Page 1 of 1

EURUSD at London-NY overlap: correlated risk cut rule

Posted: Tue Sep 22, 2026 12:20 pm
by LondonScalper
London–NY overlap on EURUSD is busy enough that correlated risk gets expensive fast.

I used to add a second EUR expression because the first scalp was working. Overlap liquidity feels forgiving until both positions lean the same way into a US data ripple. The cut rule is mechanical: shared budget across EURUSD and the EUR crosses I actually touch in that window.

Overlap specifics
  • Max combined open risk in R, not "two small ones are fine"
  • No adding correlation after the first fill without flattening something else
  • If DXY is driving, treat USD legs as the same story
Overlap is a session for clarity, not for collecting tickets.

What correlated-risk rule survives contact with your London–NY EURUSD book?

I also watch whether GBPUSD or gold is carrying the same USD story. Overlap days can load the book with one macro idea across three symbols. The cut rule is really an honesty rule about how many names one thesis gets.

When the cut rule triggers, I reduce the secondary name first and keep the cleaner invalidation if I keep anything at all.

Re: EURUSD at London-NY overlap: correlated risk cut rule

Posted: Sun Sep 27, 2026 1:40 pm
by PropScalpDesk
Mine is blunt: during the US data hour, any two USD pairs count as one position. Intraday correlation between EURUSD and GBPUSD on a USD-driven move is high enough that two 0.5R tickets behave like one full R, and on a data surprise it's effectively 1:1.

Outside that hour I allow a second USD pair only if its stop is set on its own structure, not just the same distance as the first. When I can't find a separate reason for the stop, the trade doesn't have one either.

Re: EURUSD at London-NY overlap: correlated risk cut rule

Posted: Sat Oct 17, 2026 2:02 am
by Shadow Trader
There is a quick way to put a number on counting two USD pairs as one position. For two tickets of 0.5R each, the combined risk is the square root of 0.25 plus 0.25 plus twice the correlation times 0.25, so at a correlation of 0.9 the pair carries about 0.97R.

Even at 0.6, which is roughly what I see between EURUSD and GBPUSD on a quiet overlap afternoon, the pair still carries about 0.89R. Two small tickets are never really two small tickets in that window.

Gold belongs in the same bucket, just inverted. On a strong US print XAUUSD and EURUSD usually move together against the dollar, so a long in each is one USD short paying two sets of spreads.

The opposite case catches people too. Long EURUSD with short GBPUSD looks hedged on the dollar, but it is really a EURGBP position, and on a data surprise the two legs rarely fill at the prices the hedge assumed.

During the US data hour I count it the simple way, one USD idea per book with gold included. Do you put gold in your USD count, or keep it separate because its stops are so much wider?

Re: EURUSD at London-NY overlap: correlated risk cut rule

Posted: Sun Oct 18, 2026 3:30 am
by Shadow Trader
On which leg to cut first, I now disagree a little with dropping the secondary name by default. When the rule triggers I close whichever ticket has its stop closest to price measured in M5 ATR, since that is the one an ordinary wiggle is most likely to take out.

Take a recent afternoon. Long EURUSD with a 6 pip stop against an M5 ATR of 3 pips sat 2 ATR from its exit, while long GBPUSD with an 11 pip stop and an ATR of 4 sat 2.75 ATR away, so the EURUSD ticket went first even though it was the original trade.

A euro specific headline is the other edge case. An ECB speaker at 14:00 can push EURUSD 15 pips while cable barely moves, so the one position count overstates it and the real exposure is the euro leg.

I keep the blunt version for the US data hour. A payrolls or CPI surprise moves every dollar pair at once, so outside that hour I check the headline source before deciding which leg goes.

PropScalpDesk, how do you count the second pair once the first trade has been moved to breakeven? I treat a breakeven ticket as zero risk, but slippage on a data spike says otherwise.