Lesson 02: Running Several Strategies at Once

Two strategies that lose in different months are worth more than one strategy with a better backtest, because the combined equity curve is smoother and a smoother curve is one you can actually keep trading. That is the entire case for running several. The difficulty is that most traders' "several strategies" turn out to be one strategy with three entry variations.

Diversification that is real

Diversification requires the strategies to lose at different times. Nothing else about them matters — not different indicators, not different names.

Looks diversified	Actually
Breakout on EUR/USD and on GBP/USD	One strategy, two correlated pairs
Trend following on H1 and on H4	One strategy at two speeds
MA crossover and MACD crossover	The same calculation twice
Trend following and mean reversion	Genuinely different — they need opposite conditions
Intraday and multi-week swing	Genuinely different holding periods and drivers

Rows 1 to 3 add trade count and cost without adding diversification, which is strictly worse than running one of them at a larger size. The test is conditions, not construction: if both strategies want the same market to do the same thing, they are one strategy.

Different market state — trending versus ranging is the strongest axis.
Different holding period, which exposes you to different drivers entirely.
Different instruments that are not proxies for each other, which after Track 4's risk-on lesson is a shorter list than it looks.
Different direction bias, if one is long-only and the other short-only.

Measuring correlation between systems

Do not assess this by eye. Two curves that look different can move together where it counts, and the measurement is straightforward.

Take each strategy's result per period — monthly is usual, weekly if you have enough history.
Line the periods up in two columns.
Compute the correlation between the columns. A spreadsheet does this in one formula.

Correlation of monthly returns, from +1 to -1

What each variable means:

+1 — They move together exactly — no diversification at all
0 — Unrelated, which is what you want
-1 — Perfect opposites; combined, they return nothing

Correlation	Verdict
Above 0.7	One strategy. Run the better one at size
0.3 to 0.7	Partial overlap; some benefit, less than it appears
-0.3 to 0.3	Genuinely diversifying — this is the target
Below -0.5	They are cancelling; check you are not hedging yourself

Caution

Correlation measured in calm conditions understates what happens in a crisis. Track 4's risk-off lesson is the mechanism, and the practical consequence is to check the correlation in your worst months specifically — two strategies at 0.1 overall and 0.9 in drawdowns are not diversified where it matters.

How much history you need

Twelve monthly observations is a bare minimum and a weak estimate; twenty-four is more defensible. Below twelve, the correlation figure is noise, and a confident allocation built on it is a confident allocation built on nothing.

Allocating between them

Allocation means splitting the portfolio risk budget from lesson 1 between strategies. Three defensible bases, in increasing order of sophistication and fragility.

Basis	How	Suits
Equal	Same risk budget to each	Two or three strategies, similar records
By confidence	More to the one with more evidence	A long-running system beside a new one
By risk-adjusted return	In proportion to return over max drawdown	Four or more, with 100+ trades each

Equal allocation is the right default, and it stays right longer than people expect. The more sophisticated methods depend on estimates from your own limited history, and an allocation fitted to twenty-four months of data has the same problem Track 6 called curve fitting.

A new strategy starts small — a quarter of a normal allocation — until it has 100 trades live.
Reallocate rarely. Quarterly at most, and only on evidence from the full period.
Never reallocate toward whatever did well last month. That is Track 7's recency bias with a spreadsheet attached.
Keep the portfolio caps binding. Allocation divides the budget; it never raises it.

Note

Running more than three or four strategies is rarely right for a retail trader. Each one needs its own 100-trade sample, its own review, and its own execution discipline — and the practical limit is not the maths but how many processes one person can run properly.

Retirement criteria

Strategies stop working. Markets change, an edge gets crowded, a structural feature disappears. Deciding in advance what would prove that is the same discipline as Track 6's abandon criteria, applied to a live system rather than a test.

Criterion	A workable threshold
Drawdown beyond the historical worst	1.5x the worst in the full record
Expectancy negative over a meaningful sample	100+ trades since the last rule change
Trade frequency collapsed	Under half the expected count for a quarter
The mechanism is gone	Judgement, stated in writing beforehand

Before retiring, check three things

Was it executed as written? Track 7's deviation rate. A strategy you stopped following has not stopped working.
Did costs change? A broker change, a spread widening, a pair becoming less liquid. That is fixable without retiring anything.
Is this inside the expected streak? Track 5's arithmetic. A 40% win-rate system produces runs that feel terminal and are not.

If all three come back clean and the criteria are still breached, retire it — and write down what would make you revisit it. Strategies whose conditions return are worth restarting, and a retired system with a documented reason is much easier to restart than one abandoned in frustration with no record of why.

Key takeaways

Diversification means losing at different times; different indicators on correlated pairs is one strategy with extra costs.
Measure the correlation of monthly returns rather than judging by eye, and check it specifically in your worst months.
Equal allocation is the right default; new strategies start at a quarter size, and reallocation happens quarterly at most.
Set retirement criteria in advance, and first rule out execution drift, changed costs and an ordinary losing streak.

Running Several Strategies at Once

By the end of this lesson

Before this lesson