Sizing for the worst realistic fill instead of the stop distance

[Fairman]

Sizing for the worst realistic fill instead of the stop distance

Position size calculators assume your stop fills exactly where you put it. Enter the stop distance and the money you want to risk, and out comes a lot size. For most trades in a liquid market that's close enough. For the trades that go badly, it often isn't.

A stop is a market order once it triggers. In a fast move, around news, at rollover or in a thin hour, it fills at whatever price is available, and that can be a pip or two beyond the level on EURUSD and a lot more on gold or GBPJPY. If your stop is 4 pips and the typical bad fill adds 1.5 pips, the real loss on those trades is more than a third bigger than planned.

So for each pair and time window I trade, I add an allowance for slippage on the stop and size off the total. Stop distance plus spread plus expected slippage. On EURUSD in London the allowance is small. On gold in the New York afternoon it's much larger, and the size drops to match.

The allowance comes from my own fills, not a guess. Comparing the stop level with the actual fill price on stopped trades over a few weeks gives a number for each pair and session.

It makes every trade a little smaller. In exchange, the losing trades cost what the plan says they cost, and the daily limit means what it's supposed to mean.

[Fairman]

The tricky part is that slippage isn't evenly spread. Most stops fill almost exactly, and a few fill much worse. An average hides that. I look at something like the worse tenth of stopped trades instead, because those are the ones that break a daily limit or turn a normal losing day into a bad one. Sizing off the average slippage still leaves you exposed to the tail, and the tail is the whole reason for doing this in the first place. With a small sample, just take the worst few fills and use something near them until you have more data.

[Fairman]

Time of day changes the allowance more than the pair does. The same pair can need almost no allowance during the London morning and a large one in the minutes after a release or around rollover. A small table with the allowance for each pair and window, kept next to the position size calculator, makes it quick to apply. It sounds like a lot of admin, but in practice most scalpers trade two or three pairs in one or two windows, so the table stays short. Updating it once a month from recent fills is plenty, and it tends to change slowly unless you switch brokers or start trading a new window.

[Fairman]

Guaranteed stops are the other way to deal with this, where a broker offers them. You pay a premium, usually as a wider spread or a fee, and the stop fills exactly at your level whatever happens. For scalping the cost is normally too high to use on every trade. On the odd trade held through a known risk, like a scheduled release you couldn't avoid, it can make sense. Most FX brokers don't offer them at all, so for most of us the allowance in the sizing is the practical answer. Check the terms carefully if you do use one, since some brokers limit how close to price a guaranteed stop can be placed, which may not suit tight scalping stops anyway.

[Fairman]

This also makes broker comparisons more honest. Two brokers with the same spread can have very different slippage on stops, and that difference shows up directly in the allowance. If one broker consistently needs a larger allowance for the same pair and hours, it is costing you more than the spread suggests, even if the headline numbers look identical. Over a few months of records, the allowance becomes one of the clearest cost measures you have, and it's built entirely from your own fills. That's a better basis for choosing where to trade than any marketing page.

[FXS HongKong Scalper]

Good topic. We add an expected slippage term into the size calculation for the times and instruments where we know it is worse. For EURUSD in London we use 0.3 pips, at rollover and in the first minutes of Tokyo we use 1.5 pips, and for gold several dollars around data.

So a 10 pip stop at rollover is sized as if it were 11.5 pips. The difference in lot size is small, but over a few bad fills a month it keeps the realised loss per trade much closer to the planned 1R.