Polymarket historical data: the full L2 order book archive

curl -H "X-API-KEY: $VIKE_API_KEY" \
  "https://data.vike.io/v1/polymarket/archive/datasets?limit=1000"

# Full history is the Parquet archive. Pick a date from manifest.json (no key needed),
# download the whole file, then read it locally:
DAY=YYYY-MM-DD
curl -H "X-API-KEY: $VIKE_API_KEY" -O \
  "https://data.vike.io/archive/venue=polymarket/date=$DAY/l1_quotes.parquet"

# Or the free sample — one complete market, no key:
curl -O https://data.vike.io/archive/samples/l1_quotes.parquet

import os, requests
import pyarrow.parquet as pq

H = {"X-API-KEY": os.environ["VIKE_API_KEY"]}
day = "YYYY-MM-DD"   # any date listed in https://data.vike.io/archive/manifest.json
url = f"https://data.vike.io/archive/venue=polymarket/date={day}/l1_quotes.parquet"

# Download the whole file once, then read it locally — a remote Parquet
# reader re-fetches the footer and column chunks on every query.
with requests.get(url, headers=H, stream=True, timeout=600) as r:
    r.raise_for_status()
    with open("l1_quotes.parquet", "wb") as f:
        for chunk in r.iter_content(1 << 20):
            f.write(chunk)

table = pq.read_table("l1_quotes.parquet")
print(table.num_rows, table.schema.names)

import polars as pl

df = pl.read_parquet("l1_quotes.parquet")   # downloaded first — see the cURL tab
df = df.with_columns(pl.from_epoch("ts", time_unit="ms").alias("time"))
spread = (
    df.with_columns((pl.col("ask") - pl.col("bid")).alias("spread"))
      .group_by("token_id")
      .agg(pl.len().alias("changes"), pl.col("spread").mean().alias("avg_spread"))
      .sort("changes", descending=True)
)
print(spread.head())

-- l1_quotes.parquet downloaded first (see the cURL tab), then read locally
SELECT token_id,
       count(*)          AS changes,
       avg(ask - bid)    AS avg_spread,
       epoch_ms(min(ts)) AS first_quote,
       epoch_ms(max(ts)) AS last_quote
FROM   'l1_quotes.parquet'
GROUP  BY token_id
ORDER  BY changes DESC
LIMIT  10;

# replace each {placeholder} with your own value
curl -H "X-API-KEY: $VIKE_API_KEY" \
  "https://data.vike.io/v1/polymarket/events?token_id={token_id}&from_ms={from_ms}&to_ms={to_ms}&event_type=price_change&limit=50"

# Full history is the Parquet archive. Pick a date from manifest.json (no key needed),
# download the whole file, then read it locally:
DAY=YYYY-MM-DD
curl -H "X-API-KEY: $VIKE_API_KEY" -O \
  "https://data.vike.io/archive/venue=polymarket/date=$DAY/book_events.parquet"

# Or the free sample — one complete market, no key:
curl -O https://data.vike.io/archive/samples/book_events.parquet

import os, requests
import pyarrow.parquet as pq

H = {"X-API-KEY": os.environ["VIKE_API_KEY"]}
day = "YYYY-MM-DD"   # any date listed in https://data.vike.io/archive/manifest.json
url = f"https://data.vike.io/archive/venue=polymarket/date={day}/book_events.parquet"

# Download the whole file once, then read it locally — a remote Parquet
# reader re-fetches the footer and column chunks on every query.
with requests.get(url, headers=H, stream=True, timeout=600) as r:
    r.raise_for_status()
    with open("book_events.parquet", "wb") as f:
        for chunk in r.iter_content(1 << 20):
            f.write(chunk)

table = pq.read_table("book_events.parquet")
print(table.num_rows, table.schema.names)

import polars as pl

df = pl.read_parquet("book_events.parquet")   # downloaded first — see the cURL tab
df = df.with_columns(pl.from_epoch("ts", time_unit="ms").alias("time"))
print(df.group_by("event_type").len().sort("len", descending=True))

-- book_events.parquet downloaded first (see the cURL tab), then read locally
SELECT event_type,
       count(*)              AS events,
       count(DISTINCT token_id) AS tokens
FROM   'book_events.parquet'
GROUP  BY event_type
ORDER  BY events DESC;