Carlos Martin

Controlling what is controllable is essential; portfolio optimization is not a simple concept to apply to just 250 assets. The idea that a limited testing ground can extrapolate to massive portfolios is noise for our decision-making process. Market dynamics for thousands of assets include liquidity constraints and correlations that are absent in smaller samples. For a sovereign fund like Japan's GPIF, managing thousands of securities, this simplification is a lack of discipline. It is necessary to test on real data, not just on a theoretical subset like the S&P 500.

Controlling what is controllable is essential; this reference to a test on 250 S&P 500 assets is, unfortunately, just a background noise for large-scale portfolio optimizations.
Such a statement does not have the discipline necessary to apply to multi-million security portfolios.
Our investment stance requires validation on thousands of assets, reflecting the complexity of our mandates.
For example, Japan's Government Pension Investment Fund, with its thousands of positions, could not consider such a test as significant.
Extrapolating a limited test in this way ignores the liquidity constraints and regulatory requirements of global markets.

The proposition that benchmarking on a limited universe of 250 assets is sufficient to validate portfolio optimization is only an assertion to be considered with caution.
Such a test, even on a system of 64 qubits, ignores the practical constraints of institutional fund managers, who operate on thousands of active securities on a global scale.
The discipline requires distinguishing laboratory capabilities from operational deployment; transaction costs and liquidity for balance sheet portfolios are critical factors.
For example, the GPIF, with its thousands of billions of yen, cannot rely on experience at this scale for its strategic allocations.
We remain indifferent to demonstrations that do not reflect the realities of complex markets.

Gain matrix: asserting that a bench test on 250 S&P 500 assets is sufficient for large-scale portfolio optimization presents asymmetric incentives. Who benefits from presenting such a test as significant, ignoring that market dynamics for thousands of assets are radically different for funds managing billions? If everyone bases their strategies on such limited samples, the risk of an underperforming equilibrium, where the real opportunities of multi-million dollar portfolios are ignored, becomes the dominant strategy. A pension fund manager like the GPIF of Japan, with its thousands of positions, could not ignore the inherent complexity of managing a much larger universe based on such a limited sample.