← All lessons

Diversification across markets

See why trend systems trade many markets · 8 min

ProGoal: See why trend systems trade many marketsBefore this: Moving-average systemsFurther reading: Clenow, Following the Trend
One trend system, four generated markets
0.88×1.51×2.14×01451
  • Trend system on market 1
  • Trend system on market 2
  • Trend system on market 3
  • Trend system on market 4

Same rule, same settings, same risk on each market. Their good and bad years rarely line up.

Predict first

A trend system has a Sharpe ratio of 0.4 on each of 9 markets, and the systems’ returns are only weakly correlated (0.1). Trading all 9 with equal risk, the portfolio’s Sharpe ratio is about…

Trends are rare in any one market and come at times nobody can pick. A trend follower therefore trades many markets: whichever trends this year carries the rest. Diversification does not raise the average return of a system; it lowers the volatility of the total, because losses on one market are offset by gains on others.

volatility of n equal-risk systems = σ × √(1 ÷ n + (n − 1) ÷ n × ρ)
Sharpe of the whole ≈ Sharpe of one × √(n ÷ (1 + (n − 1) ρ))

The formula shows the limit too. With ρ = 0.3 the volatility can never fall below √0.3 ≈ 55% of one system's, however many markets you add. Counting markets is not enough: what matters is how many independent bets they are. And correlations are not fixed: in a shared selloff they jump (lesson D4-1), so the diversification is smallest exactly when it would help most.

Try it: add markets and watch drawdowns shrink
0.92×1.37×1.82×01451
  • 1 markets, equal risk
  • Market 1 alone
Sharpe, one market0.62
Sharpe, 1 market0.62
Max drawdown, one market-24.5%
Max drawdown, 1 market-24.5%
Avg correlation0.09
0.540.640.7409
  • Sharpe ratio by number of markets

Each market gets the same risk budget (scaled to 15% volatility, then split evenly). The lower line is indexed from 0 (one market) to 9 (ten markets).

Generated markets, not real data. They show how a rule depends on the market it meets; they do not predict how real markets behave.

Worked calculation

Risk of four systems traded together

Each system: volatility, return
15% a year, 6% a year (Sharpe 0.4)
Four systems, correlation 0.2
n = 4, ρ = 0.2
Volatility of the equal-weight total
15% × √(0.25 + 0.75 × 0.2) = 15% × 0.632 = 9.5%
Return of the total
Still 6%: the average of the four
Sharpe of the total
6% ÷ 9.5% = 0.63, up from 0.4
Check yourself
3 questions

Two markets have daily returns correlated 0.9. What do you expect of the same trend system run on each?

Your trend system trades three stock indices. Which candidate adds the most diversification? Its trend system’s correlation with your current portfolio is: A 0.85, B 0.40, C 0.05.

One trend system had a worst drawdown of 30%. You now trade five markets with equal risk and low correlation between them, keeping total volatility the same. The worst drawdown of the five-market version is most likely…

Missed questions come back for review after one day, then three, then a week.

Practice

Strategy 1 has 20% volatility, strategy 2 has 20%, and their returns have a correlation of 1. You put 50% of capital in the first and 50% in the second. What is the blend's volatility (in %)?

Hint: σ² = w₁²σ₁² + w₂²σ₂² + 2·w₁·w₂·ρ·σ₁·σ₂, then take the square root.

More in Drills →

Closing round

Three years of a fresh generated world are on screen; the next two years do not exist yet. Tune the rule on what you can see, commit to two predictions, then generate the future.

0.97×1.02×1.07×0625
  • Your rule so far
Sharpe before split-0.11
Return before split-1.7%
Max drawdown before-9.5%

Over the next two years, after costs, this rule will…

Will the deepest drawdown after the split be worse than before it?

Generated markets, not real data. They show how a rule depends on the market it meets; they do not predict how real markets behave.

Want to feel it candle by candle? Try the mystery market: which regime is it? →