Correction (27 September 2026): earlier versions of this post said Kalshi doesn't produce a per-trade tape for tennis and esports, described the slices' coverage file as a hand-listed dark-window list, dated every table row to the Jul-19 launch vintage and implied trade prints for every captured family; the affected sentences now match the ZIPs on sale.
We sell a bundle called the Kalshi Microstructure Tape — L2 order-book depth across every Kalshi ticker family the recorder polled, plus deduplicated trade prints for the families whose prints it captured (15 of the 92 depth families in the 2026-09-06 vintage). Each cycle records the top 150 markets by 24-hour volume across all sports (top 50 before Jul 12), so the family mix varies by day. It's $199, and if you're doing cross-sport microstructure work it's the thing to buy.
But most researchers aren't cross-sport. If you're building a tennis model, the WNBA depth in that bundle is dead weight you paid for. So we've carved the tape into per-sport slices — $39 each, same live capture, same schema, and a coverage file measured from each slice's own rows. Buy the sport you actually study.
The six slices, with real numbers
The tennis, MLB, WNBA and esports counts below were measured against the Jul-19 2026 launch vintage; the NFL and college-football rows come from their current vintage, which ends 2026-09-12. Later vintages are larger, and the current counts are always on each product page and on /kalshi-historical-data. The archives are dated snapshots, not live feeds — capture continues and we cut refreshed vintages periodically. None of this can be rebuilt after the fact — Kalshi retains no order-book history, a wall we map vendor-by-vendor in our honest comparison of historical order-book data sources.
| Slice | What's in it | Depth rows | Trade tape |
|---|---|---|---|
| Tennis (ATP/WTA/ITF) | singles match + ATP game-spread books | 46.3M | — |
| MLB | game/spread/total/F5/team-total | 70.1M | 1.32M prints |
| WNBA | game markets | 8.1M | 713K prints |
| Esports (LoL/CS2) | match + map-total | 10.2M | — |
| NFL | game, spread, total, half and team-total markets | 26.2M | 1.06M prints (KXNFLGAME, KXNFLSPREAD, KXNFLTOTAL) |
| College football | game, spread, total and team-total markets | 31.9M | 1.34M prints (KXNCAAFGAME, KXNCAAFSPREAD, KXNCAAFTOTAL) |
Two things stand out. Tennis is the biggest depth-only slice — 46.3 million depth rows of ATP, WTA and ITF singles at launch — and it's the one sport with no off-the-shelf equivalent anywhere: Kalshi keeps no history, and the raw-L2 vendors that cover Polymarket don't carry Kalshi tennis. MLB pairs book and prints — 70M depth rows and 1.3M deduplicated moneyline trade executions with taker side at launch (WNBA pairs them too), so you can study fills against the ladder, not just quotes. MLB spread/total prints live in the full tape, not the MLB slice.
Two slices (tennis, esports) are depth-only because our recorder did not capture trade prints for those families. That is a limit of our capture, not a claim that Kalshi publishes no prints for those markets — we say so on the product page and in the ZIP rather than let you find out after paying.
Same honesty as the bundle
Every slice states its limits on its product page and in its ZIP, with the numbers measured from its own rows rather than maintained by hand: the 20-levels-per-side cap (Kalshi exposes more; we store the top 20); the selection rule (depth is the sport's share of a global top-150-by-24-hour-volume poll, top 50 before Jul 12, so lower-volume markets drop in and out); the measured median interval between observed snapshots, about 30.5 s before Jul 12 and about 43.2–43.3 s after (a measurement, not a fixed polling cadence; each slice's exact medians are in its manifest); every interval longer than 120 seconds between its observed depth timestamps (OBSERVED_GAPS.csv in the tennis, MLB, WNBA and esports slices, the derived gap table in COVERAGE.md for NFL and college football), which records gaps in our observations, not proof of recorder outages; and — for the tape-bearing slices — the fact that the per-print game state is capture-time, not event-time (join on created_time for event-time work). One overclaim costs more trust than ten disclosures, so the limits are on the label.
Try before you buy
Each slice has a free 400-row sample and the full schema on the samples page, no signup. For the recorded fields and how the slices fit together, see the Kalshi historical order-book data overview. Six $39 slices cost $234 separately; the $199 full tape costs less if you need all six. For a smaller selection, compare the actual prices and each product’s recorded vintage and coverage.
As always: research and backtest data, not a trading signal. Our own published work shows these markets are efficient — the value here is microstructure and execution research, not alpha.