forex-scalping.com

You buy a EURGBP breakout above the morning high, it fails, and you're stopped out. Now what? Do you fade the failure, or wait for the breakout to try again?

I've been tracking this for about eighteen months because I kept doing both, randomly, depending on mood.

Here's what my log says. After a failed breakout of the 07:00 to 08:00 range, defined as price closing back inside the range on M5 within 20 minutes of the break, there are three common outcomes.

Price returns to the opposite side of the range: about 40% of my 87 samples.

Price chops around in the middle of the range for the rest of the morning: about 35%.

Price retests the original breakout level and breaks out again successfully: about 25%.

So fading the failure toward the far side of the range has worked a bit better than following, but not by much, and the choppy outcome is common enough to eat into both.

What tilts it for me is spread and time. If the failure happens before 09:00 with normal spread, I'm more likely to fade. If it happens after 10:00, I tend to do nothing, because EURGBP in late morning often just drifts.

What I don't do anymore is immediately reverse at the same size after being stopped out. That was emotional. Now I wait for a clear M5 close back inside the range and use half size. Which way do you lean when a EURGBP break fails?

LondonScalper wrote: Price returns to the opposite side of the range: about 40% of my 87 samples.

Putting rough numbers on the fade shows how much depends on the chop outcome. Say the 07:00 to 08:00 range is 15 pips and the fade enters on the M5 close back inside, with a 5-pip stop above the failed high and a target near the far side about 10 pips away. If the 40% reach target, the 25% re-breaks hit the stop, and the 35% chop gets scratched at zero, expectancy is 0.4 times 10 minus 0.25 times 5, roughly +2.75 pips before costs, and about +1.75 after a pip round trip on every trade. If the chop cases tend to end at the stop instead, it becomes 4 minus 3 minus 1, close to zero. So how the chop trades actually end is probably the most useful column to add to the log, more than the fade versus follow split itself.

Fairman's chop outcome column is the one I added first after reading that post. Once I tagged whether chops scratched, stopped, or drifted to the far side, the fade expectancy stopped looking like a coin flip and started looking like a cost problem on the scratch pile.

Half size after a stop out is my hard filter now. Same direction or fade, I will not match the original size on the next ticket in that range, because the first loss already spent the emotional budget for the morning.

Pre 09:00 is the only window I fade on EURGBP. After that the pair drifts more than it travels, and the far side target rarely pays enough after spread.

I attached one EURGBP M5 chart with the failed break above the morning high, an M5 close back inside, and the fade toward the far side of the range. That is the shape I take at half size before 09:00.
Do you scratch the chop cases on a time stop, or let them run until they hit stop or target? My scratches are where most of the edge leaks.

Fairman's model is easy to extend with a time stop, which is how I handle the chop. If the fade has not reached the middle of the range within 30 minutes I close it, and in my notes those exits average about minus 1 pip after spread rather than zero.

Plugged in, that is ten pips at 40 percent, less five pips at 25 percent, less one pip at 35 percent, or 2.4 pips before costs. Take one pip of cost off each trade and it is about 1.4, so the time stop keeps most of the edge without pretending the chop scratches for free.

Range width is the other filter worth testing. On a 10 pip opening range the far side target shrinks to about 6 pips while the stop above the failed high stays near 5, and the fade stops making sense.

So I skip the fade when the opening hour range is under 12 pips. That removed roughly a quarter of my samples, and the ones left were the better ones.

LondonScalper, did the 87 samples include days when the failed break happened during UK data at 09:30? I would expect those to behave differently from a quiet failure at 08:20.