Kalshi · Polymarket · Live news
The intelligence layer for prediction markets.
Event contracts reprice the moment news breaks — and the edge is gone minutes later. Resolv watches the wire across Kalshi and Polymarket, reprices every affected contract in real time, and tells you what to trade, why, and how much to risk.
PRIVATE BETA NEWS WIRE · CONNECTING
Why now
Prediction markets became an asset class. The tooling didn't follow.
Regulated event exchanges went from curiosity to serious volume — macro, rates, elections, and crypto contracts now trade alongside traditional risk. But the people trading them still work out of a dozen browser tabs, a news feed, and instinct. Every other asset class got a terminal. This one hasn't.
Fragmented discovery
The same question trades on two venues at two prices, under two sets of resolution criteria. Nobody is reconciling them in real time.
Unpriced information
News moves these contracts faster than equities, but there's no layer mapping a headline to the specific contracts it should reprice.
No risk discipline
Position sizing is guesswork, and resolution ambiguity — the risk of being right and still not paid — goes almost entirely unmodeled.
How it works
From headline to sized position, in one pass.
Five deterministic stages. Click any one to see what happens inside it — the same pipeline runs on every refresh.
Evidence graph
Retrieval you can actually see.
Most AI tools retrieve context invisibly and ask you to trust the summary. Resolv makes the corpus a typed graph — markets, headlines, and entities joined by explicit links — so retrieval is a traversal you can inspect, and a claim without an edge behind it can't be written.
Evidence links join headlines to the contracts they bear on. Cross-venue links pair the same question across exchanges. Topic links group by entity.
Each claim in a generated brief carries a footnote that resolves to a node in the graph. Click it and the source is highlighted.
Hide stale news and fair values recompute — because the visible graph literally is the model's input, not a picture of it.
The model
Two model families. One auditable verdict.
Event contracts are binary options, so Resolv uses the right instrument for each — not one formula stretched over everything it doesn't fit.
d₂ = [ln(S/K) + (μ − σ²/2)T] / σ√T
"Bitcoin above $120K on August 1" has a spot, a strike, and an expiry — so it is priced with the Black-Scholes digital-option formula. N(d₂) is the event probability, driven by distance to strike, volatility, and time remaining.
"Will the Fed cut in July" has nothing to hedge against, so Black-Scholes is the wrong tool. Resolv starts from the de-biased market price as a prior and updates it with evidence in log-odds space — where every factor is one additive term.
Because the pipeline is additive in log-odds, the factors provably sum to the answer. No "the model says so."
A forecast is only as good as its calibration. Resolv fits the model's parameters to resolved contract history by gradient descent, then reports honest scoring rules on held-out data — log loss, Brier score, and a reliability curve showing whether 70% forecasts actually happen 70% of the time.
Figures from the current model-lab validation run. Calibration is refit as real resolved history accumulates.
When the evidence isn't enough, Resolv says so
A verdict only fires when the edge clears a threshold and confidence clears a floor. Otherwise the output is HOLD with a machine-readable reason — insufficient edge, low confidence, or no mapped evidence. A tool that always has an opinion isn't a model, it's a slot machine.
Resolution risk
Right thesis. Wrong payout.
Event contracts don't settle on what happened — they settle on what the resolution criteria say happened. Vague criteria are a real, priceable risk, and almost nothing else models it.
Criteria are scored
Resolution text is scanned for discretionary language, undefined terms, and missing authoritative sources. Every contract carries an ambiguity score.
Ambiguity shrinks the estimate
High ambiguity pulls fair value toward 50¢ and caps confidence — visible as its own line in the attribution, never hidden in a fudge factor.
Flagged before you size
Contracts whose payout could contradict a correct thesis are surfaced in the risk panel before position sizing, not after settlement.
Two interfaces, one model
Simple for traders. Deep for desks.
Most tools either hide too much or drown you. Resolv splits the surface by intent — the same fair value underneath.
Real-time intelligence
In event markets, the information is the edge — and it decays in minutes.
Equities absorb news gradually across a continuous price. Event contracts don't: a single headline can permanently move a contract from 44¢ to 61¢, and the gap between the wire hitting and the book catching up is the entire opportunity. Miss the window and you're trading a price that already knows.
Five states, never faked
Speed is worthless if you can't trust it. Every source reports its own health, so you always know whether a number is live, cached, or missing.
A stale number presented as fresh is the most expensive bug a trading tool can have. Timestamps tick live on every row above.
Packaging
Start with signal. Grow into infrastructure.
Private beta pricing. Early access seats include the full Pro surface while we calibrate against live history.
Basic
Everyday traders
- ✓Cross-venue market discovery
- ✓Plain-English verdict and confidence
- ✓Live news mapped to your markets
- ✓Suggested maximum risk
- ✓Watchlist and price alerts
Pro
Analysts and professional traders
- ✓Everything in Basic
- ✓Full factor attribution and reasoning surface
- ✓Evidence graph explorer
- ✓Cross-venue arbitrage scanner with fee math
- ✓Portfolio exposure and VaR-style risk
- ✓Monte Carlo and Kelly sizing tools
- ✓Session export for audit
Institution
Desks and funds
- ✓Everything in Pro
- ✓Pricing-engine API access
- ✓Team seats and entitlements
- ✓Compliance logs and audit trail
- ✓Custom alerting and data feeds
FAQ
What Resolv is, and what it isn't.
Is Resolv an exchange?
Where does the data come from?
Are the model estimates predictions?
Why should I trust a model I can't see?
What happens in the private beta?
Get early access to Resolv.
We're onboarding a small group of traders and desks during calibration. Leave your email and we'll be in touch with a seat.
No spam, no list-selling. We'll only email you about beta access.