Strategy portfolios: one pool of money

The perfect strategy does not exist. Every strategy has periods when the market does not go its way — and whoever tries to “fix” that with more parameter tuning usually ends up memorizing the noise of one history. The real path to smoother equity leads elsewhere: place strategies side by side whose weak periods do not coincide. The most valuable property of a second strategy is not its return — it is its lack of correlation with the first one. And you can compose at many levels: the same strategy on multiple instruments, several strategies of different types side by side, different asset classes — indices, equities, crypto, commodities — and even different settings of the same logic.

Why non-correlation smooths

The intuition is simple: when strategy A goes through a drawdown, an uncorrelated strategy B has no reason to go through its own in the same week. One's losses meet the other's gains or calm — and the portfolio breathes more steadily than any of its parts. The mathematics agrees: for two strategies with equal risk and zero correlation, the volatility of an equally weighted portfolio drops by roughly 30 % (a factor of √2) — with no discount on expected return. Every additional genuinely independent strategy strengthens the effect.

Correlation here is a number between −1 and +1, computed from returns, not from the similarity of rules: +1 means the equity curves breathe in the same rhythm, 0 means independence, −1 a mirror image. Correlation near zero is plenty for smoothing — negative correlation is a rare luxury. One thing always holds, though: correlation is not a constant. It tends to be low in calm markets and rises toward +1 in a panic — markets fall together. So we size the portfolio for crisis correlation, not the average one; it is the same humility that black swans teach.

This is the right place to smooth equity and reduce drawdowns. Not cosmetics on a single strategy — portfolio composition.

Levels of composition: instrument, universe, type, class

A portfolio is not composed merely of “strategy A + strategy B”. Uncorrelated sources of return can be found at several levels — and the best portfolios combine all of them:

• Multiple instruments, same strategy. A trend logic deployed on ten markets is not one system but ten related sources of return: the rules are the same, the stories of the markets are not. Bitcoin, gold and an equity index do not trend in the same weeks — even this simplest level smooths equity noticeably.
• Whole universes. One step further: we do not deploy a strategy on a hand-picked market but on a defined universe — a basket of instruments meeting the strategy's conditions (liquidity, volatility, available history). Diversification is then a property of the deployment from day one, not an afterthought — and the universe also guards against the strategy standing or falling with one chosen market.
• Multiple strategy types. Trend following, mean reversion, breakout, seasonality: different types profit from different market behaviour, which makes them naturally the least correlated. This is where most of the smoothing comes from.
• Multiple asset classes. Indices, equities, commodity futures, currencies, crypto, prediction markets — each class lives its own macro story. A crypto panic may mean nothing to grain seasonality.
• Settings and combinations of the same logic. The finest level: the same strategy in several settings or timeframes. Siblings stay similar — correlation will be higher than between unrelated strategies — but it spreads the sensitivity to a single “tuned” number we described in overfitting.

And composition does not stop at a pair of strategies — the same logic repeats floor by floor upward. Strategies compose into sub-portfolios: say, trend with mean reversion, longs with shorts, or a profit engine with its hedge. Sub-portfolios compose into segments with clear roles — for example bonds as the calm foundation, equities with dividends and indices as the core, crypto as the declared risk component. The specific combinations are always just examples — the principle is general: compose uncorrelated sources and test the concurrency honestly. At the very top sits the only number that is actually lived: one equity curve of the whole portfolio.

Why you cannot add two backtests

Now the treacherous part. A shortcut suggests itself: I have a backtest of strategy A, a backtest of strategy B — I add the equity curves and I have a “portfolio”. I do not. I have a chart of a portfolio that could never have existed, for three reasons:

1. Capital is one pool. When strategy A opens a position, it locks funds — and at that moment they are not available to strategy B. Standalone backtests never see this collision: each assumes it has the whole pool to itself. Adding two such runs produces a portfolio that, in total, spends more money than it has.
2. Collisions over entries. Both strategies may want to enter at the same moment. Live, the one with funds remaining enters — the other does not get the trade. Which trade is lost is decided by the sequence of events, not arithmetic.
3. Compounding over time. Capital compounds and reallocates continuously: strategy A's January profit changes the position sizes of strategy B in February. Two separate backtests know nothing of this coupling.

Running strategies concurrently is, plainly, the real handling of a single pool of money — and that is how it must be tested.

Time in the market: capital tetris

In the sizing article we showed why to measure the share of time in the market (market exposure). When composing a portfolio, this metric becomes the key: low-exposure strategies complement each other. One strategy's gaps are filled by the other's positions — the same capital serves two independent sources of profit, and its utilization rises without concentrating risk. The ideal pair for a portfolio: uncorrelated signals and complementary exposures.

High utilization by itself is not the goal, though — it becomes the real thing only when the capital works in the best, or at least decently profitable, strategies the portfolio has to offer. And every strategy deserves one more forward-looking question: how much capital will it probably demand over its whole life cycle — from deployment through growing positions to the day it is retired. Strategies are not eternal: they are born, they carry, they end. Plan a strategy's capital appetite all the way to retirement, and concurrency will not surprise you with two strategies asking for the same money at the worst possible moment.

How we test concurrent strategies

Our engine runs multiple strategies at once, over a single account and a single pool of capital — each with its own logic, its own timing and its own risk profile. Collisions are resolved the way an exchange resolves them: funds locked by an open position are locked for everyone, and an order without sufficient free funds is rejected — in testing and live, along the same path. The result is one real portfolio equity curve plus a per-strategy breakdown — and since a portfolio is just another testable unit to the engine, the whole chain applies to it: walk-forward and Monte Carlo over the portfolio as a whole.

That is exactly what our BXF platform is built around: you compose a portfolio at all of the levels above — across strategies, instruments, whole universes and asset classes, from stocks and futures to crypto and prediction markets — and test it as a single whole, on a single equity curve.

What to watch in a portfolio

• Correlation of strategy returns — computed from realized returns (daily or weekly), not from signal similarity. Low correlation is the ticket into the portfolio.
• Exposure overlap in time — how often both want to be in the market at once. High overlap means crowding over the same capital and less benefit from composition.
• Portfolio drawdown vs. component drawdowns — the goal is for the portfolio to fall less than its worst part. If it does not, the strategies are not as independent as they looked.
• Capital utilization — how much of the time the pool actually works, and whether it works in the most profitable strategies available.
• Risk concentration — no single strategy should dominate total risk, or the portfolio is just one bet dressed up as diversification.

Composition > tuning

One last thought worth keeping: when the equity curve is not smooth enough, instinct says tune the strategy. Resist it. Tuning beyond reason manufactures overfitted fragility; composing uncorrelated, individually sober strategies manufactures robust portfolios. You do not buy a smooth equity curve from one perfect strategy — you assemble it from several imperfect ones that complement each other.

Want to see how your strategies would work side by side — over one real pool of capital? Get in touch →