How we measure the engine

Effective May 26, 2026 (pre-registration lock)

CentoFlow's conviction engine is evaluated against a pre-registered, time-stamped methodology. Everything that could bend a result after the fact — strategy definitions, data-handling rules, and what we are allowed to publish under each outcome — is locked before any result exists.

You will not find performance figures on this page. That is deliberate: numbers appear only when a pre-registered study window completes and its publication criteria — written before the study ran — say they may.

The standard, locked before results

• Pre-registered. Strategy rules, entry/exit thresholds and position sizing were written down and time-stamped before any simulation ran.
• Look-ahead guards. Signals are read only as they were logged on the day; prices only from that day’s bar. No hindsight, no revised data.
• Publication criteria fixed in advance. What we may publish under each outcome was decided before the results existed — the outcome doesn’t get re-litigated.
• Nulls get published. A study that comes back flat or negative is reported, not buried. Falsification criteria are part of the registration.

How a study runs

Each study starts as a written pre-registration: the hypothesis, the exact strategy rules, the data sources, the evaluation window, and the falsification criteria that would count as a negative result. The document is time-stamped and locked before the first simulation runs.

Simulations then replay the engine's own logged signals exactly as they were recorded on the day, priced only with that day's bar. Nothing is recomputed with hindsight, and revised or backfilled data is excluded by construction.

The publishing tree

Before a study's results exist, its pre-registration fixes a decision tree: which outcomes may be published, where, and in what form. A strong result does not loosen the rules, and a weak one does not get quietly shelved — the branch we land on was chosen in advance, and null results are reported like any other outcome.

This page describes the standard. When a pre-registered window completes and its criteria authorize publication, results will appear with their methodology attached — and not before.