Esports Match Markets: Maps, Handicaps and Best-of Series as Labels

Esports markets look like sports markets and settle like them, but the object being priced is layered: a match is a series of maps, each map a game with its own winner, its own kill or round count and its own handicap. The venues price every layer. That makes an esports match the densest source of correlated labels in the prediction-market archive — and the easiest place to leak a backtest by treating those labels as independent. This guide lays out the vocabulary per title and venue, how a best-of series decomposes into labels, which contracts are spreads and totals in disguise, where in the file the information arrives, and how to split the data so that a model trained on one tournament is tested on another.

Everything below runs on the CryptoStruct archive: every message Kalshi, Polymarket and 35+ crypto venues publish — every Level-2 snapshot and update at full depth, every trade, every contract, every day since we added the venue — captured co-located with nanosecond venue and receive timestamps and a gap-audited event-id chain, in one normalized schema. It is the same capture our own trading engine and enterprise feeds run on, sold as €1 day bundles you buy as a guest in the Data Shop with instant download — no subscription, no minimum, no sales call. Free full-day samples let you run every command in this guide before paying.

Match, map, game — the vocabulary per title

In League of Legends a match is a best-of-one, -three or -five of games; Polymarket slugs say game1, game2, game3 (lol-kc-mvk-2026-10-18-game2) and the bare slug (lol-kc-mvk-2026-10-18) is the series winner. In Counter-Strike 2 and Valorant the unit is a map, and a best-of-three has a map-1, map-2 and map-3 winner plus the series winner; Polymarket keeps the game token for maps too (cs2-g2a-66661-2026-10-17-game1). Dota 2 follows the LoL pattern. On Kalshi the layers are separate series: League of Legends Game for the series result, Counter-Strike 2 Map Winner and Valorant Map Winner for individual maps, and total-maps series for the length of a best-of. predict.fun codes name the layer in the description part (lol-rb-mr-2026-10-27_game-3-winner_…).

The contract identifiers are therefore the first feature you parse: title, the two teams as the venue's codes, the event date and the layer. The series pages render that reading next to every identifier in their contract tables, and the same parse tells you which contracts of a day belong to one match.

A best-of series as label structure

A best-of-three yields up to four settled contracts: the series winner and one per map played. Each is a label in the same sense as a game moneyline — the last recorded print at or above 0.97 reads as Yes in our statistics, at or below 0.03 as No, always labelled implied, the venue's official settlement being the authoritative record. But the labels are not independent. The team that wins map one is more likely to win the series; the series contract reprices sharply on each map result; and the map-3 contract exists only if the series goes the distance, so its presence is itself information about the earlier maps.

For training this means the match is the example and the maps are its parts. Count labels per match when you size a dataset — a tournament weekend with forty best-of-threes is forty examples with up to 160 labels, not 160 examples — and never let two maps of one series fall on different sides of a train/test split. The upside is real: a map contract's whole book path from listing to settlement is short (typically a day or less in our 2026-10 measurement of the top-10 sample), so the features per label are compact and the examples accumulate quickly across a season of tournaments.

Map handicaps and kill or round totals

Beyond winners, the venues list two families that mirror traditional sports lines. A map handicap (-game-handicap-away-1pt5, -game1-round-handicap-<team>-6pt5) asks whether a team wins the series or a map by more than a line — a spread. A kill total in LoL or Dota (-game3-kill-over-45pt5) or a round total in CS2 and Valorant asks whether the count clears a number — a total. The 0..1 price, the book and the settlement work exactly as for a moneyline; what changes is the label noise, because a total settles on a count that can land on the line and a handicap on a margin.

Between maps: where the book moves

The most informative minutes of an esports series-day are the breaks between maps. A map ends; its contract runs to 0.99 or 0.01 within seconds; the series contract reprices to the new state of the best-of; and the next map's contract, which has been quoted thinly, suddenly carries the market's updated view of both teams. The day file records all three books through that transition with nanosecond timestamps — the depth pulled on one side before the official result is in, the spread widening and collapsing, the taker-buy share of the prints that drove the move. The contract tables on each series page show the open, high, low and close of a map contract on its day; a map that opened at 0.74 and closed at 0.999 spent most of its life somewhere between, and the file is where.

Features that follow from this: time since the previous map's settlement, the series contract's move on that settlement, the next map's book depth and spread at the moment the previous map settled, and the lag between the map result in the market and the venue's listed settlement. In-play features inside a map — kills, objectives — are not in the market data; they are an exogenous join you align by timestamp, exactly as polls are joined to election books in the election-markets guide.

Venue structure

Polymarket lists one family per title — LoL, CS2, Dota 2, Valorant — with every layer of every match inside it; in our 2026-10 window the Polymarket CS2 and LoL families were among the most traded sports families in the archive. predict.fun mirrors that structure with its own codes. Kalshi splits the layers into series — game, map, total maps — so a Kalshi series-day is one layer across every match of the day, and a reader who wants the whole match buys the same day across the sibling series. OG.com lists esports as one family. The esports topic hub groups all of them, and each title has its own hub beneath it.

Labels and leakage

Split by tournament and series, never at random. Maps of one best-of share form, draft and the same crowd; matches of one tournament share the patch, the format and often the venue. A random split over map contracts puts map two of a series in training and map three in test and learns the series result backwards. The honest protocol walks forward over events: train on tournaments that have finished, validate on the next one, purge any series still open at the cut-off, and report per tournament. Patches are regimes — a model trained before a major balance change has not seen the game it is tested on.

Data plan

1. Pick a title and the layer you can label — series winners first, maps and totals once the match parser groups contracts correctly.
2. Download the free sample contract of the series page and the full-day samples; parse the identifiers into title, teams, date and layer before anything else.
3. Buy tournament days as ranges with /shop/bundles?bundle=<series-key>&d=<from>..<to>; on Kalshi, buy the same days across the game, map and total-maps series.
4. Group contracts into matches, count matches as examples, and build the between-map features from the three books of a transition.
5. Walk forward over tournaments, purge open series, treat patches as regimes, and report per event.

Coverage

Everything is sold the same way as every other prediction series: €1 per series-day, no subscription, no minimum, no account — a series-day is every contract of the family listed on that UTC day.

Limitations

The statistics on series pages cover the ten most traded contracts per day — the bundle holds every contract, the statistics do not. Measurements quoted here are from 2026-10 and change as the archive grows. Polymarket files between 2025-09-16 and 2026-01-08 are affected by an adapter fault that left about half of them without order-book deltas (trades are complete); those days are sold as-is and marked in the shop. Kalshi trades between 2026-02-26 and 2026-03-10 were recorded twice by the first recorder — de-duplicate on the trade id. All venues have a late start on 2026-06-17. Map and series labels in our statistics are implied from the last recorded print, never the venue's official settlement. In-game events are not part of the market data. Venue APIs expose some historical market and trade data, but not the tick-by-tick book path — check the current docs. Nothing here is betting or investment advice.