← high signal
public hit-rate ledger

Track record

Sample warning: the live cohort below has only 7 scored predictions. Any rate on a sample this small is statistically meaningless — read it as “direction of travel,” not as a reliable accuracy claim. Wait until n ≥ 10 (per signal type) before trusting it. The number is what it is; we'd rather expose that than dress it up.

Every published market signal scored against subsequent market moves. Read Live first; use Backfill only to calibrate the scoring system.
Hit-rate excludes pushes. Push means the market move was too small or inconclusive.

Use for trust
Live

Forward predictions made before the scoring window closed.

Use for tuning
Backfill

Historical replay. Useful, but not proof of product quality.

Do not overread
Combined

Mixed view for debugging only; it should not be marketed yet.

Live predictions

real forward calls
67%
hit-rate
2
hit
1
miss
0
push

This is the only section that should count for public trust later.

typenhitmisspushhit-rate
guidance_change3200100%
export_restriction10100%
new_product_launch2000
foundry_capacity_announcement1000

Backfill calibration

historical replay
33%
hit-rate
8
hit
16
miss
1
push

Use this to spot weak signal types and scoring bias, not to claim accuracy.

typenhitmisspushhit-rate
partnership18100100%
analyst_revision4100100%
restructuring711050%
earnings_surprise512033%
export_restriction1024133%
guidance_change2028020%
gpu_lead_time_shift1000
foundry_capacity_announcement2000
regulatory_change3000
capex_change_neocloud6000
capex_change_hyperscaler5000
hbm_supply_shift3000
data_center_announcement11000
insider_transaction_cluster20100%
new_product_launch8000
antitrust_action1000
design_win8000
ip_lawsuit1000