WATCHTOWER / PREDICTION MARKET INTELLIGENCE

Market Radar

Methodology prediction-risk-method-v1 · October2026 · Open Watchtower

Market prices describe exact contract outcomes. They are not Watchtower forecasts, calibrated confidence, confirmed events, or a blended global-risk probability. A rising ceasefire price and a rising military-action price have different implications.

Coverage and collection

The published catalogue combines two bounded activity-ranked pages from eight Polymarket tags with open events in35 reviewed Kalshi risk-relevant series. Accepted contracts are standard binary questions, currently open and closing within five years. The source view discloses empty series, excluded records, source pagination and caps. This is a curated sample, not an exhaustive census.

Polymarket tags: geopolitics, economy, finance, politics, climate/science, technology, crypto and energy. Up/down micro-markets, mention games, sports and entertainment awards are excluded.
Kalshi series cover Fed policy, inflation, unemployment, GDP/recession, credit/default, oil/trade/Hormuz, government shutdowns, selected geopolitical conditions, health emergencies and technology policy. Series with no accepted open contracts contribute no manufactured coverage.
Catalogue cadence: Polymarket5min, Kalshi10min. Direct books: up to48 diversified contracts on5min cadence. Provider histories: up to64 contracts on30min cadence, requesting the last week at hourly resolution. Source failures and sparse histories remain visible.
The background collector uses fixed reviewed hosts, bounded requests, persisted due-time reservations, provider pacing and pauses. HTTP429 honors Retry-After with at least30min backoff; access-denial responses pause the host for review. Public page/API reads do not initiate external collection or bypass pauses.

Prices, activity and quality

Field	Meaning
Yes price	A two-sided best-bid/ask midpoint when usable, otherwise a separately labelled provider reference price. Midpoint is not an executable quote. No missing-price fallback to50% or zero.
Spread/depth	Ask minus bid in percentage points. Order-book levels are displayed in provider share/contract units. Fees, slippage and participant concentration are not inferred. A wide spread is a market-quality flag, not a risk probability.
24h reference change	Polymarket's reported one-day reference-price change, or Kalshi's last-price field minus its previous-day field when supported. These are percentage points, not investment returns, and not necessarily changes in the displayed midpoint. Unknown/placeholder baselines remain absent.
Activity	Polymarket reported USD volume versus Kalshi contract counts. The two units are not added or treated as equivalent turnover. Combined screening ranks activity within each venue.
Binary payout	Kalshi scalar, combination and nonstandard/floored payouts are excluded. Standard nominal payout isUSD1. Settlement exceptions and early-close provisions still require the original terms.

Clocks and expiry

Observed/retrieved time, native book timestamp, non-trading metadata update, trading close, expected expiration, latest expiration and historical candle period remain separate. A metadata update is not a quote timestamp. Quotes are withheld after30minutes or contract close; catalogue/reference details have a24-hour maximum age. A retained source is labelled and retains its original successful-read clock.

Kalshi book responses do not supply a native quote timestamp; Watchtower labels the observation time instead. Its Yes ask levels are derived from one minus No bids. Polymarket books must match the exact outcome asset and supplied condition, with a valid native timestamp. ProtocolV1 token IDs and V2 position IDs remain strings.

Histories and event ladders

Polymarket historical observations retain source timestamps and bucket widths, including irregular final points. Kalshi uses last-traded hourly candle closes only when the source supplies a valid close and positive period volume. Missing/no-trade periods remain gaps. Charts do not forward-fill missing values, infer trades or manufacture past performance.

Threshold and deadline contracts can overlap. A CPI “above3%” market and an “above4%” market are not mutually exclusive. Event grouping does not establish an exhaustive probability distribution; prices are not normalized or summed to100%. Cross-contract comparisons show separate questions and terms rather than automatic “edge” or arbitrage claims.

Risk interpretation and related context

Theme prompts are authored monitoring guidance. Current prices, moves, liquidity/depth and deadlines are observed market inputs. No automatic polarity is inferred: high Yes prices can represent escalation, stabilization, policy changes, or a narrow measurement threshold.

Related Watchtower reporting and calendar references use bounded keyword/country matching against the current public Geopolitical Outlook snapshot. They are context, not independently verified settlement evidence, causal price explanations or confirmation that an event occurred. Country associations are mentions/curated series scope, not country-wide event probabilities. Source calendar text retains its undeclared timezone.

Workspace behavior

Data and exports are read-only. The page polls collected data while visible and respects replay/account transitions. A watchlist contains up to50 public contract IDs stored locally per browser/account; it does not place wagers or deliver external alerts. Exports preserve selection, source clocks and methods, and stop if the snapshot/freshness selection changes while paging.

The original Polymarket tables remain in a labelled legacy section. Their older authored model/confidence adjustments are uncalibrated heuristics and are excluded from Market Radar's prices and analysis.

Primary documentation

Provider contracts, rules, settlement sources and terms remain authoritative. Watchtower is an independent research interface.