Help & FAQ

MarketClaims is an evidence-based research tool for prediction markets. We read the news 24/7, extract checkable claims with sources, and summarise them into a conservative directional lean for a given market.

We are research only - no trades, no wallets, no funds held.

There are two ways in:

• Browse Discover and add markets you care about to Watch.
• Paste a Polymarket or Kalshi URL into the box on Discover to run a one-off brief.

New accounts start with a small pool of free analyses so you can try the product before adding credits.

Polymarket and Kalshi markets. Paste a market URL, or browse Discover by category.

A claim is a discrete, checkable statement extracted from a news story - pinned to its source, publication time, and language - that bears on a market's resolution criteria.

Each claim is labelled as supporting YES, supporting NO, or neutral / mixed.

Watch is the list of markets you have chosen to follow. Each row shows how much the research behind it can be relied on, how our read compares with the price, and - once the market settles - what it actually did.

Markets are grouped by whether we have a usable read at all. Most of the time we do not, and the page says so rather than filling the space.

Watch is the markets you are following now: what we make of each one, and what to do about it. Reports is every brief you have already run, newest and most serious first.

Think of Watch as the live list and Reports as the archive.

On Discover, use "Analyze" on any market card, or paste a Polymarket or Kalshi URL into the box at the top.

Watching a market is always free. Running an analysis on it costs one credit, taken when the report completes.

Because that is usually true. On roughly four markets in five, too few independent sources argue a direction clearly enough for a read to mean anything, so the report says so instead of dressing up a number.

The price may be perfectly good in those cases - we simply have nothing of our own to add.

An analysis is one end-to-end research run on a single market: collect evidence, score it, build the report, save it. Each completed analysis produces a Report that you can revisit, share, and grade against the eventual outcome.

Most analyses finish in 1-3 minutes. Markets with sparse coverage or complex resolution rules can take longer. You can navigate away - the finished report will appear in Reports when it is done.

The fast polling window elapsed without a final status, but the server job has not failed. We do not declare a false failure - that would let you retry and spend a second credit on the same market.

Leave the tab open or check back later from Reports. No additional credit is used while you wait.

Open Reports from the top nav. Reports are listed newest first and stay there even after your free analyses are used up - they will not disappear because you ran out of credit.

Yes - every report has a stable URL of the form /reports/<report-id>. Sharing the URL lets a teammate read the same evidence, sources, and claim breakdown.

Online news coverage monitored 24/7 in 20+ languages, spanning real-time stories and recent-weeks context. Every claim links to its source so you can read the original article.

The lean is a directional summary of the claims we collected:

• YES - the weight of sourced evidence favours the YES outcome.
• NO - the weight of sourced evidence favours the NO outcome.
• INSUFFICIENT EVIDENCE - coverage is too thin or conflicting to call.

A lean is not a prediction and not a recommendation to trade.

When evidence is thin or conflicting we say so - the verdict can be insufficient evidence. We would rather abstain than overreach. Abstentions are not failures; they are honest scorecards.

Confidence (low / medium / high) reflects the quantity, recency, and agreement of sourced claims, plus how directly they speak to the resolution criteria. The report explains the main confidence factors in the Why this confidence panel.

Resolution risk is the chance the market resolves in an unexpected way because of how its rules are written - for example, ambiguous wording, oracle behaviour, or edge cases the news coverage does not address. Reports surface this separately from the directional lean.

• Live - the report was built against fresh evidence at run time.
• Cached - evidence came from a recent cache that is still within the freshness window.
• Stale - the live evidence provider was temporarily unavailable, and we fell back to an older cache. Treat stale reports with extra caution.

New accounts get ten free analyses. They are a one-time pool, not a monthly allowance. Each completed analysis uses one. Watching markets, browsing Discover, and re-reading existing reports never cost a credit.

Only running a new analysis costs a credit. Specifically:

• Pasting a URL and clicking Analyze.
• Clicking Analyze on a Discover card, which also adds the market to Watch.
• Re-running an analysis from a report page.

The credit is taken when the report completes, so a failed job never costs one.

Nothing is charged. The credit is taken only when a report completes, so a job that ends in a failed state never costs you one.

If a job stays in "still running" we keep checking it, and you are not charged a second credit for that market.

You can buy a pack of credits, or ask for some.

Every account starts with ten free analyses, once, not as a monthly allowance. When those run out, the Pricing page lists the packs and what each one costs. There is no subscription: a pack is a one-off purchase and the credits in it do not expire.

You can still ask for credits from the Pricing page instead, and we grant them by hand.

Credits you have not spent can be refunded within thirty days of the purchase that provided them, at the price you paid for them. Email us and we will sort it out.

Refunds are issued by Paddle, who sold you the pack, back to the card you used. The full terms are on the Refund Policy page.

No. Card details are entered on a page hosted by Paddle, who sell the credit packs, and they never reach our servers.

What we keep is the order, the amount, Paddle's reference for it, and the last four digits of the card, which is enough to find your payment if you write to us about it.

No. MarketClaims is a research tool. Our leans are evidence summaries, not recommendations to trade. We place no trades and hold no funds.

No. There is nothing to connect. MarketClaims is read-only research - it holds no funds and connects to no wallets or exchange accounts.

We do not promise that, and you should be wary of anyone who does. What we promise is sourced evidence and an honest scorecard - including the misses.

Calls are timestamped before resolution and graded against actual outcomes - hits, misses, and abstentions alike. Markets you watched that have settled appear on Watch, with what they actually did.

The full Terms of Service, Privacy Policy, Risk Disclosure, and Research Disclaimer live under /legal/. You accept them when you first run an analysis. Versioned acceptance is recorded so you always know which version you agreed to.