FXS Prop Desk, three London EURUSD scalps are not a portfolio, and I learned that when two of my setups shared 11 of their 14 red days in the same quarter. The combined drawdown sat near the sum, which is exactly what a fixed floor punishes.
I now export daily P and L for each setup over at least three months and compute the correlation before I allocate size. Anything above 0.6 gets treated as one strategy with a shared risk budget.
Real diversification for me came from mixing an Asia range fade with a New York momentum idea on gold. Their red days rarely lined up, and the equity curve smoothed enough that the soft stop stopped feeling crowded.
Moving a pair of strategies from roughly 0.7 correlation down toward 0.3 made a bigger difference than adding a third London FX setup ever did. The equity path stopped looking like a single bet dressed as three.
I check overlapping losing days as well as the coefficient before sizing. Is the number alone enough for your decisions, or do you count shared red days too?
FXS Prop Desk, one trap in the daily correlation method came up when I recomputed mine after my last reply. Days where one strategy does not trade at all enter as zeros, and a long run of zeros drags the coefficient toward nothing even when the two setups lose together whenever they both fire.
Over 64 sessions my Asia fade and a London breakout showed 0.25 across all days. On the 31 days when both actually traded, the same pair came out at 0.52, which is a very different risk picture.
The worst days matter more than the average as well. I now look at correlation on the worst tenth of combined days only, because two setups can be loosely related most of the time and still fail together on a CPI morning.
For sizing I use the higher of the two figures. If the shared trading day number crosses 0.5, both setups draw from one risk budget, even if the headline correlation looks comfortable.
That one choice moved my number more than adding another month of data did. Do you compute correlation on all calendar days, or only on days when both strategies were active?
Running several strategies seems to diversify risk, but the benefit depends on how correlated their returns are. Two systems that both profit in trending markets and lose in ranges will suffer together, offering little protection.
Measuring correlation requires daily or hourly return series for each strategy. A correlation matrix shows which pairs move together. Low or negative correlations indicate genuine diversification, and high ones suggest redundancy.
Beyond correlation, tail dependence matters. Strategies may appear uncorrelated in normal times and then lose simultaneously during stress. Reviewing the worst days of each system and checking whether they coincide reveals this risk. Combining strategies that exploit different mechanisms, such as mean reversion in the Asian session and breakout trading at the London open, may provide more durable diversification. Allocating capital in proportion to risk contribution, instead of equally, further improves balance. Periodic recalculation is necessary, since relationships change as markets evolve.
A simple annual review of the correlation table, alongside a look at the worst shared days, keeps the portfolio honest as markets and strategies change.