Diversification is the one improvement available to a trader that costs nothing and asks nothing of the market. It also fails most of the time it is attempted, because it is attempted in the wrong place.
Adding instruments does not diversify. Ten currency pairs that all include the dollar are one bet on the dollar, and five stock indices are one bet on equities; on the day that bet goes wrong, every position goes wrong together. Adding timeframes does not diversify much either, since a breakout on the four-hour chart and on the daily chart are usually the same breakout.
Adding strategies does. A system that buys pullbacks in a rising market and a system that fades weekly extremes in a ranging one make their money at different times, because they need different conditions. Their bad periods rarely coincide, and when one is losing the other is usually somewhere ordinary. Combined, their drawdown is smaller than either alone, while the return is the average of the two. The return per unit of pain has improved without either strategy changing.
The arithmetic is simple. Two uncorrelated strategies with equal 20 per cent drawdowns, run at half size each, produce a combined drawdown closer to 14 per cent than to 20. Three produce something smaller again. The gain is real only while the strategies are genuinely different: different logic, different markets, different timeframes, and, on the record, results that do not rise and fall together.
The test is on the record. If a multi-strategy system can show its components separately and their losing months fall in different places, the diversification exists. If every component lost in the same month, they were one bet with several names.