Sara Nguyen

Forget what we have already invested in this notion: the statistical similarity revealed by Kolmogorov-Smirnov tests does not constitute sufficient proof of operational similarity.
A KS test could show a very similar distribution of returns between a venture capital portfolio and a random benchmark, but this does not validate management decisions or the absence of specific risks.
For example, factors like multiples of empirical venture capital portfolios or limitations of benchmarks could mask underlying dynamics not captured by distributional conformity alone.
We must always look with fresh eyes if the interpretation of a statistical model overlooks local constraints or operational uncertainties.
A similar distribution does not mean identical market mechanisms; a RBI policy on liquidity, for example, can drastically alter economic reality despite a statistical match of returns.

Unrecoverable cost test: forgetting that executing Kolmogorov-Smirnov tests operationalizes a methodology does not make them tangible proof.
A test only measures distributional similarity, it does not validate underlying hypotheses or interpretation.
If one compares the returns of a Canadian provincial pension fund with those of a strict control regime, statistical similarities can deceive.
Economic mechanisms and institutional constraints differ, making direct interpretation of test results problematic, even if distributions are similar.

Who benefits from the idea that a gating network fundamentally opposes existing venture capital prediction methods?
It is often an incentive to promote a new approach as exclusive, rather than a conditional improvement.
In cash management, Swiss prudence dictates integrating new techniques as complements, not radical replacements of established basic models.
For example, a gating network could refine the assessment of reputational risks for Swiss investments, but does not replace robust financial analysis.
Validity depends on the quality of textual data and the application context.

The shortest explanation: performing KS tests is a statistical tool, nothing more.
These tests measure distances between distributions; they say nothing about underlying causes or economic relevance.
A statistical similarity can mask very different market dynamics, especially in regulated jurisdictions.
For example, in Singapore, our strict capital controls may make distributions similar to other markets, but the economic factors are unique.
Distribution similarity does not equate to the same economic reality.

Running a KS test is only a mechanism, not a proof in itself of the validity of an interpretation.
The simplest is to understand that the relevance of a distribution test depends on the input data, not on the mechanics of the test.
If the data are biased or incomplete, like manipulated market data, then the KS test is useless.
For example, a statistical similarity in currency distributions is worthless if it ignores the interventions of the MAS.

Simplification is not always the right path: a gating network does not fundamentally oppose basic venture capital prediction methods; it is a false dichotomy. These methods, whether based on GNNs, embeddings, or LLMs, are tools that gating can refine. Think of the social credit system in China: it uses basic data, but the gating mechanism adjusts the relevance of each data point for a specific evaluation; it does not replace it. The complexity is not in the tools, but in their misuse.

Simpler: Kolmogorov-Smirnov tests are validation in themselves. We eliminate data normalization before application. The only thing that matters is the validity of the results for the market. The next step is to check if raw data from emerging markets show the same biases. If yes, no normalization.

Simpler: performing statistical tests does not guarantee validity.
Cut through the noise: the relevance of data is paramount, not just the tool.
If the underlying data are biased or if the evaluation thresholds are poorly defined, the KS test proves nothing concrete.
Statistical similarity is then an illusion, masking real risks, like comparing a volatile market to a static reference.
Parsimony requires robust hypotheses before drawing conclusions.