Kalshi and Polymarket vs sportsbooks in one call

Request the books and the fee fields together

Select the moneyline with markets=h2h, name the books, and ask for oddsFormat=decimal. fees=taker adds net_price without changing price. Authentication uses the X-API-Key header.

curl --fail-with-body -H "X-API-Key: $ODDSLINK_API_KEY" \
  "https://oddslink.app/v1/sports/nba/odds?markets=h2h&bookmakers=kalshi,polymarket,draftkings,fanduel,pinnacle&fees=taker&oddsFormat=decimal"

The response is an array of fixtures. Each fixture has bookmakers, each book has markets, and each market has outcomes. The same outcome name identifies the side within that fixture. See the fee field definitions for supported schedules and rounding.

The fee flips the leading price

Read on 31 October 2026 at about 22:22 UTC: Oklahoma City Thunder at Houston Rockets, scheduled for 2026-11-01T00:30:00Z. Fixture id: nba:houstonrockets-oklahomacitythunder:2026-11-01. These are recorded quotes from that read, rather than a current offer.

On the Thunder, Kalshi and Polymarket both show 2.27 before fees. Kalshi's per-contract coefficient is 0.07; Polymarket's is 0.05. Their returned net prices are 2.184 and 2.208. FanDuel at 2.24 then leads the recorded comparison, followed by Pinnacle at 2.23.

On Houston, Pinnacle's 1.75 exceeds Kalshi's 1.701 and Polymarket's 1.715 after fees. Keep the raw and net columns separate. The decimal raw quote is rounded to two places, while net_price has three; recalculating from the displayed raw number can differ from the returned net quote.

A shared fixture id solves the join. You still need equivalent settlement rules, the same full-game contract and enough available size before comparing executable orders.

Read the raw price, net price and timestamp in Python

import json, os
from urllib.parse import urlencode
from urllib.request import Request, urlopen

params = urlencode({
    "markets": "h2h",
    "bookmakers": "kalshi,polymarket,draftkings,fanduel,pinnacle",
    "fees": "taker",
    "oddsFormat": "decimal",
})
request = Request("https://oddslink.app/v1/sports/nba/odds?" + params,
                  headers={"X-API-Key": os.environ["ODDSLINK_API_KEY"]})
with urlopen(request, timeout=20) as response:
    fixtures = json.load(response)
for fixture in fixtures:
    for book in fixture["bookmakers"]:
        for market in book["markets"]:
            if market["key"] != "h2h" or market.get("period"):
                continue
            fee = market.get("fees")
            for outcome in market["outcomes"]:
                raw = outcome.get("price")
                if raw is None:
                    continue
                net = outcome.get("net_price")
                # Missing net_price is not proof of a zero fee.
                comparable = net if net is not None else (
                    raw if fee and fee["model"] == "included" else None)
                print(fixture["id"], book["key"], outcome["name"],
                      "raw", raw, "net", comparable,
                      "last_update", market.get("last_update"))

The script reads current NBA fixtures when you run it; it does not replay the October read. It prints an unknown comparison value when the fee fields cannot establish a net quote. Resolve those books' fee terms before sorting by price. An absent net_price alone cannot justify a zero-fee assumption.

Keep size kinds and timestamps attached

The recorded Thunder sizes were 8,807.4 at Kalshi and 134,170.59 at Polymarket, both resting_depth. FanDuel showed 12,097 as posted_limit. Resting depth is the amount at the quoted price; a posted limit is the book's stated maximum. Neither promises that an order will execute unchanged.

Read size.kind before using size.amount, and preserve an absent size as unknown. Keep the market's last_update next to the price. The timestamps in the table belong to the recorded response and are all UTC.

For fee arithmetic across repeated bets, see the long-term cost of exchange fees.

Questions

More from the blog