forex-scalping.com

For a long time I treated a wide spread purely as a cost, something that made a trade more expensive and nothing else. It took me an embarrassingly long time to start reading it as information.

Here is the pattern that changed my view. GBPUSD, quiet Asian range, price creeping toward the high around 06:40 London. Spread on my raw account, normally 0.3 to 0.5 at that hour, suddenly sits at 1.1 to 1.4 as price pokes through the range high by a few pips. Breakout traders pile in. Twenty minutes later price is back inside the range.

When I went back through my notes from the first half of this year, I found 27 Asian range breaks before 07:00 that happened while the spread was at least double its usual level for that hour. Nineteen of them failed and returned inside the range within thirty minutes. Of the breaks that happened with a normal spread, a bit under half failed.

My reading, which may be wrong, is that a break during wide spreads often reflects thin liquidity rather than real interest. Liquidity providers pull back, the price moves easily through a level, and nobody important is actually behind it.

So the rule I use now is simple: a range break that happens while spread is more than double the hour's median doesn't count as a break. I wait for the spread to normalise and see whether price is still outside. More often than not it isn't.

One mechanical point that might explain part of your 19 out of 27. If your chart shows bid prices, then with a 1.2 pip spread the ask is trading above the range high long before the bid gets there.

So any buy stop parked a pip above the high gets triggered before the bid chart has even broken the high. The "breakout" you see on the chart is partly just the spread, and the stops it fills are the fuel for the move back inside.

Rough sketch of the pattern below, with the spread underneath. The break happens while spread is about three times normal, and price is back inside the range well before the 07:00 candle.
I'd also be careful with the sample size. 19 of 27 is about 70 percent, but with 27 events the honest range is roughly 50 to 85 percent. It still looks better than your normal spread breaks, just not by a settled amount yet.

Worth checking whether your 27 cluster on UK data days. A 07:00 release is a good reason for liquidity providers to step back at 06:40, and that would make it a news effect rather than a spread effect.

LondonScalper wrote: Spread on my raw account, normally 0.3 to 0.5 at that hour, suddenly sits at 1.1 to 1.4 as price pokes through the range high by a few pips.

There's a mechanical reason that pattern shows up. MT4 and MT5 charts are drawn from the bid, but buy stops trigger on the ask. When the spread widens from 0.4 to 1.3 pips, the ask can rise most of a pip even if the bid hardly moves, so buy stops sitting just above the range high fire without the bid, and the chart, breaking the level by much. Those traders are then long at a widened ask in a thin market, and when the spread comes back in they're straight under water and plenty get stopped. That's a big part of why breaks during wide spreads fail more often. Sell stops under the low work the same way on the bid side. It also means the spread filter at the moment of the break works as an entry signal, not just a cost check.

Fairman's last line, about the spread filter working as an entry signal, raises the obvious next question. If most wide spread breaks fail within thirty minutes, is fading them back toward the range tradeable?

The problem is the entry cost. On a 15 pip Asia range, a poke 3 pips above the high is about 10 pips from mid, so selling at a 1.3 pip spread gives up over a tenth of the target before any slippage.

My version would wait for two things: spread back under 1.5 times the hour's median, and an M1 close back inside the box. Then short with a stop above the poke high.

That leaves less on the table, but at least the cost is known. Some of your examples won't offer the entry at all, because the return happens while the spread is still wide.

For measuring the hour's median, the bid chart won't help. I run a small script that logs the spread on every tick to a file, and work out the median for each hour and weekday over the last 20 trading days.

Monday mornings need their own numbers. Spreads there are wider for most of Asia, so comparing them against a midweek median makes every Monday break look suspicious.