Skip rules for NAS100 when ATR percentile is extreme

[LondonScalper]

Most of my skip rules are about the calendar. This one is about volatility, and it's the only purely statistical filter I use on index trades.

What I measure. Each morning before the US open I calculate the 14-day ATR for NAS100 on the daily chart and compare it with the past year of daily ATR values. That gives a percentile: where today's volatility sits relative to the past twelve months.

The rule. If the percentile is above 90 or below 10, I skip NAS100 that day.

Why the high end. When volatility is in the top decile, the moves are big enough that my normal stops get hit by noise, and widening the stop means cutting size so far that the trade isn't worth the attention. My results on high-volatility days were sharply negative over the past three years, mostly because of a handful of brutal losers.

Why the low end. When volatility is in the bottom decile, the index often drifts in a narrow range all afternoon. My setups need a bit of movement. On those days I took lots of small losers and scratches, and the costs added up.

I'll admit the thresholds are rough. I tried 85 and 15, and the results were similar. I chose 90 and 10 because it means skipping fewer days.

[Shadow Trader]

Using a percentile rather than a fixed ATR level is smart, because NAS100 volatility shifts so much from year to year. The attached chart is an illustration with made up data of how I picture it: a year of daily percentile readings with the 10 and 90 lines, where about one day in five falls outside the band.

img_12998.png (109.35 KiB) Viewed 36 times

That one in five is the part I'd check. If you're skipping around 20% of days, it's worth knowing which side most skips come from. In my case the low side was skipped far more often during summer, while the high side clustered in a few short bursts.

I'd also look at intraday conditions. A day in the 50th percentile can still open with a 200 point gap after earnings, so I have a second rule: if the first 15 minutes after the cash open exceed twice the average, I wait an hour.

The 85 and 15 test giving similar results is reassuring. It suggests the filter isn't fitted to one threshold.

Do you recalculate the percentile using a rolling year, or a calendar year?

[Shadow Trader]

The day after an extreme reading turned out to be more interesting in my log than the extreme day itself. The first session back inside the band after a run above 90 was often one of my better NAS100 days that month.

My guess is that volatility is still high but falling, so the moves are bigger than normal but more orderly. Stops sized on the inflated ATR have room, and targets get reached faster than usual.

So instead of a plain on and off switch, I run three states. Above 90 is a skip, the first day back below 90 is half size, and after that it's normal size, while the low side stays a simple skip.

One more detail worth checking is which bars feed the ATR. A CFD daily candle includes the quiet overnight hours, so I calculate mine from the cash session only, 14:30 to 21:00 London.

On a few days a year the two versions put the same day on opposite sides of the 90 line. Do you use the broker's daily bars, or something closer to the cash session you actually trade?