How we report
Methodology: how W.E.T. reports prediction-market odds
W.E.T. (World Event Trading) is the sentiment intelligence layer for prediction markets — we report what the market says about real-world events: the implied probability, the Kalshi-vs-Polymarket split where it exists, and the traded volume. Every figure is a dated snapshot of live market prices. It is data, not a forecast, and not advice. This page documents exactly how those numbers are sourced, computed, and dated so anyone — journalist, researcher, or reader — can judge and cite them. It also documents the W.E.T. Score — our 0–100 measure of how much verifiable signal sits behind a market's number — and the calibration program being built on top of it.
What we publish
For each tracked event we publish the market's implied probability for the leading outcome, the outcome ladder (each priced side), the combined traded volume, and — when the same market trades on both venues — the cross-venue divergence. Each page is a durable, dated snapshot with an explicit “as of” date. Since July 2026 each event page also carries a W.E.T. Score — a 0–100 signal-quality measure, documented in full below.
Where the data comes from
Prices are read from the public APIs of two prediction-market venues: Kalshi, a CFTC-regulated US event exchange, and Polymarket, a crypto-native prediction market. We read only publicly available market data (prices, volume, resolution metadata) — no private or personal data. Neither venue pays for or influences our coverage; where W.E.T. earns referral commissions on sign-ups, we disclose it.
How implied probability is computed
A prediction-market contract that pays $1 if an outcome happens trades between $0 and $1. Its price is a direct estimate of probability: a contract at 63¢ implies a 63% chance the market resolves YES. We read the venue's live YES price for each outcome and express it as a whole-number percentage. The reported favorite is the highest-probability who-wins (moneyline) or main-ladder outcome — never a side prop. Events with no genuine who-wins market, or whose leading outcome is effectively settled (≥96% or ≤4%), are not published.
Cross-venue divergence
When the same event trades on both Kalshi and Polymarket, we show the gap between their implied probabilities on the favored outcome. A divergence can reflect different liquidity, fee structures, or — importantly — subtly different resolution rules, so it is a signal to read both rulebooks, not an arbitrage guarantee. In practice the two venues list mostly different events; where they do not overlap, a page shows a single venue and says so.
Freshness & update cadence
Snapshots are regenerated daily from live market data. To keep freshness signals honest, a page's “last updated” date advances only when the implied probability actually moves by at least one point — a daily rebuild that re-stamps an unchanged number would be a false freshness signal. When a number moves, we show the delta (for example, 61% → 63%) and the date it last changed. The original publish date is preserved; it is not reset on every refresh.
Score methodology
The W.E.T. Score
The W.E.T. Score is a 0–100 measure of how much verifiable signal sits behind a market's headline number. It answers one question: how seriously should you take this price? A 63% implied probability backed by a million dollars of traded volume, two agreeing venues, and a week of stable daily history is a different object from a 63% resting on $4k and one thin order book — the score is how W.E.T. tells them apart, computed identically for every market on the site.
It is built from five observable components — nothing subjective, nothing editorial. Every input already appears on the page it scores, so any reader can audit any score by hand.
| Component | Max pts | Data source | How it's computed |
|---|---|---|---|
| Liquidity | 30 | Combined traded volume across the event's markets, both venues | log₁₀-scaled: $1k → 0 pts · $10k → 10 · $100k → 20 · $1M+ → 30. Ten points per decade — a market must trade 10× more to earn each next block. |
| News corroboration | 20 | Wire headlines matched to the event from W.E.T.'s live news feed | Count of matched headlines: 0 → 0 pts · 1 → 10 · 2 → 16 · 3+ → 20. |
| Cross-venue agreement | 20 | Kalshi and Polymarket implied probabilities, aligned to the same outcome | Both venues priced: 20 − 2 × gap in points, floored at 4, capped at 20. Single venue: 8 — neutral, because agreement is unknowable, not absent. |
| Market coverage | 15 | Number of distinct priced markets on the event | 1.5 pts per market, capped at 15 (reached at 10 markets). |
| Track-record depth | 15 | Days of daily odds history W.E.T. has captured | 1.5 pts per day, capped at 15 (reached at 10 days). |
| Range | Tier | Read it as |
|---|---|---|
| 85–100 | Prime Signal | Deep, corroborated, multi-venue. The strongest reads on the site. |
| 70–84 | Strong Signal | Well-traded with meaningful corroboration; minor gaps. |
| 55–69 | Solid Signal | A legitimate read — check which component is missing before leaning on it. |
| 40–54 | Developing Signal | Early or thin; treat the number as provisional. |
| 0–39 | Thin Signal | A price exists, but little verifiable weight sits behind it yet. |
Alongside the score, each market can carry flags — binary conditions with fixed trigger thresholds. Flags never change the score; they annotate it. They are listed here in priority order — the same order the site ranks them when space is limited.
| Flag | Trigger | Read it as |
|---|---|---|
| Thin market | Total volume < $10,000 | A handful of trades can move this number. Treat with caution. |
| Deep liquidity | Total volume ≥ $250,000 | Enough money at stake that this price is expensive to be wrong about. |
| Divergence | Cross-venue gap ≥ 7 pts | A checkable arbitrage gap — run it through the arbitrage calculator. |
| Venue consensus | Both venues priced; gap ≤ 3 pts | Two independent trader pools converged on the same probability. |
| News-backed | ≥ 1 matched wire headline | Independent reporting exists for what the market is pricing. |
| Big mover | Repriced ≥ 5 pts across the tracked movement window | The market repriced hard; look for the catalyst on the news wire. |
| Early read | < 3 days of captured history | Tracking started recently — the depth component will grow if the market stays listed. |
What the score is not
It is not a prediction of the outcome. It is not a rating of whether the favorite wins, and not trading advice — a Thin Signal market can still resolve YES, and a Prime Signal market can still resolve NO. The score measures signal quality — how much verifiable weight sits behind a price — never outcome likelihood. 63% at Prime Signal and 63% at Thin Signal are the same probability with different evidentiary weight behind them.
News matching is lexical, not semantic. Headlines are matched to markets by text overlap with the event title. That is fast, auditable, and disclosed — and it can miss a relevant story phrased differently, or match a merely similar one. We call it corroboration, not verification.
The weights are choices. The five components and their point allocations are W.E.T.'s own design — reasonable, fully disclosed above, and applied identically to every market, but not the only defensible weighting.
It is not investment advice. Not a trade rating, not a buy/sell signal, and no substitute for reading the venue's resolution rules yourself. See the full disclaimer.
Why prices know things
The crowd-sourced signal
A prediction-market price is not a poll. Nobody is asked for an opinion — traders stake money on a claim, and the price is simply where buyers and sellers stop disagreeing. That mechanism has two properties no survey can replicate:
Skin in the game
Being wrong costs money, so cheap talk gets filtered out. Anyone holding information the price does not yet reflect is literally paid to trade against it — correcting the price is the profitable move, which is why dispersed information ends up in the number.
Continuous repricing
A market never files a final answer and goes home. Every headline, poll, injury report, and rumor gets priced within minutes, around the clock — the number you see has already absorbed whatever the crowd knows right now.
W.E.T.'s edge is reading two crowds at once. Kalshi (a CFTC-regulated, dollar-funded US exchange) and Polymarket (crypto-native, global) draw different trader pools under different constraints. For every event both venues price, we align their books to the same outcome — even when one venue lists a combined “A vs B” market and the other prices each side separately — and publish the split. Agreement within a point or two is two independent crowds converging on the same probability — the strongest form of the signal, scoring at or near the full 20 points in the score's cross-venue component. Divergence is information too — different resolution rules, different liquidity, or one crowd repricing first — which is why a gap of 7+ points trips the divergence flag and links out to the arbitrage calculator rather than being averaged away.
Once a day, W.E.T. captures one implied-probability point per tracked market — the favorite's price, dated. Capture began July 14, 2026. Those daily points are the odds paths on every event page, and once a market settles they freeze into the archive's permanent record: what the crowd believed, day by day, before anyone knew the answer. Today the history feeds the score's track-record component; as it accumulates, it feeds the calibration program below.
What ships next
Simulations & calibration (roadmap)
Everything above this line is live today. This section is a commitment, not a feature list: as the daily odds corpus matures, W.E.T. will publish the analysis that corpus makes possible — and nothing before the data supports it. Capture began July 14, 2026. A calibration claim built on a few weeks of settlements would be noise dressed as research, so these ship on the data's schedule, not a marketing calendar.
01 · Accuracy
Calibration curves
Were the market's 70% calls right about 70% of the time? Every settled market pairs a final pre-close probability with a known outcome. Binned across enough settlements, that pairing is the market's honesty test — the calibration curve — published by category and by venue.
Requires: Hundreds of settled markets per bin
02 · Scoring
Brier-score tracking
One number for forecast accuracy: the mean squared error between a day's implied probability and the eventual 0-or-1 outcome, tracked at fixed horizons (30 days out, 7 days out, final read) by category and venue. Lower is better; an always-50% coin-flipper scores 0.25.
Requires: Settled outcomes across multiple capture horizons
03 · Simulation
Monte Carlo settlement paths
Observed daily paths give an empirical distribution of how markets in each category actually move. Resampling those real paths lets us simulate thousands of routes to settlement and publish ranges instead of point estimates — built strictly on W.E.T.'s own captured paths, never on assumed volatility.
Requires: Deep per-category path history
No fabricated results, no backfilled precision, no survivorship edits: until each item's data threshold is met, this page will say “not yet” rather than show a number we cannot defend. When the analyses ship, they will be reproducible from the archive's published data.
What these numbers are NOT
They are not a prediction or forecast by W.E.T., not betting or investment advice, and not a recommendation to trade. Implied probability reflects what traders are collectively paying at a moment in time — it moves continuously and can be wrong. Prediction-market trading involves risk of loss. Always verify live prices and resolution rules on the venue itself. See our full disclaimer.
How to cite W.E.T.
Journalists and researchers are welcome to cite our figures with attribution and a link. Because prices move, please include the “as of” date shown on the page. A citable form:
“As of [date], the prediction market gives [outcome] a [X]% chance (Kalshi [a]% / Polymarket [b]%), according to World Event Trading.”
Every event page also exposes a copy-paste embeddable odds widget and supports oEmbed, so a pasted link unfurls into a live, auto-updating card with attribution. For data partnerships, corrections, or questions, reach us via the contact page.
If you cite a W.E.T. Score, cite it as a signal-quality measure — “a W.E.T. Score of 82/100 (Strong Signal)” — never as a probability of the outcome. The score's components and thresholds are documented in full above and apply identically to every market we track.
Maintained by Corbin King. Last reviewed July 15, 2026. Browse the live prediction desk or the interactive event dashboard.