How do you handle post-news reversion scalps

[LondonScalper]

After a big release the first move often overshoots, and some of it comes back. I've traded this reversion for years with mixed success, and I'm curious how others approach it, because my own rules feel more like guidelines than a system.

What I do now. I only look for reversion on releases where the first M1 candle is at least twice the size of the average M1 candle in the previous hour. Smaller moves don't have enough to give back.

I wait for two things before entering: the spread has dropped back to within 50% of normal, and price has made at least one M1 candle that fails to make a new extreme. Before that I'm guessing.

Target is a retracement of about a third of the initial move, measured from the release price to the extreme. Stop goes beyond the extreme plus a little room for the spread.

My numbers from the past twelve months, mostly on EURUSD and USDJPY after US data: 34 trades, 19 winners, small positive expectancy but heavily dependent on three big winners after the March and May CPI prints. Without those, it's roughly break-even.

What I keep wondering is whether the edge is real or whether I'm just picking up a few nice moves and calling it a strategy.

How long do you wait after a release before you even consider fading it, and do you treat CPI differently from NFP?

[Fairman]

What I keep wondering is whether the edge is real or whether I'm just picking up a few nice moves and calling it a strategy.

With 34 trades and the result resting on three of them, the honest answer is probably that it can't be told apart yet. One quick check is to recompute expectancy with the top three winners capped at the size of your average winner. If it goes negative, the strategy depends on outliers, which isn't necessarily bad, but then it's a CPI strategy more than a general reversion one. That matters because it changes how you'd size it and when you'd trade it. Splitting by release type would answer your CPI question directly too. Even 8 or 10 CPI trades against the rest tells you whether the rest is just paying spread while waiting for the big ones.

[Shadow Trader]

Your two entry conditions match mine closely, but I'd add a third. I want the release price itself to be at least 10 to 15 pips away from the extreme before I fade. On smaller first moves, a third of the move barely covers the spread plus commission.

The attached chart is an illustration of how I measure it on a USDJPY release: candle at 13:30, extreme two minutes later, entry on the first M1 bar that fails to make a new high, and the target at a third of the move back.

img_13065.png (71.5 KiB) Viewed 20 times

On CPI versus NFP, I treat NFP as worse for reversion. The first move on payrolls often gets revised in the next few minutes when people read the revisions and wages, so the "extreme" isn't really the extreme. CPI tends to settle faster.

I also wait longer than I used to, about three minutes minimum. Most of my bad reversion trades were the ones taken inside the first 90 seconds.

How many of your 15 losers were stopped beyond the extreme, as opposed to just timing out?

[Shadow Trader]

Fairman's capping test is a good one, so I ran it on my own reversion log. Of 41 trades, the top three winners were each about four times my average winner.

Capping those at the average took the expectancy from roughly +0.15R to slightly negative. So by that test mine is an outlier strategy too, and I'd rather know that than keep calling it general reversion.

The split by release type told me more. CPI was 11 of the 41 trades and carried most of the result, while the other 30 were close to just paying spread.

Looking at the losers by timing gave me one more filter. When the extreme came more than about five minutes after the release, it was usually a repricing rather than an overshoot, and fading it lost more often than it won.

Now I only fade if the extreme is in by 13:35 on a 13:30 release. After that the market has had time to read the details, and I'm arguing with a new price.

LondonScalper, did your March and May winners come on big misses against the forecast, or on moves that looked bigger than the miss justified? That might tell you which of the two to filter for.