[DEMO] US-China bilateral trade agreement signed by Q3 2026?
Market 54.0% against model 62.0%. Resolves in 96d 11h, data updated 13d ago.
Why this is not actionable
The model can still be informative here, but one or more gates blocks a trade call.
Decision layer
No-trade decision
The model may still be informative, but at least one gate blocks an action-style signal.
Expected value after costs, not raw probability spread.
How much support the model sees across available inputs.
Thin markets can erase apparent edge through spread and slippage.
Resolution ambiguity, timing, and data quality pressure the decision.
weak feature coverage.
Why / why not trade
One decision layer for the market read.
This public box mirrors the internal diagnostic style without exposing execution controls: decision, probability gap, cost-adjusted edge, blocker, and next thing to monitor.
no side selected
62.0% model / 54.0% market
fees, spread, slippage, risk
Category exposure cap already reached in the paper portfolio — concentration risk gate.
Category exposure cap already reached in the paper portfolio — concentration risk gate.
Read this market in three passes
Model 62.0% vs market 54.0%.
Raw disagreement is reduced by fees, spread, slippage, and risk controls.
No trade
Why this read matters
The model may disagree with price, but the gates say the disagreement is not actionable right now.
[DEMO] US-China bilateral trade agreement signed by Q3 2026?
Volume $108,900
Why the engine declines to trade this market
- - Category exposure cap already reached in the paper portfolio — concentration risk gate.
Declining to trade is a feature: most markets are priced fairly within costs, and the risk gates run before any edge is considered.
Market-implied vs model probability
Factor attribution
The model estimates a 8-point higher probability than the market, primarily driven by historical base rate.
| FACTOR | SIGNAL | WEIGHT | LOG-ODDS ΔLog-odds contribution measures how much each factor shifted the model's probability estimate in log-odds space — the mathematically correct way to stack independent evidence. Formula: Δlog-odds = weight × signal. Positive values push the probability up; negative values push it down. Log-odds are converted back to probability via the logistic function at the end. | DIRECTION | DESCRIPTION |
|---|---|---|---|---|---|
| Historical base rate | 25% | — | −1.099 | Bearish | Historical frequency for this kind of event — the prior before any market-specific evidence. |
| Model probability | 62.0% | Prior: 25% · Market: 54.0% | |||
| Confidence (λ)Confidence λ (lambda) controls how much weight to give the model vs. the market. Formula: p_final = λ·p_model + (1−λ)·p_market. λ is derived from data quality, factor agreement, and liquidity. When inputs are weak, the model shrinks toward the market — not toward 50%. | 0.60 | Final: 62.0% = λ·model + (1−λ)·market | |||
Comparable eventshistorical base rate 60.0% - n=10
| Event | Date | Outcome | Prior mkt prob. |
|---|---|---|---|
| US-China tariff pause — 90-day pause announced | 2025-05-12 | US and China agreed to 90-day tariff pause. Rates reduced 115pp each. | 45% |
| US-China tariffs — Trump 2.0 tariff escalation (Liberation Day) | 2025-04-02 | Sweeping tariffs announced. Reciprocal tariffs on most trading partners. | 70% |
| Hamas attacks Israel — October 7 2023 | 2023-10-07 | Attack occurred. 1,200 Israelis killed. 250 taken hostage. | -- |
| Kosovo-Serbia tensions 2022-2023 — No armed conflict | 2023-07-15 | No armed conflict despite major escalations. NATO KFOR maintained peace. | -- |
| Taiwan Strait crisis 2022 — China military drills, no invasion | 2022-09-01 | No invasion. China conducted unprecedented military exercises around Taiwan then stood down. | 12% |
| Russia invades Ukraine — Full-scale invasion | 2022-02-24 | INVADED. Full-scale invasion began. | 50% |
| Armenia-Azerbaijan war 2020 — Nagorno-Karabakh conflict | 2020-11-10 | Full-scale war. 44-day conflict. Azerbaijan recaptured most of Nagorno-Karabakh. 5,000+ killed. | 55% |
| India-China Galwan Valley military clash 2020 | 2020-06-15 | Armed clash occurred. 20 Indian soldiers killed. First fatalities on India-China border since 1975. | 20% |
| US-Iran escalation 2020 — No full-scale war after Soleimani killing | 2020-01-22 | No US-Iran war. Iran retaliated with missile strikes on US bases in Iraq (no US fatalities); standdown followed. | 35% |
| North Korea nuclear standoff 2017 — No armed conflict | 2018-01-15 | No armed conflict. Crisis de-escalated into diplomacy (Singapore Summit 2018). | 30% |
Real historical events from the comparable-events library (showing 10 of 10 matched). The model's base rate is the realized frequency over the full matched set.
Scenario treeEngine template
Node probabilities are conditional on the parent; hover for cumulative path probability. Leaf EV is per $1 YES contract at the current price, before fees (fee-adjusted EVs in the table on the left).
| Path | Path prob. | YES pays | EV (YES, after costs) |
|---|---|---|---|
| Threshold hit in first half of window | 27.9% | $1 | +42.6c |
| Threshold hit in second half | 34.1% | $1 | +42.6c |
| Never reaches threshold in window | 38.0% | $0 | -57.4c |
Root-implied probability 62.0% reconciles with the model's 62.0% (±1pt invariant).
Description
GEOPOLITICS category exposure cap already reached in the paper portfolio.
Resolution criteria (verbatim, with analyzer flags)
analyzed by heuristicResolves YES if the United States and China sign a bilateral trade agreement before September 30, 2026.
- Deadline without timezone A deadline is stated without a timezone — the cutoff moment is undefined.
Resolves Thu, 01 Oct 2026 03:48:52 GMT. The contract pays on these exact criteria, not on the thesis.
Suggested paper position
The engine sizes NO TRADE markets to zero. Sizing never overrides the risk gates.
Paper position only. No real-money execution
Live open-market tracking
Since the first stored model read on 2026-06-09, the market has moved from 54.0% to 54.0%.
This is a directional diagnostic for unresolved markets, not final performance. Resolved outcomes still determine the official live record.
Data quality58/100 - weak
When features are unavailable, the model increases uncertainty and weights the final estimate closer to the market price. Lower data quality does not mean the market is wrong. It means the model is being appropriately humble.
Risk factor breakdownsim
| Inverse liquidity | 34 | |
| Price volatility | 33 | |
| Resolution proximity | 0 | |
| Data quality | 27 | |
| Category base risk | 80 | |
| Resolution ambiguity | 20 | |
| Regulatory exposure | 0 | |
| Portfolio concentration | 0 |
Composite score 44/100, higher = riskier.
Related markets
| Market | Mkt | Delta |
|---|---|---|
| Category context | ||
| [DEMO] Crypto regulatory bill passes Senate vote this week? same event: same venue event + wording overlap | 55.0% | +8pt |
Divergences > 5pt flagged in amber. For cross-venue pricing, see the Scanner.