Real-time analytics terminal for Polymarket & Kalshi
We price every crypto Up/Down window from five live exchange feeds, show you where that price and the order book's ask disagree — and publish our own report card every night, losses included.
Free, no card, nothing to install.
Five exchange books over direct websockets — Binance, Coinbase, Kraken, OKX, Bybit — re-based onto the feed that settles the market and folded into one composite price. Four assets, three horizons.
Our probability minus the real ask — the price you would actually pay, spread included, not the midpoint. Both prices are written down, timestamped, while the answer is still unknown.
The official result grades both prices — Brier, calibration, paper P&L with 95% intervals — and the night lands on /performance. Nothing archived before the outcome can be changed after it.
Why the order matters: grading only counts because step 02 happens before step 03 — every prediction is committed while the answer is still unknown, which is what makes the record impossible to retrofit.
Three cards, recomputed from the archive every morning: the average, the tails, and the paper profit and loss.
On average
Brier score at sixty seconds to expiry, pooled across the 174,317 resolved windows archived between 2026-07-13 and 2026-11-07. Lower is better.
On the average window the book is better calibrated than we are, by 0.0026 of Brier. A liquid book pools thousands of opinions; beating it on every window is not a claim any honest shop can make.
In the tails
Paper result per fill of the rules that buy only when the model sits at least that many points above the ask. Every case is archived before its outcome and graded after it.
Paper P&L
A coin flip loses 1.0¢ a fill — the cost of crossing the spread. Buying the favourite loses 0.5¢ while winning 93% of its fills. 13 of the 13 live rules clear zero.
Data through · consolidated
Paper profit and loss per fill, with 95% interval
Every rule is replayed over every archived session at one dollar a fill, and reported with a 95% confidence interval. Two are deliberate controls — a coin flip and a buy-the-favourite. If the harness were leaking, the decoys would drift above zero. They do the opposite: the coin flip loses 1.0¢ a fill — the cost of crossing the spread — and the favourite loses 0.5¢ even while winning 93% of its fills, because comfort trades at a premium. That baseline is what every live rule has to clear.
Every window archived and graded the night it resolves. The record is 116 sessions deep and has been public since the first one — read it before you believe a word of this.
A web terminal — mispria.com/terminal, nothing to download — that lists every crypto Up/Down window currently in play on Polymarket and Kalshi, prices each one independently, and shows you where its answer and the book's answer stop agreeing.
The same archived tick as the replay above — BTC 15m, 5 Aug 2026, T−02:34. Under about two points the edge is inside the noise of estimating volatility, and we say so.
Pro: a message when the executable edge on a window you follow crosses your threshold — never in the last minute, never more often than your cooldown. Free: the morning recap and three /market lookups a day.
Our probability minus what it costs to buy — the ask, spread included, not the midpoint. That is the only version of the number you could have acted on, so it is the one the panel puts in 58-pixel type. Under about two points it is inside the noise of estimating volatility, and we say so.
Ours and the book's, on one axis, for the life of the window. When they sit on top of each other — which is most of the time — you can see that too. A measurement that only speaks up when it disagrees isn't a measurement.
Five exchange books over direct websockets, each one re-based onto the scale of the feed that settles the market, then reduced to their median. The second tile is the lead: how far in front of that feed the composite currently sits.
The distribution of the terminal move over the time that is left, with the distance to the price to beat drawn on it in σ. A probability with no distribution behind it is a number you have to take on trust.
Price to beat, distance in dollars and in σ, σ over the remaining horizon, annualised against realised volatility. Everything the probability was computed from, so you can disagree with it — and a footer that states the resolution rule in full.
The venue layer is built to be swapped — feeds, order book and resolution — so the same measurement can run anywhere binary markets trade. The terminal reads Polymarket and Kalshi today.
If you can't check it, you shouldn't trust it — so here it is, including the parts where it breaks.
Crypto never closes, so volatility is annualised over 365×24×3600 seconds. Borrowing the equity market's 252 trading days here is a factor-of-two error, and it survives in a lot of published crypto option maths.
Variance is additive in time, so the standard deviation grows as σ·√t. Over a fifteen-minute window drift is negligible against noise, so we set it to zero rather than pretend to estimate it. Above two hours we say so on screen.
A Gaussian tail underprices large moves by one to two orders of magnitude. We also run Student-t with ν=4, renormalised so its variance still matches σ. LTCM is the reference case: a thin-tailed model in a fat-tailed world, at 25× leverage.
Liquidation cascades are jumps; no diffusion captures them — the fat tail only softens the blow. Sub-minute sampling picks up bid-ask bounce rather than signal. If returns autocorrelate, √t itself breaks. We publish through those nights too.
Binance quotes bitcoin against a dollar-pegged token, Coinbase against actual dollars, and the feed that settles the market publishes a third number on its own clock. Left alone they sit tens of dollars apart, and none of them is wrong — they are quoting different things. Before any of it becomes a probability, every source has to be pulled onto one scale.
Five exchange books arrive over direct websockets; the blue line is the raw feed behind the resolution series, which is the scale the market will actually be settled on. Each source carries a standing basis — the USDT venues trade at a premium to the USD ones — so we hold a per-source offset, updated as an exponential moving average (α = 0.07) against that feed, and apply it before anything else happens. What converges here is not the price of bitcoin; it is five quoting conventions being pulled onto one ruler. The composite is the median of the five corrected books, which is what the model prices off — a median so that one stalled venue moves nothing. The tick sequence is a seeded illustration of the mechanism; the values it lands on ($64,444.96 composite, 81.7% model, 61¢ to buy) are the archived tick this whole page follows.
The same composite, probability and gap run live in the terminal — on every window in play.
Open the free terminalNone of these are opinions. They fall straight out of the maths, and they are the reason a window that looks settled with two minutes left often isn't. Each one plays when you open it.
A binary market resolves against a single published price. That feed is authoritative, and it is also slower than the venues it summarises — between two of its updates it is quoting a price the market has already left. We blend five exchange books into one composite, recalibrated to the feed's own scale, which usually means we know where the feed is going to land before it lands there.
The pale lines are five exchange books; the amber line is the composite — median of the five, recalibrated to the feed's scale — which is what the model prices off. The blue staircase is the resolution feed — flat until it republishes, then a jump to catch up. Every shaded band is an interval where the two disagree, and the direction of that disagreement is what we measure. Measured over the archived session of 13 July 2026: correlation 0.62 between the composite lead and the subsequent feed move, 75% directional hit. The tick sequence below is a seeded illustration of the mechanism, not a recording — the two statistics are the measured ones.
σ scales as √t, not as t. Burn three quarters of the clock and you have not burned three quarters of the risk — you still carry $90.86 of it, exactly twice the $45.43 a linear reading would give you. It is why the last ninety seconds of a window move the price more than the first ten minutes, and why a market that looks decided usually isn't yet.
Both curves have the same σ — the Student-t is renormalised so its variance matches. They only disagree about how often the extremes arrive, and the disagreement is invisible until it isn't: they track each other to about 2σ, cross 10× at 3.3σ and 100× at 4.1σ. Log scale on the vertical axis, or the tails would be an invisible smear along the floor. This is the LTCM shape: a thin-tailed model in a fat-tailed world.
120 paths of the same model, all starting half a σ below the price to beat, all seeded identically so this figure is the same every time you load it. The reflection principle says P(touch) = 2 × P(finish beyond) for a driftless walk; the sample lands on 74 and 37. That is why the panel quotes both numbers — a window can spend most of its life on the right side of the line and still resolve on the wrong one.
The free tier is meant to be enough to convince you the measurement is real. It is not meant to be enough to act on.
Today: €4.95 · On : €24.99/month · Cancel in one click from your account
Try a month for €4.95Today: €4.95 · On : €119.88/year · Cancel in one click from your account
Try a month for €4.95Not on sale yet. Join the list and we will email you when it opens.
Join early accessPrices in EUR, taxes included.
At most one short email a week: the track record in three numbers, one research note, one market. You confirm by email; every email has a one-click unsubscribe link. Privacy