Compare the Optimization Methods
Behind Portfolio Software.

Q72 optimizes on CDaR, Confidence, and Correlation instead of the Sharpe ratio. The Sharpe ratio measures return per unit of volatility, but it treats upside and downside movements symmetrically — a fund that gains 30% and a fund that loses 30% contribute the same "volatility." CDaR (Conditional Drawdown at Risk) focuses on tail risk: the magnitude of the worst drawdowns, which is what actually forces wealth managers to sell at the bottom.

Classical mean-variance optimizers — including Bloomberg PORT and BlackRock Aladdin — treat all return forecasts with equal confidence. A return estimate for a large-cap equity with ten years of clean data and twelve analyst models is weighted identically to a forecast for a recently-listed small-cap with sparse data and no coverage. The optimizer amplifies noise as eagerly as it amplifies signal, which is why most institutional managers impose ad-hoc constraints (position limits, turnover caps, sector bounds) to make the output usable. The constraints are a workaround for the input problem, not a solution to it.

Q72 Confidence Alpha assigns a proprietary confidence score to each asset's forecast and re-weights the optimization objective accordingly. Assets with high-confidence signals retain full weight. Assets with low-confidence signals have their estimates shrunk toward a confidence-weighted consensus. No asset is excluded; all are weighted by the reliability of what is known about them. The result is allocations that are more stable across rebalancing periods and systematically better behaved out of sample.

Q72 validates every allocation out of sample, not just in sample. In-sample Sharpe ratios — the number most optimization tools put in their summary box — are almost always misleading because the optimizer finds the allocation that looks best on the training data by construction. Q72 uses a rolling walk-forward framework: a portion of historical data is held out entirely, the optimizer runs on the training window, and the resulting allocation is evaluated on the held-out window as if it had been deployed. The summary statistics in every Q72 output come from this out-of-sample distribution.

Q72 shows four engines side by side, not one black-box answer. Markowitz, Risk Parity, Black-Litterman, and the Q72 Confidence Alpha engine all run on the same input data. The quantum-assisted refinement is only used when it demonstrably improves on the best classical solution without increasing volatility. If it doesn't, the classical result stands — and the user sees both.