Risk parity is the principle that each component of a portfolio should be sized so that it contributes the same amount of risk, rather than the same amount of capital. A volatile component gets a small allocation and a calm one a large allocation, so that neither dominates the result.
At trade level
The risk-based position sizing described on its own page is risk parity between trades: each trade is sized so that its stop represents the same fraction of the account, whatever the instrument or the stop distance. A trade in gold with a $40 stop and a trade in EUR/USD with a 40-pip stop carry the same risk in pounds, and neither can decide the month on its own.
At portfolio level
Between strategies, the same principle sizes each so that its typical drawdown, or its volatility, contributes equally to the whole. A strategy whose results swing widely is scaled down; a steady one is scaled up. Without this, a portfolio of three strategies weighted equally by capital is dominated by the most volatile one, and the diversification that was the point of holding three is lost.
The alternative and its problem
Equal capital weighting is simpler and usually wrong. Three strategies with volatilities of 5, 10 and 25 per cent a year, weighted equally, produce a portfolio whose risk is mostly the third strategy. If it has a bad year, the portfolio has a bad year, and the other two barely register. Risk parity would allocate roughly five times as much to the first as to the third, and each would then matter about equally.
The limits
Risk parity relies on estimates of each component's volatility or drawdown, which are themselves uncertain and change over time. Sizing to a thin estimate, a strategy with 40 trades, produces an allocation fitted to noise. And equal risk is not the only sensible target; a component with a better return per unit of risk arguably deserves more. Risk parity is a default that avoids the largest mistake, not a theorem.
A system runs three strategies with historical maximum drawdowns of 8, 16 and 32 per cent when each is run at the same nominal risk. Sized for risk parity, their allocations are roughly in the ratio 4:2:1. The combined drawdown, with the strategies uncorrelated, comes in around 9 per cent, and no single strategy can produce more than about a third of it. Weighted equally instead, the third strategy alone could produce a 20 per cent hole.
A multi-strategy EA that describes its sizing as risk parity, or says each strategy is scaled to its own risk before a single setting scales the whole, is applying this principle, and it is the right one. Ask what the scaling is based on, since drawdown estimates from short records are the weak point, and expect the forward record, rather than the backtest, to be the eventual test of whether the allocations were right.