Independent of every exchange

The independent record of prediction markets.

Every market on Kalshi and Polymarket, checked: how it settles, what it costs, whether its twin on another venue pays the same, and whether the trading looks wrong. Built for traders, researchers and investigators.

Request an invite See a sample record

15 Minute Markets is independent and is not affiliated with, endorsed by, or sponsored by Kalshi or Polymarket.

I trade I research I investigate I build agents or systems

Market Record · Sample valuesKalshi · KXBTC15M

Bitcoin above $67,250 at 14:15 ET?

rec:btc-15m/kalshi/KXBTC15M-26OCT241415-T67250

How it settlesBRTI, average of the last 60 seconds before close. Rule quoted, unchanged since listing.

Rule read

What it costs100 contracts at 0.62, taker: $1.6492 fee. Break-even 0.6365.

Fee to $0.0001

Its twins Polymarket US, same index, same window, 412 of 412 settled alike.

The recordSeries settlements reproduced from the rule: 23,827 of 23,828.

Audited

The tradingNo open flags. Each detector's false-alarm rate: not yet measured on real tape.

Not graded

As of 14:09:30 ET · record v1 · hash 7c1e…a90Outage rule: a data outage resolves every strike to No

19,104 / 19,107Kalshi crypto 15-minute settlements reproduced from the rulesettlement_audit · as of 2026-11-13

484,835markets catalogued daily across three venues, rules and sources keptmarket_catalog · as of 2026-11-13

4 in 600cross-venue matches wrong in the last two audits, publishedxv_match audit · as of 2026-11-04

51 trials, 0 proven, 0 findingsstrategy trials registered before we looked; every one publishedtrial registry · as of 2026-11-13

Proven holds the chance of any false pass in its family to 5% and is the only tier a strategy gate reads. A Finding holds the false discovery rate to q 0.10: research only. How the tiers work

The Market Record

One market, five answers, each with its source

A raw feed tells you the price. The record tells you whether to trust it. Every field names the method behind it, when it was measured, and how often that method is wrong.

01 · Rule

How will it pay?

The rule quoted word for word, the settlement source, the window, and every change to the rule since listing.

market_catalog · catalog_diff

02 · Cost

What does it cost?

The exchange fee for your size and price, maker or taker, from the series' own fee type. Break-even beside it.

fees.py, pinned to real fills

03 · Twins

Does its twin pay the same?

The same question on other venues, graded A to D on the rule text and past settlements. In our own audit, 50 of 8,911 matched pairs did not settle alike.

xv_identity · xv_equivalence

04 · Record

Has it paid by the rule?

Past settlements of the series reproduced from the source, and calibration by horizon for its family.

settlement_audit · calibration

05 · Integrity

Does the trading look wrong?

Flags across all three venues on one clock, grouped into one case per event. A flag shows its false-alarm rate once measured on real tape; until then it says not yet measured.

integrity_cases · base-rate atlas

Three desks

The same record, three jobs, three steps each

For traders on Kalshi and Polymarket US

Trade with fewer expensive mistakes

1
WatchA watchlist of records and saved views. Hear when a rule changes, a twin stops paying the same, or news lands on a market you hold.
2
CheckBefore an order and while you hold it: rule, fee, best venue, settlement meter and outage exposure on one screen.
3
JournalEach fill with the record as it stood, reviewed weekly.

Research plan · $500 a yearSee the desk Members

For quants, funds and AI labs

Research that survives the trial count

1
AskA question in plain words, or a filter in the catalog.
2
PullThe dataset with its method and the number of trials behind it.
3
CiteA permalink with its date, its method and its corrections.

Institutional $5,000 · Data $15,000See the desk Members

For counsel, regulators and integrity teams

Show what a market knew, and when

1
CaseOne screen per incident, keyed to the market: the verdict, the reasons against its baseline, what rules it out and what is missing.
2
TimelineBooks, trades and the first public news on one clock, across venues, with the book replayed around the event.
3
BundleA derived-only evidence bundle with a hash chain, ready for counsel.

Every detector publishes its measured false-positive rate on the methods page, or says not yet measured.

Per matter, through counselSee the desk Members

For investors

What prediction markets priced before the macro print

Portfolio managers ask what these markets know about rates, inflation and elections. We record them on three venues and measure how well they priced past outcomes, what moved after each release, and where two venues disagreed.

See what we measure

Calibration of macro marketsCPI, Fed and payrolls markets, priced against what happened
Event ChainsEvery market a release touches, and what moved last time
Midterms case studyHow close the markets were before polls closed

For systems and agents

Your code and your agent read the same record

One schema, one version, one content hash, whether you read it on the site, pull it from the API, load it into your warehouse or hand it to an agent.

Data API

Records, twins and series over HTTPS

GET /api/derived/record
  ?id=KXBTC15M-26OCT241415-T67250
→ record.v1 · hash 7c1e…a90

OpenAPI 3.1 · Institutional and Data plans

Files and warehouse

Daily Parquet and a Snowflake share

records/2026-10-24.parquet
family_grades/2026-10-24.parquet

Derived fields only, never raw books

MCP for agents

Read-only tools for your agent

get_market_record · find_twins
watch_record · list_datasets

Order tools are gated: off by default, site owner only, and need your approval

Browse the data catalog Members · every dataset with its schema, history, freshness and licence status

Coverage

Seven families, three venues, never pooled

Family	Kalshi	Polymarket US	Records since
Crypto, 15 minutes	Full books, every trade	BTC books and trades	by Nov 7
Crypto, hourly and daily	Full ladder depth	Books	Nov 9
Sports	Books, in-play, play-by-play	Books	Oct 5
Economics and politics	Books, release calendar	Books	Oct 5
Weather and mentions	Books, NWS reports	Not recorded	by Nov 7

Gaps in the record are counted and shown, not filled. Catalog and archive read Nov 13 · ledger as of Nov 13 · 23 of 23 recorders running.

How we compare

Others sell the data. We sell the check.

Buy raw books from the venues and their distributors. Come here for what a raw feed cannot tell you.

Question	Venue feeds and distributors	Unified APIs	Wallet trackers	15 Minute Markets
Does the twin on another venue pay the same?	One venue each	Matched by title	No	Graded A to D on the rule, audit published
Did past settlements follow the rule?	Venue's own word	No	No	Reproduced from the source
How often is this flag wrong?	Internal	No	Not stated	On every flag
What would explain it innocently?	Internal	No	A score, no tests	Four tests on every trading-pattern case (public news, copy trading, market making and chance), each marked
All three venues on one clock?	No	Prices only	Global Polymarket only	Books, trades and news
Raw tick history and full depth	Yes, licensed	Some	No	No. We do not resell raw data
Wallet identities	No	No	Yes	No. Markets, never people

What we are not

Four things we will not do

Not a venue

We list no markets and take no side, so we are free to grade any venue's markets.

Not a data reseller

No raw books, trades or tape leave our systems. You get checks and derived facts.

Not a tipster

Impersonal research, not advice tailored to you. No strategy has passed our own bar yet, and we say so.

Never a name

Integrity work describes markets. We never name a person in public.

Plans by role

Annual plans. Access is by invite.

Traders and researchers

Research

$500 a year

Every Market Record
Pre-trade check, position screen and journal
Watchlists, saved views and email alerts
Ask and MCP

Desks and research teams

Institutional

$5,000 a year

Firm licence
Data API, 10 keys
Webhooks and the settlement monitor
Quarterly call

Funds and AI labs

Data

$15,000 a year

25 keys, 100,000 calls a day each
Daily Parquet of records
Eval sets with settled answers

Counsel and integrity teams

Matters

By matter

One-screen case with what rules it out
Case timeline across venues
Watch a redacted account under your matter
Derived-only evidence bundle
Through counsel, by request

Compare every plan

Free research by email

Independent research on Kalshi and Polymarket, every week

New studies and findings, each with its sample size and method, and what failed as well as what held. Written by a firm that runs no venue.

One email a week with what we published
Free, no account; we send a confirmation link first
Unsubscribe in one click

Get the weekly research

Bring us a market. The record will say no when the answer is no.

Independent of Kalshi and Polymarket. One founder, Scott Gaines, and a conflicts policy, published once adopted.

Request an invite See a sample record