War Markets | Interactive Deterrence

Price of a $1 “Yes” share, daily

Priced vs happened

Each dot is a resolved contract, placed at its price a week before the outcome was decided (the day it settled at Yes, or the deadline for a No). Click a dot to open it.

Betting on war

Jonathan Walberg’s Betting on War project starts from a gap in the theory of crisis bargaining. States show resolve with costly signals, and the literature since Fearon (1997) has two kinds: tying hands, where a leader stakes her reputation in public, and sinking costs, where a state pays up front by mobilizing. Both assume the receiver knows who sent the signal.

Prediction markets add a third channel. A trader can buy a large position in “Will the US strike Iran by June 30?” and move its price. The position costs real money, the price is public, and news outlets and officials quote it. The buyer stays anonymous. An observer who sees the price jump has to ask where it came from: someone who knows what the sender will do trading on it, or the sender trading to be believed. Walberg’s theory paper, “Anonymous Costly Signals: A Third Mechanism in Crisis Bargaining,” models this as a signaling game in which the receiver’s problem shifts from reading a signal to judging its source. Everything turns on one parameter, λ, the share of anonymous price movements that come from informed trading rather than from the sender’s own bet or ordinary noise.

The companion methods paper, “Estimating the Deception Parameter,” asks whether λ can be measured. No single price move can be attributed, but the share of informed moves is a mixing weight, and contract resolutions make it estimable: in contracts that resolved without war the unresolved side does not bet, so movements there separate informed trading from noise.

Working-paper estimates of λ by theater (primary sample)
Theater	λ̂	95% interval	Reading
Iran/Israel	0.244	[0.147, 0.367]	At the upper edge of what a market with no informed traders produces (5% of such simulations reach it). Weak evidence of informed flow.
Russia/Ukraine	0.053	[0.012, 0.139]	Inside the no-information benchmark: no signal.
Taiwan	0.530	[0.102, 1.000]	The data fail a premise of the method; the estimate carries no information.

Source: Jonathan Walberg, “Estimating the Deception Parameter: A Finite-Mixture Approach to Source Attribution in Thin Prediction Markets,” qualifying paper draft, University of Virginia, 29 September 2026, Table 2 (profile-likelihood intervals). These are working-paper estimates, not peer reviewed, and may change. In the author’s words, for a receiver in Taipei or Washington “the evidence does not yet justify reading a sharp overnight move in an invasion contract as leaked knowledge of Beijing’s intentions; for the Middle East contracts, it gives modest grounds to weight such moves as more than noise.”

This page does not estimate λ. It shows the raw material: what the archived contracts priced, day by day, and how each one resolved.

How to read it

Lines are daily “Yes” prices in cents. Contracts on one question with different deadlines form a ladder; the dashed line in each contract’s color is its deadline.

Filled dots at 100¢ or 0¢ mark the day prices settled at the outcome and stayed there.

Dots along the top are dated events from Walberg’s event lists. Hover or tap the chart to read them, or focus it and use the arrow keys.

The panel shows the focused contract’s outcome, its price a week before, its Brier score and its resolution rule.

Gaps are days with no archived trade. The collector started on 5 August 2026 and only the last trades of earlier contracts survive, so some lines are short.

This page is for research and teaching. It is not trading advice and makes no prediction about any open contract.

Method

Prices and trades come from Walberg’s conflict-market-data collector, which archives Polymarket’s geopolitical contracts hourly because Polymarket stops serving price history once a contract resolves. Daily price: the last hourly bar of the UTC day where the archive has hourly bars; otherwise the size-weighted median of that day’s fills, with a “No” fill at price q counted as a “Yes” price of 1 − q. Days with no trades stay empty.

Outcome of a closed contract: the final-price rule used in the working paper. If the last 20 archived trades average 95¢ or more for Yes, it resolved Yes; 5¢ or less, No. This is not Polymarket’s official resolution notice. A check against Polymarket’s public API on 6 October 2026 found all 23 resolved contracts resolved the same way.

“Settled” day: the first day from which every later daily price stays at 95¢ or more (Yes) or 5¢ or less (No). “A week before” is the last price on or before seven days before the outcome was decided: the settled day for a Yes, the deadline for a No. If the archive starts later, the first archived price is used and marked with an asterisk.

Brier score: (price − outcome)², with price as a share of $1 and outcome 1 or 0. Always quoting 50¢ scores 0.25. The panel’s “Mean Brier” averages the score of every archived daily price before the day the outcome was decided, for the focused contract. The scorecard’s “Mean Brier score a week out” scores one price per contract, the price a week before, and averages across the resolved contracts.

Deadlines are read from each contract’s question. Market IDs are Polymarket’s. Resolution rules are quoted from archived copies of each contract’s page on polymarket.com, linked in the panel.

Events are the tier-1 dated events from Walberg’s Betting on War replication files (2026 only) for the Iran/Israel and Russia/Ukraine theaters, each with the one source URL the list records. Two fact-checks in September 2026 opened every source and rewrote the descriptions that went beyond what their sources say, moved one event to the date its source gives (19 February, not 20 February) and dropped one event whose source did not support it.

Scripts: scripts/fetch_rules.py retrieves the rules; scripts/build.py reads the collector database and writes data/markets.js. No value on this page is notional.

Sources

Market data: Polymarket, via github.com/jonawalb/conflict-market-data (hourly bars and trade tape; the archive for these 28 contracts was topped up from Polymarket’s public API on 7 October 2026, UTC).

Resolution rules: archived Polymarket pages on the Internet Archive Wayback Machine, one link per contract in the panel.

Events: Walberg, Betting on War replication data (events_iran_israel_v5.csv, events_russia_ukraine_v5.csv), with the source listed for each event in the chart tooltip data.

Jonathan Walberg, “Anonymous Costly Signals: A Third Mechanism in Crisis Bargaining,” working paper.

Jonathan Walberg, “Estimating the Deception Parameter: A Finite-Mixture Approach to Source Attribution in Thin Prediction Markets,” qualifying paper draft, University of Virginia, September 2026.

James D. Fearon, “Signaling Foreign Policy Interests: Tying Hands versus Sinking Costs,” Journal of Conflict Resolution 41, no. 1 (1997): 68–90.

Betting on war

How to read it

Method

Sources

Events shown