When I explain correlation risk to newer traders, I keep it to one rule on a sticky note: one USD idea per hour.
The example I always use: long EURUSD, long GBPUSD and long gold at the same time. To a beginner these are three different trades on three charts. In reality, all three are mostly a bet on the dollar falling. If a strong US data release comes out, all three go against you together, and a planned 1R risk becomes 3R.
Teaching notes I use:
- Before opening a second trade, ask what currency the two trades share. If both are short USD, they count as one idea.
- Total risk on one idea cluster stays at 1R or less, however many charts it is spread over.
- If you want two USD trades, split the risk: 0.5R each.
- Crosses like EURGBP are a different idea, but check what they are doing: long EURGBP plus long EURUSD is double exposure to the euro.
I leave correlation coefficients and portfolio maths for later. Mentees who start with spreadsheets of correlation numbers usually miss the simple point, which is that most retail FX trades are a dollar trade underneath.
The sticky note does more good than the maths.
Correlation caps for discretionary scalpers: mentee teaching notes
-
LondonScalper
- Posts: 965
- Joined: Sat Sep 05, 2026 7:54 am
-
Shadow Trader
- Posts: 989
- Joined: Wed Sep 09, 2026 5:19 pm
Re: Correlation caps for discretionary scalpers: mentee teaching notes
A sticky note is the right size for a mentee. One case I'd add is the trade whose stop has already moved to entry, because that's when newer traders usually ask to open the next USD idea.
On paper its risk is zero, so a second 1R trade looks allowed. But on a strong US release both can gap through their stops together, so I let a breakeven trade free up half the cluster, not all of it.
The other habit that helped the traders I've worked with was writing the legs on the ticket. Long EURUSD becomes long EUR, short USD, so the shared currency is visible before the order goes in.
That catches the pairs where the dollar sits on the other side. A long EURUSD and a short USDJPY don't look related on two charts, but both are short USD and belong in the same cluster.
Gold is where I'd soften the wording a little. On risk off days it can rise with the dollar, so I still count it in the USD cluster but tell mentees not to expect it to behave like a currency pair.
How do you handle a mentee who wants to count EURUSD and GBPUSD as partly correlated, say 0.6R each instead of 0.5R? I've had that argument more than once.
On paper its risk is zero, so a second 1R trade looks allowed. But on a strong US release both can gap through their stops together, so I let a breakeven trade free up half the cluster, not all of it.
The other habit that helped the traders I've worked with was writing the legs on the ticket. Long EURUSD becomes long EUR, short USD, so the shared currency is visible before the order goes in.
That catches the pairs where the dollar sits on the other side. A long EURUSD and a short USDJPY don't look related on two charts, but both are short USD and belong in the same cluster.
Gold is where I'd soften the wording a little. On risk off days it can rise with the dollar, so I still count it in the USD cluster but tell mentees not to expect it to behave like a currency pair.
How do you handle a mentee who wants to count EURUSD and GBPUSD as partly correlated, say 0.6R each instead of 0.5R? I've had that argument more than once.
-
Shadow Trader
- Posts: 989
- Joined: Wed Sep 09, 2026 5:19 pm
Re: Correlation caps for discretionary scalpers: mentee teaching notes
LondonScalper, the 0.6R argument has a number behind it, so I answer it with numbers. Two trades of 0.6R each only behave like 1R combined if their correlation is around 0.4, and EURUSD and GBPUSD rarely sit that low.
I checked 5 minute returns on the pair over a few months, split by condition. Quiet Asia came out near 0.55, a normal London session near 0.70, and the 15 minutes after a US release near 0.92.
The chart above is an illustration of those readings with the combined risk of the two trades. At 0.92 the two 0.6R positions behave like roughly 1.18R, while 0.5R each stays just under 1R.
The cap has to hold in the minute when it matters, and that minute is the US number. That is why I keep the 0.5R split even though an average day would allow a bit more.
The one condition that breaks the pattern is a UK data morning, where my reading fell to about 0.35. Cable trades its own story then, but that is exactly when a mentee should be sitting out cable anyway.
Would you let an experienced trader use 0.6R on days with no US data at all? I can see the case, though I would not teach it.
I checked 5 minute returns on the pair over a few months, split by condition. Quiet Asia came out near 0.55, a normal London session near 0.70, and the 15 minutes after a US release near 0.92.
The chart above is an illustration of those readings with the combined risk of the two trades. At 0.92 the two 0.6R positions behave like roughly 1.18R, while 0.5R each stays just under 1R.
The cap has to hold in the minute when it matters, and that minute is the US number. That is why I keep the 0.5R split even though an average day would allow a bit more.
The one condition that breaks the pattern is a UK data morning, where my reading fell to about 0.35. Cable trades its own story then, but that is exactly when a mentee should be sitting out cable anyway.
Would you let an experienced trader use 0.6R on days with no US data at all? I can see the case, though I would not teach it.