Prediction-market intelligence
Polymarket and crypto market structure, on one screen. EdgeMarket reads the capital that arrives first — whale positioning, order flow, funding, liquidations.
The premise
Every market has a moment where information exists but price has not yet agreed. That gap is the edge. We instrument it — every wallet, every fill, every shift in conviction — and hand it to you while it still pays.
The base rate is unforgiving. Prediction-market traders lost that much, net, across roughly two million Polymarket addresses, while bots racked up the gains. EdgeMarket exists because that number is what it is. It is the distribution you are entering, not evidence you will beat it.
Bloomberg, April 28, 2026 — read the report →
EdgeMarket — measured, published, and named where it fails
Capabilities
Three products on one dataset. Each ships with the measurement behind it — coverage, hit rate, and the cases where the edge disappears.
Continuous surveillance of the wallets that actually move markets. Not by trade size — size predicts nothing, and we publish the measurement. By measured alpha, validated out of sample.
A directional read on BTC, ETH, SOL and XRP, resolved every five minutes. Every call ships with a calibrated confidence band, and the full hit-rate table is public — including the bands that pay nothing.
Funding, open interest, liquidations and positioning across four venues, sampled every minute. Not a bare live number — a live number with the window it covers and the age of the reading it came from.
The animated panel on each card illustrates the instrument — it is not a live readout, and no figure is read from it. Every number above is measured; Oracle values are from the August 15, 2026 snapshot.
Oracle
A five-minute call, confirmed two minutes before the slot resolves. Below, the calibration table in full — how these markets actually resolve at each price, including the bands where the market is priced better than we are.
| Price at T+180s | Slots | Resolved that way | Gap (pts) |
|---|---|---|---|
| 0.00 – 0.40 | 1,184 | 31.76% | −4.12 |
| 0.40 – 0.50 | 3,906 | 43.63% | −2.62 |
| 0.50 – 0.60 | 9,907 | 55.57% | +0.51 |
| 0.60 – 0.70 | 12,702 | 66.92% | +1.84 |
| 0.70 – 0.80 | 15,650 | 78.24% | +3.18 |
| 0.80 – 0.90 | 20,375 | 88.72% | +3.51 |
| 0.90 – 0.95 | 15,780 | 95.35% | +2.67 |
| 0.95 – 1.00 | 27,542 | 98.88% | +1.24 |
Rows marked † (0.70 to 0.90) carry the widest measured gap — about the size of the 3.5% it costs to take liquidity. It is a map of where short-dated markets misprice, not a return.
Confirmed hit rate · all-time
One snapshot, August 30, 2026. The rate covers the slots that reached settlement; the rest had no retrievable outcome and are excluded, not counted wrong.
Below 0.50 the gap turns negative: those sides resolve less often than they are priced — by 2.8 points between 0.40 and 0.50, and by 5.0 points below 0.40. Above 0.95 the gap is real but small, 1.4 points against roughly 3.5% to take liquidity. The widest gap is 3.46 points, between 0.80 and 0.90, and is about the same size as that fee. We put the whole table on the page instead of one flattering number.
Read the public register →The price is read 180 seconds into a 300-second slot, on the side the signal takes, and compared with how the slot actually resolved. Gap is the resolved share minus that price, in points — negative means the market overprices that side.
Live signals
A directional call on four crypto markets, once per five-minute slot, scored against the settled outcome. Every number below is recomputed from the public register — nothing here is illustrative.
Live
Verified accuracy
82.31%
25,108 settled slots
Verified accuracy
83.02%
27,378 settled slots
Verified accuracy
82.77%
26,856 settled slots
Verified accuracy
82.88%
28,051 settled slots
Direction is held back on the public register —see plans
Accuracy is measured on slots that reached settlement. It is a record of past calls, not a forecast of future ones, and it is not investment advice. Read the full register
Live tape
Every line is a qualifying trade pulled from the live feed, timestamped in UTC and rendered on the server — so a crawler, a visitor without JavaScript and you all read the same rows.
Live
| Time | Wallet | Market | Side | Price | Size |
|---|---|---|---|---|---|
| Ghostly Viper #80xee88...2826 | SS Lazio vs. Genoa CFC: O/U 2.5 AND Cagliari Calcio vs. FC Internazionale Milano: O/U 2.5 | BUYYES | 0.78 | $6.3K | |
| Scarlet Neptune0xb595...5922 | Counter-Strike: SAW Youngsters vs Noir Verse - Map 1 Winner | BUYYES | 0.97 | $5.1K | |
| Scarlet Neptune0xb595...5922 | Counter-Strike: SAW Youngsters vs Noir Verse - Map 1 Winner | BUYYES | 0.97 | $5.8K | |
| Iron Cobra #140xb87b...08c5 | Counter-Strike: SAW Youngsters vs Noir Verse - Map 1 Winner | BUYYES | 0.97 | $12.4K | |
| Ghostly Falcon #30x92d8...9762 | Will Eyüpspor win on 2026-08-30? | BUYYES | 0.94 | $5.8K | |
| Storm Triton #20x8806...2e83 | Eyüpspor vs. Alanyaspor: O/U 2.5 | SELLYES | 0.96 | $22.7K | |
| Cobalt Hunter #140x6373...a0b7 | Dota 2: Level UP vs Power Rangers - Game 1 Winner | BUYYES | 1.00 | $6.9K | |
| Ghostly Barracuda #60x5a05...12a2 | Dota 2: Level UP vs Power Rangers - Game 1 Winner | SELLYES | 1.00 | $6.9K | |
| Scarlet Titan #50x58f8...dc70 | LoL: Fnatic vs GIANTX (BO3) - LEC Regular Season | BUYYES | 0.77 | $13.5K | |
| Onyx Hydra #50x1f71...b2e4 | Will Eyüpspor win on 2026-08-30? | BUYYES | 0.94 | $20K |
10 lines · refreshed every 12 seconds
The public feed is capped: this section shows what the API concedes at that moment, never a padded window.
How it works
Every settlement and order event, normalised and resolved to persistent wallet identities — including the multi-counterparty fills most trackers collapse into one.
Each fill is scored against the wallet's own measured history and the depth of the book at that instant. Accumulation, exit, hedge or noise.
What survives the filter reaches you on the channel you chose — and it is written to a public register whether it works out or not.
Operators
Ranked by measured alpha — the price movement that follows their trades — not by displayed P&L. Only 1.2% of wallets have a computable win rate, and three quarters of those show exactly 100% on a handful of trades. That is not skill, and we do not publish it.
| Entity | Style | Markets | Alpha | t | Hit rate | Volume · 90d |
|---|---|---|---|---|---|---|
| 01Obsidian Hydra #4 | Whale | 29 | +44.9 | 10.1 | 93.1% | $101.5K |
| 02Arctic Atlas #2 | Whale | 39 | +43.3 | 8.2 | 89.7% | $138K |
| 03Amber Emperor #2 | Whale | 29 | +39.4 | 7.9 | 93.1% | $73.8K |
| 04Stealth Orca | Whale | 30 | +31.6 | 4.2 | 83.3% | $114.4K |
| 05Crimson Poseidon #2 | Whale | 53 | +31.6 | 8.5 | 86.8% | $956.1K |
| 06Iron Blade #5 | Shark | 26 | +30.8 | 4.4 | 80.8% | $253.6K |
| 07Iron Juggernaut #4 | Whale | 32 | +27.0 | 6.7 | 84.4% | $154.7K |
| 08Thunder Tsunami #2 | Whale | 31 | +23.6 | 6.2 | 93.5% | $226.9K |
Measured on trades from 2026-04-01 to 2026-06-24, prices between 0.05 and 0.95. Aggregated per market before per wallet, so one active market cannot dominate. 16,837 wallets evaluated; entries require at least 25 distinct markets and a t-statistic above 2.5. These figures describe the measured window — they are in-sample, and promise nothing about the next one.
Bar length is the row's alpha relative to the top entry. It carries no information the number beside it does not already state.
New on the desk
One reads a chart you hand it and writes the whole plan out. The other listens to X, asset by asset, and never prints a score without the count it rests on.
Chart analysis
Paste a screenshot, drop a file, or open one of the crypto assets already quoted on the desk. Crypto, foreign exchange, equities — if the chart is legible, it can be read.
Market pulse
What is being said on X, asset by asset, read from public posts. Crypto opens first; foreign exchange and equities sit behind it, as context.
The two figures above are drawings, not screenshots. They carry no price, no score and no measurement — the real ones are computed on your own assets, inside the product.
Access
One dataset, three depths. What changes between tiers is not the data — it is how early you see a signal, and how far into the raw series you are allowed to go.
For anyone who wants to check the measurements before paying for them.
$0.00/mo
Oracle signals
BTCETHSOLXRP
Public register only — signals are not delivered live on this tier
For the trader who works one book daily — Polymarket, crypto, or both.
$39.00/mo
Billed every month. Cancel any time.
Oracle signals
BTCETHSOLXRP
5 min slot · 82.68% verified on these markets
For the desk that executes size and wants the raw series behind the screens.
$79.00/mo
Billed every month. Cancel any time.
Oracle signals
BTCETHSOLXRP
5 min slot · 82.76% verified on these markets
Bring your own thesis — we supply the flow, the measurement, and the register that says whether we were right.