Prediction market data API: Kalshi and Polymarket chances for finance questions
By Faizan Ahmad. Updated .
The Tradytics prediction market data API serves every live finance question on Kalshi and Polymarket with the chance of each outcome: where the S&P 500 and Nasdaq close, CPI, jobs and the Fed, and earnings beats for single stocks.
It is built on an archive of 5.95 million Kalshi and Polymarket markets and 41.4 million trades back to June 2021, settled ones included: 5,748,581 Kalshi markets since June 30, 2021 with 30.2 million trades, and 202,780 Polymarket markets with 11.2 million trades. It comes with Pro at $69 a month, 300 calls a minute and 50,000 a day, on the same key as options flow, dark pool and dealer exposure.
What the prediction market API returns
| Endpoint | What it returns |
|---|---|
GET /probabilities | Every live finance question from Kalshi and Polymarket, filtered by kind (close, range, stock, earnings) or symbol, each with its outcomes, their chances and, for price questions, the chance per price range |
GET /probabilities/{id} | One question: the chance of every outcome, where it is likely to land, and for earnings questions the stock's past report-day moves |
GET /probabilities/markets/{market}/history | One outcome's chance through time, hour by hour or day by day, for live and settled markets |
GET /probabilities/outlook | The S&P 500 outlook: where prediction markets put the index at the next close, month end and year end |
GET /probabilities/track | The index close questions over the last 30 sessions: where people put the close at 10 AM, and where it closed |
GET /probabilities/settled | The settled archive, 5,748,581 Kalshi and 202,780 Polymarket markets, filtered by source, series and date, with each result |
Base URL https://tradytics.com/api/public/v1. Every request sends Authorization: Bearer YOUR_KEY. Keys are made on the API keys page.
A real response
The Kalshi question on the S&P 500 at 3 PM on November 5, 2026, read at 2:07 PM New York with the index at 7,373.37: a 90% chance of 7,320 or above, 27% of 7,390 or above, 19.5% of 7,400 or above. Trimmed to four outcomes and three price ranges.
GET /probabilities?group=close
{
"at": 1793905665,
"data": [
{"id": "spx-close-2026-11-05-1500", "src": "kalshi", "s": "SPX", "name": "S&P 500",
"title": "S&P 500 at 3pm on Thu, Nov 5", "group": "close",
"at": "2026-11-05T15:00:00-05:00", "spot": 7373.37,
"top": [
{"market": "KXINXU-26NOV05H1500-T7319.9999", "label": "7,320 or above", "k": 7320.0, "chance": 0.9},
{"market": "KXINXU-26NOV05H1500-T7329.9999", "label": "7,330 or above", "k": 7330.0, "chance": 0.85},
{"market": "KXINXU-26NOV05H1500-T7389.9999", "label": "7,390 or above", "k": 7390.0, "chance": 0.27},
{"market": "KXINXU-26NOV05H1500-T7399.9999", "label": "7,400 or above", "k": 7400.0, "chance": 0.195}
],
"dist": {"kind": "close", "bins": [
{"lo": 7365.0, "hi": 7370.0, "p": 0.05},
{"lo": 7370.0, "hi": 7375.0, "p": 0.07},
{"lo": 7375.0, "hi": 7380.0, "p": 0.05}
]}}
]
}The day before, the 7,515 line for the November 4 close went from a 58% chance at midnight to 99% by 2 PM, and settled yes:
GET /probabilities/markets/KXINXU-26NOV04H1600-T7514.9999/history
{
"data": {"outcome": "7,515 or above", "points": [
{"at": "2026-11-04T00:00:00Z", "chance": 0.58},
{"at": "2026-11-04T02:00:00Z", "chance": 0.35},
{"at": "2026-11-04T08:00:00Z", "chance": 0.75},
{"at": "2026-11-04T10:00:00Z", "chance": 0.94},
{"at": "2026-11-04T14:00:00Z", "chance": 0.99}
]}
}The fields
| Field | Meaning |
|---|---|
id, src | The question's id (for /probabilities/{id}) and its source, kalshi or polymarket |
s, name, title | The symbol it is about, its name and the question |
group | The kind of question: close, range, stock or earnings |
at, spot | When the question resolves, and the underlying's price now |
top[] | The outcomes: market (its Kalshi ticker or Polymarket id, for /history), label, the level k and chance, the market's chance as a fraction |
dist.bins[] | The chance p that the answer lands between lo and hi |
points[] | On /history: the chance at the end of each hour or day |
What sits behind the chances
Probabilities reads the questions traders care about and turns them into what is likely and what happened. The S&P 500 outlook is a fan of where people expect the index at the next close, Friday, month end, three months and year end, with jobs, CPI and Fed releases on the time line. Index close questions are scored against the last 30 sessions, and earnings-beat questions come with each stock's past report-day moves. The settled archive back to June 2021 is what lets a chance be read against how the same kind of question resolved before. The app view is at Probabilities; more in Tradytics Probabilities.
How it joins the other datasets
- The options market: SPY's expected move and term structure beside the chance of each S&P 500 close.
- Dealer exposure: SPX gamma, walls and flip beside the levels the questions are about.
- Bars and the earnings calendar: how the index closed, and which reports are coming.
The backtester reads the same Kalshi and Polymarket archive.
From Claude, ChatGPT or Cursor
In Cursor or any coding agent, GET /probabilities with your key returns the questions as JSON. In Claude and ChatGPT, the Tradytics MCP server at https://tradytics.com/mcp gives the agent the same questions, chances and history as tools, next to options flow, dark pool and dealer exposure, on the same Pro plan. More in the Tradytics API and MCP server.
Limits and plan
| Tradytics Pro | |
|---|---|
| Price | $69 a month or $499 a year, with the full app included |
| Rate limits | 300 calls a minute, 50,000 a day |
| Archive | 5,748,581 Kalshi markets since June 30, 2021 with 30.2 million trades; 202,780 Polymarket markets with 11.2 million trades |
| Also on the key | Options flow since March 31, 2025, ranked dark pool, dealer gamma, delta, vanna and charm, the options market, Signals, AI Fund trades, the scanner, research and market data |
| MCP server | Included |
How it compares
Checked on each platform's own site, November 5, 2026.
| Platform | Market data APIs compared |
|---|---|
| Tradytics | In Pro, $69 a month with the full app: 300 calls a minute, 50,000 a day; options flow since March 2025, ranked dark pool, dealer exposure, Probabilities, Signals and AI Fund trades; MCP server |
| Unusual Whales | Separate: $99 a month (1-year lookback, 25,000 requests a month), $150 (2 years), $375 (WebSocket); MCP with 100+ tools; Polymarket only |
| Quant Data | $149.99 a month, 365+ days of history, 240 requests a minute; MCP |
| Bullflow | $129 a month or pay per request; flow, dealer gamma and dark pool; MCP |
| Quiver Quantitative | From $30 a month; alternative data, no options flow or dark pool prints; MCP |
| Barchart OnDemand | Usage-based pricing; option history and unusual activity; no MCP |
| Cheddar Flow, SpotGamma, OptionStrat, Market Chameleon, MenthorQ, TradeAlgo, InsiderFinance, BlackBoxStocks, Koyfin | No API |
Common questions
Does Tradytics have a Kalshi API?
Yes. GET /probabilities returns every live finance question on Kalshi, and on Polymarket, with the chance of each outcome, in Pro for $69 a month. More in Kalshi data API.
Is there a Polymarket data API for finance questions?
Yes. The same endpoint carries Polymarket's finance questions next to Kalshi's, each with its source in src, on an archive of 202,780 Polymarket markets with 11.2 million trades.
How far back does the prediction market archive go?
To June 30, 2021 for Kalshi: 5,748,581 markets with 30.2 million trades, settled ones included. Together with Polymarket that is 5.95 million markets and 41.4 million trades. GET /probabilities/settled lists them with their results, and /probabilities/markets/{market}/history gives any one's chance through time.
What chance does the market give the S&P 500 close?
Each S&P 500 close question in /probabilities lists its levels with the market's chance, and dist gives the chance per price range. On November 5, 2026 the 3 PM question put 90% on 7,320 or above.
Can I use it from ChatGPT or Claude?
Yes. A coding agent calls the API with your key, and the Tradytics MCP server gives Claude, ChatGPT and Cursor the same questions and chances as tools, on the same Pro plan.