Data
Polymarket weather market data
Test a Polymarket weather bot against the real books of every temperature bucket of every city chain since July 2026. Marketlens holds 106,027 markets across 108 series, 103,264 of them resolved, with $333.6M of traded volume. Endpoint reference is in the docs.
Backtest a weather bot on real books
client.backtest replays every bucket of a city's day together, and ctx.books holds all of them, so a bot can read the whole distribution. Orders fill against the real book, with limit order queue priority and latency modeled, not at mid prices. See Polymarket backtesting for the fill model.
Marketlens holds the market side; forecasts are yours to bring. Many Polymarket data vendors stop at crypto and sports, and the comparison shows who carries what.
from datetime import datetime, timedelta, timezone
from marketlens import MarketLens, Strategy
class FavoriteBucket(Strategy):
def __init__(self):
self._bought = set()
def on_trade(self, ctx, market, book, trade):
# buy the top bucket once its midpoint passes 0.6
top = max(ctx.books.values(), key=lambda b: b.midpoint)
if top.midpoint > 0.6 and top.market_id not in self._bought:
ctx.buy_yes(size=100, market_id=top.market_id)
self._bought.add(top.market_id)
client = MarketLens()
end = datetime.now(timezone.utc) - timedelta(days=4)
result = client.backtest(
FavoriteBucket(), "nyc-daily-weather", initial_cash=10_000,
after=end - timedelta(hours=6), before=end,
)
print(result.summary())How weather markets work
Each city day is a chain of bucket markets: adjacent temperature ranges 1 degree Celsius or 2 degrees Fahrenheit wide, with open ended "or below" and "or higher" tails at the ends. Exactly one bucket resolves Yes when the day's reading is in. Because the buckets partition the outcome space, the set of bucket prices at any moment is a market implied probability distribution over the day's temperature, and its evolution as forecasts update is the interesting object, not any single price.
Marketlens records the full L2 order book of every bucket, with trades, candles, and the resolved outcome once the day settles (see resolution history for how outcomes are stored).
The end date quirk
Weather markets carry a listed end date of 12:00 UTC on the target day, but trading actually runs until midnight in the city's local time zone, and much of the volume trades after the listed end date. Collection that trusts the metadata stops hours early and misses the session where the answer becomes clear. Marketlens follows each chain through local midnight, so the late trading is in the archive. Chains also open the prior day, so the record captures forecast driven trading before the target day begins.
City coverage and access
Coverage spans city chains worldwide plus temperature index and precipitation series; the full list with per series numbers is in the weather category. A few of the most traded city series:
from datetime import datetime, timedelta, timezone
from marketlens import MarketLens
client = MarketLens()
# one day of the NYC chain, every bucket's book
end = datetime.now(timezone.utc) - timedelta(days=1)
for market, book in client.orderbook.walk(
"nyc-daily-weather",
after=end - timedelta(days=1),
before=end,
):
print(book.as_of, book.best_bid, book.best_ask)Latest chains in the snapshot run through November 2, 2026. The same data feeds the backtester for strategies trading the distribution against forecast updates.
FAQ
Common questions
Does Polymarket have weather markets?
Yes. Polymarket lists daily temperature markets for cities worldwide: each day's chain is a set of buckets (1 degree Celsius or 2 degrees Fahrenheit wide, with open ended tails) covering where the day's high or low temperature will land. Marketlens records 106,027 weather markets across 108 series.
Where can I get Polymarket weather market data?
Marketlens has collected order books, trades, candles, and resolved outcomes for Polymarket weather markets since July 2026: 106,027 markets, 103,264 resolved, $333.6M of traded volume. It is served through a REST API, a Python SDK (pip install marketlens), and bulk Parquet exports.
How do Polymarket temperature buckets work?
Each city day is split into adjacent temperature ranges, one market per bucket, with "or below" and "or higher" tails at the ends. Exactly one bucket resolves Yes. Read across all bucket prices at any moment and you get the market implied probability distribution over the day's temperature.
Why does trading continue after the listed end date?
Polymarket stamps weather markets with a 12:00 UTC end date, but trading actually runs until midnight in the city's local time zone, and much of the volume trades in that window. Marketlens collection follows markets through local midnight rather than stopping at the listed end date, so the late session is captured.
Is every temperature bucket collected, or only the likely ones?
Every bucket of every chain is collected, including tails trading near zero. Distribution work needs the whole set: the implied probability mass in unlikely buckets is exactly what tail studies and calibration analysis are about.
Can I backtest a Polymarket weather bot?
Yes. The Marketlens Python SDK replays every bucket of a city's day together from the recorded order books, across 103,264 resolved weather markets. Your bot reads the whole chain at once and its orders fill against the real book, not at mid prices.
Does Marketlens include weather forecasts?
No. Marketlens holds the market side: order books, trades, candles, and resolved outcomes. Bring your own forecast series and join it to the market data on time.
Try it
Backtest your weather bot on last week's chains
The free tier includes 2M data rows and every market open in the last 7 days, no card required.
$ pip install marketlens