Cost of holding a scalp 3 minutes vs 30 seconds on XAUUSD

[LondonScalper]

Cost of holding a gold scalp three minutes versus thirty seconds.

On XAUUSD the spread+commission ticket is paid immediately. Extra minutes only help if expectancy improves. My log said many three-minute holds were just thirty-second ideas I refused to close -- giving back edge while gold wiggled through my stop.

What I compare
Same setup family, split by hold bucket: under 60s, 1-3 min, over 3 min. After costs, the long bucket was often worse for my scalp definition. So I tightened time stops on "liquidity grab" tickets and kept longer holds only for a separate continuation tag with different size.

Mixing styles on one chart without tags made gold look random. Separating them made costs honest.

If you scalp gold, have you bucketed results by hold time -- and did it change your exits?

Practical change: default time stop on XAU scalps is short unless the ticket is tagged continuation with a pre-written hold plan. Untagged "see how it goes" holds were the leak. Gold will always offer another tick; your costs already charged you for the ones you took.

[PropScalpDesk]

I've bucketed by hold time too, and there's a trap in reading the buckets. Hold time is partly an outcome rather than a choice. Trades that work quickly hit the target inside a minute and land in the short bucket. Trades that don't work drift around and end up in the long bucket. So "over 3 minutes was worse" is partly guaranteed by how the buckets fill, even if the extra time itself cost nothing.

The cleaner test is to apply each exit rule to every trade from the moment of entry. Take all your gold scalps, then use M1 or tick data to see what each would have made with a 60-second time stop, a 3-minute one and no time stop. Then you're comparing rules on the same set of entries instead of comparing different trades.

When I did that, the short time stop still came out ahead after costs, but by less than the raw buckets suggested. So your conclusion probably holds; the size of the leak may be smaller than it looks.

[Shadow Trader]

Replaying every entry under each rule is what I did with my own gold scalps, and it hit one complication. With a 60 second time stop, one M1 bar often contains both stop and target, and the bar cannot say which came first.

I counted those ambiguous bars as losses, which is the conservative choice. It affected 14 of 180 trades, so tick data is worth the effort for very short holds.

img_12415_r1.png (71.12 KiB) Viewed 2 times

The chart puts both views side by side, in dollars per ounce after about 30 cents of costs. The raw buckets made short holds look far better, +1.10 under 60 seconds against minus 0.65 over 3 minutes, while the same 180 entries replayed under each rule gave +0.42 for the 60 second stop, +0.31 for 3 minutes and +0.05 with no time stop.

So the conclusion survived, but the gap shrank from 1.75 to 0.37. That is close to what PropScalpDesk found.

One more split helped. Entries in the first 30 minutes of London did slightly better with the 3 minute stop, while Asian and late New York entries favoured 60 seconds.

Has anyone tried a time stop scaled to M1 ATR instead of fixed seconds? That seems the obvious next test.

[Shadow Trader]

I ran the ATR scaled time stop I asked about on the same 180 gold entries, and it did worse. The stop was 60 seconds at median M1 ATR, longer in proportion when ATR was lower and shorter when it was higher, with a floor of 30 seconds and a cap of 3 minutes.

img_12415_r2.png (67.8 KiB) Not viewed yet

The chart splits it by session in dollars per ounce after costs. Overall it made +0.36 against +0.42 for the fixed 60 seconds, and the damage sat almost entirely in Asia, down from +0.30 to +0.12, and the London open, down from +0.45 to +0.33.

Looking back, my earlier session split already pointed that way. Quiet Asian trades did better with short holds and fast London open trades with longer ones, which is the reverse of what scaling by volatility does.

So volatility does not seem to decide how long a gold scalp deserves. My best guess is that follow through matters more, and the London open simply has more of it.

The tempting fix is to scale the other way round, but with 40 Asian trades that would be fitting the rule to the sample. Has anyone tested a time stop that resets each time price makes a new extreme in the trade's favour?