Page 1 of 1

Cointegration and scalping: worth combining?

Posted: Fri Oct 30, 2026 2:55 pm
by FXS London Scalper
I've been reading about pair trading with cointegrated currency pairs (e.g. AUDUSD and NZDUSD) and wondering whether it works on a scalping timeframe.

The idea: if two pairs are cointegrated, the spread between them tends to return to a mean. When the spread stretches, you short the stronger and buy the weaker, and wait for it to come back.

Problems I see for scalping:
  • Costs double. Two positions, two spreads, two commissions.
  • Cointegration is unstable. It holds over months and breaks over weeks, especially when central banks diverge.
  • Intraday the spread is noisy. On M1 or M5, mean reversion can take hours, which isn't scalping anymore.
  • Execution. Two legs have to be entered at the same moment, or the spread you trade isn't the one you saw.
Where it might be useful: as a filter. If AUDUSD rallies and NZDUSD doesn't, maybe the AUD move is local news and more likely to fade.

I tried a simple backtest on M5 with a z-score of the spread. Before costs it looked okay. After costs for two legs, it was basically flat.

Has anyone traded cointegration intraday successfully? Or is it a strategy for longer timeframes only?

Re: Cointegration and scalping: worth combining?

Posted: Fri Oct 30, 2026 3:07 pm
by Amanhasnoname
I spent a couple of months on AUDUSD against NZDUSD on the five-minute chart and came to the same conclusion you did. The z-score signals looked lovely before costs, and after paying both spreads twice the curve was basically a flat line with some ugly steps down whenever the relationship broke after a central bank surprise.

Where I've had more luck is using the pair as a filter, very close to your idea. If Aussie jumps ten pips in a few minutes and Kiwi barely moves, I treat the move as Australia-specific and lean toward fading it on a single leg with a tight stop. When both move together, I leave it alone because that's usually broad dollar flow and it tends to keep going.

Trading one leg instead of two halves the cost and removes the leg risk completely. It's not a true cointegration strategy any more, but it keeps the useful information and drops the part that was killing the returns.