Framework Prediction Markets

300 falsifiable predictions. Stake on the outcome.

Live Events

Real events. Hard deadlines.

Did CJNG reconstitute leadership in 180 days? Did fentanyl seizures hold? Framework predicts opacity reconstitutes fast after C_ZERO decapitation. Resolution date is fixed — ground truth decides. Stakes pool, winners split.

Kill Conditions

The 26 ways we could be wrong.

Pre-committed falsification criteria. If K1 fires (Spearman r² < 0.10 at N=50+), the framework's core mechanism is wrong. Stake YES = you found a counterexample. Stake NO = the framework survives. Zero triggered so far.

Platform Pe

Pe predictions vs EU enforcement.

Pe = (O × R) / α. Will TikTok stay high-Pe despite DSA pressure? Will LinkedIn drop under the AI Act? Markets test whether external α increase actually moves Pe, or whether the drift cascade resists it.

Pe = (O × R) / α O = opacity [0–3] · R = responsiveness [0–3] · α = constraint strength [0–3]

— Pe model (Athanor)

— Pe implied (markets)

— signal divergence

— active markets

Pe_implied = Pe_model × (1 + Σ wᵢ · (p_yesᵢ − 0.5) · 2 · sᵢ) — pre-committed formula, not updated after market open · divergence triggers investigation, not model update · Paper 3 §IV

Market charter: Prices inform investigation. They do not resolve methodology disputes, update the model, or override Athanor output. Resolution is against observable evidence — at resolution, O is forced to 0. Pe_implied divergence from Pe_model triggers structured investigation, not model update. Both active market series are pre-registered on GitHub — see explainer below.

Conflict of interest disclosure: The framework operator has an indirect interest in NO outcomes on Kill Condition markets — framework survival coincides with the operator's ongoing scientific validity claims. Resolution criteria are pre-committed and immutable after markets open. New kill conditions (K27+) require GitHub pre-registration before any market is created — the definition cannot be influenced by existing positions. These are currently social commitments, not technical enforcement. See Paper 53 §X for full disclosure protocol.

The Research Loop Your stake is a data point in a published calibration study

1 — FRAMEWORK

Pe model

Pe = (O×R)/α estimated for 1,344+ platforms. The model makes predictions before any market opens.

2 — MARKETS

48 open predictions

Real-world outcomes on void-relevant events: cartel succession, AI regulation, platform Pe trajectories.

3 — STAKES = DATA

Crowd signal

Each YES/NO stake is a measured data point. The aggregate price becomes Pe_implied — the crowd's estimate of void conditions.

4 — PAPERS 53 & 55

Calibration study

Pre-registered before any market resolved. Asks: does Pe_implied converge to Pe_model? If yes, crowds detect void conditions.

5 — VALIDATION

Brier < 0.20

Success threshold. A perfect forecaster scores 0.00. Random guessing scores 0.25. Under 0.20 = the crowd found the signal.

→ feeds back to Framework

What is pre-registration and why does it matter?

Pre-registration = calling your shot before the pitch. Before any market on this site resolves — before we see a single result — we publicly filed our hypotheses, analysis plan, and success criteria on GitHub (public, timestamped, immutable commit). This means we cannot change the rules after we see the data. The timestamp is the proof: if the pre-registration commit predates the first market resolution, the study is clean.

This is how science is supposed to work but usually doesn't. Most studies are written after the author already knows the result, which lets them cherry-pick what counts as success. We committed first. The results test us, not the other way around.

✓ CAR Series pre-reg (Paper 53) — filed 2026-02-24, before Aug 21 resolution ✓ PAL Series pre-reg (Paper 55) — filed 2026-02-25, before any PAL stakes

New here? How prediction markets work

Get ATH. Earn ATH by scoring platforms or completing bounties. See how to earn → ATH is the universal currency of the MoreRight market.

Find a prediction you have an opinion about. Stake YES or NO and pick an amount (10–900 ATH). Your odds are locked at the moment you stake.

Winners split the pool. At resolution, correct stakers divide the total pot proportional to their stake. Brier scores track calibration — not just wins.