like NZDCAD or EURNZD looks like stairs: price sits at one level for a few seconds, then jumps three or four points in one step.

Those gaps matter for entries and stops. A stop order at a price that sits inside a gap won't trigger until the next tick, and it fills at the new price. The loss is bigger than planned by the size of the jump.

It also changes what "tight" means. A 5 point stop on a pair whose ticks regularly jump 4 points is basically a coin flip about where the fill lands.

Counting the typical tick-to-tick change on a pair, in points, is a quick way to judge whether the instrument fits a tight style. The average and the 95th percentile both matter.

It also explains why backtests on bar data flatter these pairs. A one-minute bar hides the staircase and lets every stop fill at its exact level.

A rough measure: take a day of ticks, compute the absolute difference between consecutive prices in points, and look at the 50th, 90th and 99th percentiles. If the 99th percentile is bigger than the intended stop, a fill at the stop price is the exception, not the rule.

Which pair showed the biggest tick-to-tick jumps when it was last measured on your feed?

I ran exactly that percentile check on a handful of crosses last spring after a run of stops that filled worse than I'd set them. EURNZD was the worst on my feed by a long way. The median tick to tick change was small, but the ninety ninth percentile was over five points, which was bigger than the stop I was using at the time. No wonder my losses kept coming in above plan.

GBPNZD wasn't far behind, while the majors were tidy by comparison. On EURUSD in London hours, the ninety ninth percentile was barely more than a point, so a stop there really did mean roughly what it said.

Since then I won't trade a pair with a stop smaller than about twice that ninety ninth percentile figure, measured in the hours I actually trade. It ruled out most of the exotic crosses for my style, and it also made me rerun a couple of old backtests on tick data, which knocked a surprising amount off the results.