Start here
What Edgecast does
Prediction markets like Polymarket let people bet on real-world events. The price of a contract is a probability: if “Will X happen?” trades at 47¢, the market thinks there is a 47% chance.
Polymarket tells you that price. It does not tell you whether the price is wrong. Edgecast estimates what each probability should be, shows you where that disagrees with the market, and — the part most tools skip — keeps a permanent record of how often it was right.
What does the market believe?
The live price, order book and volume across every tradeable Polymarket market. Plain fact, observed and timestamped.
What should the probability be?
Eight independent signals combine into one fair-value estimate, with an explicit confidence attached.
Why is there a difference?
Every estimate shows which signals moved it, by how much, and what would prove it wrong.
Anatomy of an opportunity
A real row from the scanner right now — not an example. Here is how to read it.
| Market | Market | Fair est. | Edge | Confidence | Score | EV / $1 | Liquidity | Resolves |
|---|---|---|---|---|---|---|---|---|
| Will the Fed increase interest rates by 25 bps after the September 2026 meeting? | 51.5% | 54.2% | +2.71pp | 0.58 | 4.63 | +0.041 | $604.9k | 7.7d |
What Polymarket says right now.
What our models think it should be.
Positive: we think the market is too LOW.
Medium — 3 signals contributed.
Out of 10. Edge × confidence × tradeability.
Profit per $1 risked, if we are right.
Money resting in the book. More = easier to trade.
Time until the outcome is known.
Click any market name to open its detail page, where you get the price chart, the live order book, a breakdown of every signal that contributed, and — importantly — a counter-thesis arguing why the estimate might be wrong.
The most important thing to understand
Fact and estimate are never mixed
Things we observed
Prices, order books, trades, volume, liquidity. Read directly from Polymarket at a known moment. If the data is late, the header says so rather than quietly showing you a stale number.
Things we computed
Fair probability, edge, confidence, signal score, expected value. These are estimates, and they can be wrong. Every field name ends in _estimate in the API for exactly this reason. We never present a model number as truth.
The numbers, in plain English
- Market probabilityFACT
- What Polymarket currently prices. 64.5% means the market thinks it is roughly two-in-three.
- Fair estimateEST
- What our models think the probability should be. The whole product is this number and how much to trust it.
- EdgeEST
- Fair estimate minus market price, in percentage POINTS (pp). +5pp on a 40% market means we think it should be 45%, not that it should be 42%. A bigger edge is more interesting and less likely to be right.
- ConfidenceEST
- How much the model trusts itself here, 0 to 1. Driven by how many independent signals agree and how liquid the market is. Low confidence deliberately pulls the estimate back toward the market price — if we do not know, we defer.
- Signal scoreEST
- A 0–10 summary combining edge size, confidence, and how tradeable the market actually is. A huge edge in a market nobody can trade scores low, because it is not an opportunity.
- EV / $1EST
- Estimated profit per dollar risked, if our probability is right. Calculated at the price you would actually pay (crossing the spread), net of fees — not the friendlier mid-price.
- Liquidity & spreadFACT
- How much money is resting in the order book, and the gap between buy and sell prices. Thin liquidity or a wide spread can eat an edge entirely.
- DislocationFACT
- When related markets contradict each other arithmetically — e.g. all candidates in one race adding up to 108% instead of 100%. This is arithmetic on observed prices, not a model opinion, which makes it the most solid signal type we have.
- Brier scoreEST
- How accurate past forecasts turned out to be. Lower is better. Only meaningful against a baseline, so we always show it next to the market’s own Brier — beating the market is the only claim that matters.
What this cannot do
Worth reading before you trust any number on the site.
It is not proven to beat the market yet.
The track record currently shows a positive score, but our own analysis flags it as concentrated in one direction on a biased sample. The site argues against its own headline number rather than selling it to you. Judge it again after more markets resolve.
It cannot execute trades.
There is deliberately no order-placement code. Edgecast is analytics only — it never touches your money or your wallet.
It is not financial advice.
These are estimates about uncertain events. Prediction markets carry real risk of loss, and a 70% probability is wrong three times in ten by definition.
It has blind spots.
No news analysis, no AI copilot, and wallet intelligence is shallow so far. Signals lean on order-book and price data, so a market moving on information we cannot see will surprise it.
Where to go next
Discovery
The home screen. Biggest estimated edges, breaking moves, trending markets.
Edge Scanner
Every opportunity, filterable by category, edge size, confidence and liquidity.
Dislocations
Markets that contradict each other. Pure arithmetic, no model needed.
Track Record
Our scorecard, caveats first. Start here if you want to know whether to believe any of it.
Tip: press ⌘ K anywhere to search markets by name. Currently tracking 9,144 markets.