Every quant stock rating platform shows you its best numbers. Backtests with triple-digit returns. Win rates above 70%. Charts that only go up. What none of them show you is the audit that did not go their way.
We audited our own Q-Score — the composite quant stock rating at the heart of StoQuant — and it got a failing grade. The rank correlation between Q-Score and forward excess returns is approximately zero over 178 scoring dates. The top-ranked bucket beat the Russell 2000 only 46.1% of the time across nine independent 90-day windows. By the standards we set for ourselves, that is a Grade F.
Most companies would bury that result. We put it at the top of our accuracy page. Here is why.
A quant stock rating is a number that tries to summarize many signals into one score. Usually it blends valuation, momentum, quality, sentiment, and sometimes macro or alternative data. The promise is that a single score can rank stocks from best to worst.
The reality is harder. A composite score is only as good as:
If any of these steps is weak, the rating is not wrong — it is just untrustworthy. And without transparent validation, you cannot tell the difference.
The Q-Score combines 149 signals across nine dimensions: valuation, quality, momentum, sentiment, insider activity, analyst activity, macro sensitivity, risk, and machine-learning alpha. The inputs include fundamentals, price data, insider filings, analyst actions, news sentiment, and macro regimes.
The score is regenerated nightly across a universe of 3,500+ US stocks. Every score is stored append-only, meaning we never overwrite a past rating to make the history look better. That audit trail is what made the F grade possible.
You can read the full construction details in our Q-Score methodology guide and the broader methodology page.
In August 2026 we re-examined how we were measuring Q-Score performance. We found two problems in the previous version of the proof page:
We withdrew those figures on 2026-08-16 and replaced them with a corrected protocol:
The corrected result: rank IC ≈ -0.001 over 178 scoring dates, with a 46.1% beat rate across nine independent windows. In plain English, the Q-Score currently shows no measurable cross-sectional edge.
The obvious move was to hide the result, rebrand the score, or quietly deprioritize the proof page. We chose the opposite for three reasons.
First, it is the only honest thing to do. If a scoring model fails its own audit and the company hides it, every other number on the site becomes suspect. The F grade protects the rest of the product.
Second, transparency is a moat. In a category full of black-box scores and undisclosed backtests, a public report card nobody else is willing to publish is hard to copy.
Third, the product still has value even if the composite score does not predict returns. The hidden-gem screener, the smart-money radar, the 3,500-stock coverage universe, the API/MCP access — these are useful independent of the Q-Score's grade. Investors need data and tools, not just predictions.
A Grade F composite score does not mean StoQuant is useless. It means the current version of the all-in-one Q-Score should not be used as a stand-alone stock picker.
Here is what we recommend instead:
This is the same standard we would apply to any external rating. If a score cannot show its report card, we do not trust it. Now our own score meets that standard, even when the result is embarrassing.
Most quant ratings platforms fail on at least one of these dimensions:
We built StoQuant's proof page to avoid every one of these traps. The methodology is public. The history is append-only. The benchmark is the Russell 2000. The windows are independent. And the current grade is F.
If you are evaluating quant stock ratings, here is a simple checklist:
If the answer to any of these is no, the rating is marketing, not measurement.
StoQuant passes the first four. The fifth is the hard one, and we are choosing to pass it too — starting with this post and the live proof page.
The F grade is not the end of the story. We continue to retrain the Q-Score, add independent windows, and publish the updated grade. If the signal improves, the proof page will show it. If it does not, the proof page will show that too.
In the meantime, the rest of the platform keeps working. The Graham screener still finds under-covered value names. The smart-money radar still tracks insider clusters and analyst drift. The API/MCP still lets agents and developers pull structured data.
We are not asking you to trust our score. We are asking you to trust our measurement.
Performance figures refer to benchmark-relative outcomes vs the Russell 2000 over date-matched 90-day forward windows. Overlapping observations are thinned to independent windows before statistics are computed. Figures are signal-level and gross of execution costs unless stated. An earlier version of our accuracy page used mis-weighted benchmarks and fixed score bands labeled as quintiles; those figures were withdrawn on 2026-08-16 and replaced with the corrected measurement. See /proof for the live numbers.
The rank correlation (IC) between Q-Score and forward excess returns is approximately zero (-0.001) over 178 scoring dates, and the top-ranked bucket beat the Russell 2000 only 46.1% of the time across nine independent 90-day windows. By the standards we set for ourselves, that is a Grade F — no measurable cross-sectional edge in the current model.
Two things: the benchmark was mis-weighted, which flattered relative returns, and fixed score bands were labeled as quintiles even though the buckets did not always contain 20% of the universe. Those figures were withdrawn on 2026-08-16 and replaced with a corrected protocol: Russell 2000 buy-and-hold matched by date, 90-day forward windows, overlapping observations thinned to independent windows, and mean/median/beat-rate reported instead of a single pooled average.
No. A Grade F composite score should not be your final answer. Use the Graham screener for deterministic value ideas, the smart-money radar for insider clusters and analyst drift, and the Q-Score as one input among many. The proof page updates as more independent windows accumulate.
Daily Q-Score changes, insider buy clusters, and ML forecast updates for the tickers you follow. Email only, no account needed. See a sample digest.