Prediction archive audit · May–July 2026
We found the May archive.
The data was never one ledger. It lived across
dated signal snapshots, council run directories, database backups,
and the current prediction store. We separated what was live from
what was pre-launch or replayed later—then scored the live calls again.
Frozen at market close · 30 Jul 2026
Quorum Pro · T+30
77 / 36
right / wrong first signals · 68.14%
Quorum Max · T+30
147 / 84
right / wrong first signals · 63.64%
T+90
Not mature
we will not publish a number before the clock runs
Four May datasets—not one track record
| Cohort | What it proves | Benchmark use |
|---|
| 10–17 May signal snapshots | Pre-launch research state |
Diagnostic only |
| 17–21 May run artifacts | Councils computed briefings before the ledger existed |
Publication cannot be established |
| 21–29 May product ledger | 301 genuine forward predictions; 263 equities after ETF exclusions |
Eligible forward evidence |
| June runs stamped 10 May | Backdate/replay experiments, including cache and web-grounding risks |
Hindcast—never marketing evidence |
Integrity check: the 301 genuine May
rows in today's ledger match the 30 May database backup byte-for-byte
on identity, ticker, tier, verdict, μ, issue time, and source run.
What the live product data supports
Quorum Pro
T+30 first signal: 68.14%
77 right, 36 wrong across decisive stocks. The same-cohort
majority-direction baseline was 61.06%: descriptive lift
+7.08 percentage points.
Quorum Max
T+30 first signal: 63.64%
147 right, 84 wrong. The same-cohort baseline was 59.74%:
descriptive lift +3.90 points. At the
event level, Max recorded 356 right and 205 wrong.
Quorum Max · T+60
32 right / 20 wrong · 61.54%
This did not beat its 65.38%
majority-direction baseline. It is reported, not advertised as a win.
Quorum Ultra-Max
T+30 / T+60 / T+90: not mature
Ultra-Max began live operation in July. Quorum Lite is also
excluded from this paid-tier bulletin while its Ornith Wide
evaluation is still running.
Alphabet: the precise version
“Quorum predicted GOOGL correctly” is true only for a specific
later call. The May upward calls missed: genuine Quorum Max calls on
22 and 27 May were wrong at T+30 and T+60. The 26 June Max
neutral calls were correct at T+30 because Alphabet moved −4.67%,
inside the audit's ±5% flat band.
A timestamped prediction is the claim. A ticker name by
itself is not.
How we scored it
- Equities only; ETFs, leveraged products, and every
backdate-* replay excluded from forward results.
- Split/dividend-adjusted Yahoo bars. Entry is the first tradable
open after the signal; exit is the first close on or after
T+30, T+60, or T+90 calendar days.
- Canonical GMM μ determines direction; stored verdict is only
the fallback. Moves inside ±5% are flat/inconclusive.
- We report both every event and the earliest mature signal per
stock within each tier. The headline uses first signals to reduce
repeated-call dependence.
- Same-cohort majority direction is the baseline. A raw hit rate
below that baseline is not presented as an achievement.
Important: This is a descriptive
paper signal-direction study, not customer returns, a trading strategy,
or investment advice. No trade was executed. Figures omit fees, spreads,
slippage, taxes, position sizing, and portfolio construction. Prediction
events are correlated. Past observed results do not indicate future
results, and AI-generated research may be wrong.