Schema doc shipped inside top_of_book_archive.zip (member top_of_book_archive/SCHEMA.md, 2,863 bytes, dated 2026-06-21), published unchanged.
Erratum (2026-09-28): the closing "Monthly drops from May 2026 onward" line describes a product that has not been released; no monthly drop was on sale on this date.
Erratum (2026-09-28, from the delivered Parquet files): integer columns are int64 and float columns float64, not int32/float32; Kalshi game_state values are pregame, live, late_game, final, unknown and status_delayed (no in_progress; unknown is 27.1% of rows); the 25- and 11-column counts exclude the datetime_utc convenience column (26 and 12 delivered).
Erratum (2026-09-28, from the delivered Parquet files): not every other family spans the full Mar 28 – Apr 30 window; families whose season or listings start or end inside it cover only their own dates (for example NCAAMB spread/total end Apr 7 and NCAAWB spread/total Apr 5; MLB team totals start Mar 31, NBA second-half markets Apr 14-15 and several soccer leagues Mar 31 – Apr 1; all UTC).

# Top-of-Book Tape — Schema

## Known gaps

- The men's & women's college-basketball moneyline (KXNCAAMBGAME / KXNCAAWBGAME) ends ~Apr 3-4, so the early-April NCAA tournament championship team-to-win markets are **not captured** (their spread & total are). All other sports moneyline/spread/total families span the full Mar 28 – Apr 30 window.

This bundle contains TWO Parquet files with intentionally different schemas.

## kalshi_snapshots.parquet (25 columns, full archive parity)

Same column layout as the Polymarket & Kalshi Orderbook Archive's Kalshi side.
Top-of-book quotes plus full game-state synchronization.

| Column | Type | Notes |
|---|---|---|
| timestamp | float64 | Unix epoch seconds, UTC |
| datetime_utc | timestamp | Convenience UTC datetime |
| ticker | string | Kalshi market ticker |
| title | string | Market question text |
| game_id | string | ESPN/source game ID |
| home_team, away_team | string | Team abbreviations |
| home_score, away_score | int32 | Live score at snapshot time |
| score_diff | int32 | home_score - away_score |
| period | int32 | Quarter/half/inning |
| time_remaining | string | Display clock |
| game_state | string | pregame / in_progress / final |
| yes_bid, yes_ask | float32 | Top-of-book quotes |
| mid, spread, spread_cents | float32 | Derived |
| yes_bid_size, yes_ask_size | int64 | Top-of-book sizes |
| min_size | int64 | Minimum tradable size |
| open_interest | int64 | Total OI |
| volume_24h | int64 | Trailing 24h volume |
| liquidity_score | float32 | Computed liquidity metric |
| tradable_now | bool | Market tradable at snapshot time |
| is_burst_mode | bool | High-frequency capture flag |

## poly_topofbook.parquet (11 columns, top-of-book only)

The Polymarket recorder during this window captured top-of-book only — no
orderbook depth, no level counts, no metadata fields beyond sport + event.

| Column | Type | Notes |
|---|---|---|
| timestamp | float64 | Unix epoch seconds, UTC |
| datetime_utc | timestamp | Convenience UTC datetime |
| token_id | string | Polymarket CLOB token (YES side) |
| sport | string | NBA, NHL, MLB, NCAAMB, NFL, etc. |
| event | string | Event title |
| outcome | string | Outcome label this token represents |
| bid, ask | float32 | Top-of-book in dollars (0.0–1.0) |
| mid, spread | float32 | Derived |
| bid_size, ask_size | float32 | Top-of-book sizes (shares) |

### NOT included on the Polymarket side

These columns from the original archive are **not in this dataset**:
`condition_id`, `event_slug`, `question`, `market_type`, `line`,
`bid_depth_5c`, `ask_depth_5c`, `total_bid_depth`, `total_ask_depth`,
`bid_levels`, `ask_levels`, `volume_24h`, `liquidity`, `is_burst`.

If you need depth, see:
- The Polymarket & Kalshi Orderbook Archive (Dec 2025 – Jan 2026)
- Monthly drops from May 2026 onward (full archive schema restored)