Betting Splits
Get public betting splits (handle % and bet %) from DraftKings, Circa Sports and BetMGM.
Each row covers one event at one source. It gives the share of tickets (bets_pct) and, where the source publishes it, the share of money (handle_pct) on each side of each market it covers: moneyline, spread and total. On soccer, a moneyline can have three sides (home, draw and away): see Three-way moneyline below. Use it to see which side the public is backing, and where the money is heavier than the ticket count. To see how one event’s splits moved over time, use the history endpoint described in Splits History below.
GET /api/v1/splitsAuthenticationPermalink for this section
Requires API key. Requires Pro tier ($229/mo) or higher.
What are betting splits?Permalink for this section
Handle % is the percentage of total money wagered on each side. Bet % is the percentage of total tickets (bets placed) on each side.
The gap between bet % and handle % reveals sharp money. If 30% of tickets carry 60% of the money, sharp bettors are on that side.
Data SourcesPermalink for this section
| Source | Type | What it carries | Update Frequency |
|---|---|---|---|
| DraftKings | Recreational book (~35% US market share) | Handle % and bet % | Every 5 minutes |
| Circa Sports | Sharp-friendly book (attracts professionals) | Handle % and bet % | A few times a day |
| BetMGM | Recreational book — derived from its own public bet-percentage field | Bet % only — handle_pct members are null | Intermittent |
Comparing DraftKings (recreational) vs Circa (sharp) splits reveals where professional money diverges from the public.
Sportsbook coverage is fixed at these three books. It does not grow with your plan’s sportsbook allowance — selecting more books, or upgrading to a higher tier, adds no further splits sources. The only other rows are consensus rows, described below. Because BetMGM carries a ticket percentage only, the bet %/handle % gap that reveals sharp money cannot be computed on BetMGM rows. History includes DraftKings, Circa, BetMGM and consensus samples captured during the retained 48-hour window. BetMGM publishes ticket percentages only; its handle percentages remain null.
consensus is a label, not a sportsbook. When every source covering an event posts identical numbers, those rows are merged into a single row carrying "sportsbook": "consensus". The API applies it synthetically — it never appears in /api/v1/sportsbooks, though /splits?sportsbook=consensus does filter for it. Merged rows are not returned by a sportsbook=draftkings filter. Soccer consensus rows are not merged book rows: they carry market-wide public betting percentages, not one sportsbook’s own, and they are the only rows with a draw.
Soccer coveragePermalink for this section
Soccer splits cover eight competitions. Filter for them with sport=soccer, or with league and the ids below.
| Competition | league | consensus (three-way) | draftkings (two-way) |
|---|---|---|---|
| Premier League | england_-_premier_league | Yes | Yes |
| La Liga | spain_-_la_liga | Yes | — |
| Bundesliga | germany_-_bundesliga | Yes | — |
| Serie A | italy_-_serie_a | Yes | — |
| Ligue 1 | france_-_ligue_1 | Yes | — |
| MLS | usa_-_major_league_soccer | Yes | Yes |
| Champions League | uefa_-_champions_league | Yes | Yes |
| Europa League | uefa_-_europa_league | Yes | Yes |
consensusrows carry bet % and handle % for the three-way moneyline (home, draw, away) and for total goals (over/under). They carry no spread.draftkingsrows carry DraftKings’ moneyline split between the two teams only. DraftKings shows no draw in its soccer splits, so these rows are two-way ("outcomes": 2, nodraw): they are not a 1X2 split. They carry no spread and no total.
Query ParametersPermalink for this section
| Parameter | Type | Description |
|---|---|---|
sport | string | Filter by sport (comma-separated). Example: basketball. Sport ids only; a league id such as nba goes in league (details). |
league | string | Filter by league (comma-separated). Example: nba,ncaab. The soccer league ids are listed under Soccer coverage above, e.g. england_-_premier_league. |
sportsbook | string | Filter by splits source. Example: draftkings,circa. Also accepts the synthetic consensus value. |
event_id | string | Filter by canonical event ID (comma-separated) |
market | string | Filter to events that carry a given split market (comma-separated). One or more of spread, total, moneyline. |
limit | integer | Max results (default 100, max 200) |
offset | integer | Pagination offset (default 0) |
ResponsePermalink for this section
{
"data": [
{
"event_id": "mlb_guardians_orioles_2026-04-16",
"sport": "baseball",
"league": "mlb",
"sportsbook": "draftkings",
"away_team": "Baltimore Orioles",
"home_team": "Cleveland Guardians",
"spread": {
"away_line": -1.5,
"home_line": 1.5,
"away_odds": -1.5,
"home_odds": 1.5,
"handle_pct": { "away": 0.22, "home": 0.78 },
"bets_pct": { "away": 0.20, "home": 0.80 },
"updated_at": "2026-04-16T19:12:40.118204+00:00"
},
"total": {
"line": 8,
"handle_pct": { "over": 0.53, "under": 0.47 },
"bets_pct": { "over": 0.57, "under": 0.43 },
"updated_at": "2026-04-16T19:20:02.553917+00:00"
},
"moneyline": {
"outcomes": 2,
"away_odds": 104,
"home_odds": -126,
"handle_pct": { "away": 0.28, "home": 0.72 },
"bets_pct": { "away": 0.33, "home": 0.67 },
"updated_at": "2026-04-16T13:48:51.640275+00:00"
},
"fetched_at": "2026-04-16T19:25:28.363825+00:00",
"available_metrics": ["bets_pct", "handle_pct"]
}
],
"pagination": {
"limit": 100,
"offset": 0,
"count": 1,
"total": 41,
"has_more": false,
"next_offset": null
},
"updated_at": "2026-04-16T19:29:38.920698424Z"
}Read the spread line (the handicap, e.g. -1.5 / +1.5) from spread.away_line / spread.home_line, the same names the historical endpoint uses. spread.away_odds / spread.home_odds carry the same value as a legacy alias: they are the line, never a price. No splits source publishes a spread price.
Response FieldsPermalink for this section
| Field | Type | Description |
|---|---|---|
event_id | string | Canonical event ID — use this to join with /odds data |
sport | string | Atlas-normalized sport name |
league | string | Atlas-normalized league name |
sportsbook | string | Source sportsbook for the splits data (draftkings, circa, betmgm, or the synthetic consensus) |
away_team | string | Away team name |
home_team | string | Home team name |
spread.away_line | number | Away spread line (e.g., -1.5) |
spread.home_line | number | Home spread line (e.g., +1.5) |
spread.away_odds | number | Legacy alias: the same value as spread.away_line (the line, not a price) — see callout above |
spread.home_odds | number | Legacy alias: the same value as spread.home_line (the line, not a price) |
spread.handle_pct | object | Money % on each side (away, home; 0.0-1.0) |
spread.bets_pct | object | Ticket % on each side (away, home; 0.0-1.0) |
total.line | number | Over/under line (e.g., 225.5; on soccer, total goals, e.g. 2.5) |
total.handle_pct | object | Money % (over, under; 0.0-1.0) |
total.bets_pct | object | Ticket % (over, under; 0.0-1.0) |
moneyline.outcomes | integer | How many outcomes the moneyline market has: 3 for a three-way (home/draw/away) market, 2 for every other moneyline. On every moneyline object — see Three-way moneyline below |
moneyline.away_odds | number | Away moneyline odds (American format) |
moneyline.home_odds | number | Home moneyline odds (American format) |
moneyline.draw_odds | number | Draw odds (American format). Only when outcomes is 3 |
moneyline.handle_pct | object | Money % on each side (away, home, and draw when outcomes is 3; 0.0-1.0) |
moneyline.bets_pct | object | Ticket % on each side (away, home, and draw when outcomes is 3; 0.0-1.0) |
spread.updated_at, total.updated_at, moneyline.updated_at | string | When that market’s numbers last changed (ISO 8601): usually the time SharpAPI first saw the new numbers, normally within a few minutes of the change, or, for some sources, the change time the source itself reports. null or absent when no change time is known, for example when the market has no numbers — see Freshness below |
fetched_at | string | When SharpAPI last polled the source for this row (ISO 8601). It moves forward on every poll, even when the source’s numbers have not changed, so it does not say how current they are: read updated_at on each market — see Freshness below |
available_metrics | array | Which split metrics this row’s sportsbook publishes at all — bets_pct, handle_pct, or both, always in that order. A statement about the book, not about this row’s values. Omitted when the book is undeclared — see below. |
event_uuid | string | The event’s stable UUID, the same value as event_uuid on /odds. Present only when the row matches exactly one event — see below. |
canonical_event_id | string | The matched event’s ID, as /events serves it in id and /odds in event_id. Usually identical to event_id. Present only when event_uuid is. |
event_start_time | string | The matched event’s start time (ISO 8601), the same value as start_time on /events. /odds rows carry the same start in event_start_time, but not always in the same format (it can omit the seconds), so compare the two as timestamps, not as strings. Present only when the row matches exactly one event — see below. |
exchanges | object | Buyer flow on prediction-market exchanges for this game. Not a betting split. Present only when an exchange shows buyer flow for the game, absent otherwise — see Exchange buyer flow below. |
Which values can be null, and which keys can be absent. The percentage keys and the moneyline odds keys above are always present (draw and draw_odds only on a three-way moneyline), but some carry null instead of a number. Other keys can be missing rather than null: total.line is omitted on an event for which the book has posted no total, spread.away_line / .home_line and their aliases spread.away_odds / .home_odds are omitted on a betmgm row that has no spread line, and available_metrics is omitted for a book SharpAPI has not declared — read these with a default rather than assuming the key.
handle_pct.away / .home (and .over / .under on totals) are null on every betmgm row, because BetMGM publishes a ticket percentage only — there is no handle figure to report. bets_pct members are null when the source has posted no percentage for that market yet; BetMGM’s feed is intermittent, so this shows up there. moneyline.away_odds / .home_odds are null when a book has pulled the moneyline for an event while still publishing its split percentages — this happens on DraftKings and Circa rows too, not only BetMGM. The OpenAPI spec declares all of these as nullable, so a generated client will accept them. spread.away_line / .home_line (and spread.away_odds / .home_odds) are null when the source has posted no spread line for the game, and that spread’s bets_pct and handle_pct members are then null too. On a three-way moneyline, draw_odds can also be null, and each share object holds either three numbers or three nulls. A soccer row carries no spread: every spread field is null.
event_uuid, canonical_event_id and event_start_time are present only when the row matches exactly one event. SharpAPI matches each row’s event_id to an event. When the match is ambiguous, the row omits all three keys rather than guess: for example, a doubleheader row whose event_id has no _bN suffix could be either game. A row whose event is not found omits them too. event_start_time can also be absent on a matched event that has no start time on record, the same case in which /events omits start_time. Read these keys with a default.
The event_id field uses the same canonical ID format as the /odds endpoint, so you can join splits with odds data directly.
For example, fetch odds for a specific game and compare with its splits:
GET /api/v1/odds?event_id=nba_thunder_timberwolves_2026-03-15
GET /api/v1/splits?event_id=nba_thunder_timberwolves_2026-03-15/odds does not include per-row public bet %.
Betting splits (handle % and ticket %) are only available through this endpoint (/splits) and its historical counterpart (/splits/history). There is no public_bet_pct field on /odds rows.
Freshness (updated_at)Permalink for this section
fetched_at is when SharpAPI last polled the source for the row. It moves forward on every poll, even when the source has stopped changing its numbers, so a recent fetched_at does not mean the numbers are current. To judge how current they are, read updated_at on each market: it is when that market’s numbers last changed. Usually that is the time SharpAPI first saw the new numbers, normally within a few minutes of the change; for some sources it is the change time the source itself reports. Compare it with the time of your request to get the age.
In the example above, fetched_at is four minutes before the response, but the moneyline last changed at 13:48, more than five hours earlier.
A market whose updated_at is null or missing has an unknown age. Read that as age unknown, not as current.
Rows are served whatever the age of their numbers. updated_at does not say why the numbers have not changed: a source that has stopped updating and a split that has held steady look the same.
Three-way moneyline (outcomes)Permalink for this section
Every moneyline object carries outcomes, the number of outcomes its market has. Branch on outcomes, not on the sport: a soccer moneyline can be two-way too (see the DraftKings row below).
outcomes | Market | Keys in bets_pct / handle_pct | Prices |
|---|---|---|---|
3 | Home / draw / away (1X2): soccer consensus rows | away, home, draw | away_odds, home_odds, draw_odds |
2 | Every other moneyline, DraftKings soccer rows included | away, home | away_odds, home_odds |
The three shares of a three-way object add up to 1, give or take rounding: in the example below, the money shares add up to 0.99. A three-way share object is served whole or not at all. When any of its shares is missing, or the three do not add up to 1 within 0.02, it is served as { "away": null, "home": null, "draw": null } and outcomes stays 3. It is never cut down to two numbers, so a moneyline with two shares is never a 1X2 split. The exchanges block carries no outcomes.
A soccer consensus row, three-way:
{
"event_id": "germany_-_bundesliga_eintrachtfrankfurt_hoffenheim_2026-11-21_b2",
"sport": "soccer",
"league": "germany_-_bundesliga",
"sportsbook": "consensus",
"away_team": "TSG Hoffenheim",
"home_team": "Eintracht Frankfurt",
"spread": {
"away_line": null,
"home_line": null,
"away_odds": null,
"home_odds": null,
"handle_pct": { "away": null, "home": null },
"bets_pct": { "away": null, "home": null }
},
"total": {
"line": 3.5,
"handle_pct": { "over": 0.90, "under": 0.10 },
"bets_pct": { "over": 0.90, "under": 0.10 }
},
"moneyline": {
"outcomes": 3,
"away_odds": 160,
"home_odds": 128,
"draw_odds": 290,
"handle_pct": { "away": 0.15, "home": 0.70, "draw": 0.14 },
"bets_pct": { "away": 0.24, "home": 0.69, "draw": 0.07 }
},
"fetched_at": "2026-11-20T22:32:18.402771+00:00",
"available_metrics": ["bets_pct", "handle_pct"]
}A DraftKings soccer row, two-way (its spread and total are left out here):
{
"event_id": "uefa_-_champions_league_astonvilla_galatasaray_2026-11-24_b2",
"sport": "soccer",
"league": "uefa_-_champions_league",
"sportsbook": "draftkings",
"away_team": "Aston Villa",
"home_team": "Galatasaray",
"moneyline": {
"outcomes": 2,
"away_odds": 145,
"home_odds": 155,
"handle_pct": { "away": 0.62, "home": 0.38 },
"bets_pct": { "away": 0.82, "home": 0.18 }
}
}DraftKings shows its soccer split for the two teams only. away_odds and home_odds are its prices for each team to win, and the two shares in each of bets_pct and handle_pct add up to 1. It is not a 1X2 split: don’t compare its two shares one-for-one with the three shares of a consensus row. Every spread and total field on these rows is null: DraftKings’ soccer split covers the moneyline only.
Reading available_metricsPermalink for this section
A splits row may carry available_metrics, an array naming the split metrics that this row’s sportsbook publishes at all. It describes the book, not the row: a book that publishes money share still lists handle_pct on a row whose handle_pct members happen to be null.
Only two names can appear, always in this order, and each is exactly a key inside the spread, total and moneyline objects — so a name here points at a field you can look up:
bets_pct— share of tickets (bets placed), 0.0-1.0handle_pct— share of money (handle), 0.0-1.0
The field is omitted entirely when SharpAPI has not declared what a book publishes. Absent means not stated — never publishes nothing.
Read the list together with the value. Taking handle_pct as the example:
available_metrics | handle_pct members | What it means |
|---|---|---|
present, lists handle_pct | numbers | The book publishes money share and we have it. |
present, lists handle_pct | null | The book publishes money share and we are missing it — a gap on our side, not a limit of the book. |
present, does not list handle_pct | null | The book does not publish money share. Nothing is broken. |
present, does not list handle_pct | numbers | Should not occur; treat it as a declaration that has fallen behind and trust the value. |
| absent | numbers or null | Nothing is stated about this book. Do not infer either way — read the values on their own, and do not record the book as publishing nothing. |
A merged consensus row carries the intersection. A merged "sportsbook": "consensus" row folds several books together, so its claim has to hold for every one of them: it lists only the metrics all of its contributing books declare, and omits the field entirely if any one of them is undeclared.
Exchange buyer flow (exchanges)Permalink for this section
A row may carry an exchanges object: buyer flow for the same game on prediction-market exchanges. It shows which side buyers took on each exchange, as a share of the money and as a share of the trades. It is not a betting split: it comes from no sportsbook, carries no sportsbook field, and is never a row of its own.
The object is present only when at least one exchange shows buyer flow for the game, and absent otherwise, so read it with a default. When an event has more than one row (books that disagree), every row of that event carries the same exchanges object.
{
"event_id": "nba_celtics_knicks_2026-11-02",
"sportsbook": "consensus",
"exchanges": {
"fetched_at": "2026-11-02T12:00:01+00:00",
"moneyline": [
{
"exchange": "kalshi",
"line": null,
"money_pct": { "away": 0.06, "home": 0.94 },
"trades_pct": { "away": 0.61, "home": 0.39 },
"usd_traded": 2759,
"trade_count": 41,
"light_trading": true,
"all_lines": false,
"updated_at": "2026-11-02T11:59:00+00:00"
}
],
"spread": [
{
"exchange": "kalshi",
"line": { "away": -3.5, "home": 3.5 },
"money_pct": { "away": 0.55, "home": 0.45 },
"trades_pct": { "away": 0.48, "home": 0.52 },
"usd_traded": 18420.5,
"trade_count": 212,
"light_trading": false,
"all_lines": false,
"updated_at": "2026-11-02T11:59:00+00:00"
}
],
"total": [
{
"exchange": "prophetx",
"line": 221.5,
"money_pct": { "over": 0.7, "under": 0.3 },
"trades_pct": { "over": null, "under": null },
"usd_traded": null,
"trade_count": null,
"light_trading": false,
"all_lines": true,
"updated_at": "2026-11-02T11:58:30+00:00"
}
]
}
}The other row fields are as in the Response example above.
| Field | Type | Description |
|---|---|---|
exchanges.fetched_at | string | When the exchange flow for this game was read (ISO 8601) |
exchanges.moneyline / .spread / .total | array | One entry per exchange showing flow on that market. A market with no entries is omitted. |
exchange | string | The exchange: kalshi, polymarket, novig or prophetx |
line | number, object or null | The line the flow is for. Always null on moneyline. A number on total (e.g. 221.5). An object with away and home on spread (e.g. -3.5 / 3.5). Also null when the exchange’s line is not known. |
money_pct | object | Share of the money that bought each side (0.0-1.0). Keys are away / home on moneyline and spread, over / under on total. Both members are numbers. |
trades_pct | object | Share of the trades that bought each side (0.0-1.0), with the same keys as money_pct. Both members are null when the exchange reports no trade split; one is never null without the other. |
usd_traded | number or null | Money traded, in US dollars. null when not reported. |
trade_count | integer or null | Number of trades. null when not reported. |
light_trading | boolean | true when the sample is thin: the entry is shown, but do not lean on its shares. |
all_lines | boolean | true when the volume is matched across every line of that market on the exchange, not only the line shown in line. |
updated_at | string | When this exchange’s reading was taken (ISO 8601). Use it to judge how old the reading is. |
Exchange shares are fractions from 0 to 1, the same scale as the split fields. money_pct and trades_pct read like handle_pct and bets_pct: 0.94 means 94%. The block reports buyer flow only; it carries no open interest.
ExamplesPermalink for this section
All NBA Splits
curl "https://api.sharpapi.io/api/v1/splits?league=nba" \
-H "X-API-Key: YOUR_API_KEY"Splits HistoryPermalink for this section
Track how splits shift over time for a specific event.
GET /api/v1/splits/history?event_id={event_id}Query ParametersPermalink for this section
| Parameter | Type | Required | Description |
|---|---|---|---|
event_id | string | Yes | Canonical event ID |
sportsbook | string | No | Filter by book (comma-separated). Applied before pagination. |
start_time | string | No | Lower bound. RFC 3339 (2026-04-16T13:00:00Z) or Unix seconds (1776344602). |
end_time | string | No | Upper bound, same formats. |
limit | integer | No | Max entries (default 100, max 200). |
cursor | string | No | Opaque continuation token from meta.next_cursor; keep event, book and time filters unchanged. |
The retained window is 48 hours only. There is no longer-term archive or backfill beyond captured, retained samples. Follow meta.next_cursor while meta.has_more is true to read the complete window. meta.total, books, oldest, and newest describe the current page; meta.limit is the effective page size (default 100, maximum 200; larger values clamp to 200). Use the same end_time on every page for a fixed time range. Invalid timestamps or mismatched cursors return 400; temporary history-store failures return 503.
ResponsePermalink for this section
Entries are sorted oldest-first. This endpoint emits the success envelope (success/data/meta) — distinct from /splits which emits data/pagination/updated_at.
The history payload uses book (not sportsbook), a known inconsistency vs /splits. Its spread carries away_line/home_line, as /splits does, but not the legacy away_odds/home_odds alias. Each entry also carries available_metrics, with the same meaning as on /splits: the metrics that entry’s book publishes at all. It is omitted for a book SharpAPI has not declared.
Each entry’s moneyline has the same shape as on /splits: outcomes on every moneyline object, and draw and draw_odds on a three-way entry. Soccer’s three-way entries are consensus entries ("book": "consensus").
History values may be unavailable. Percentage members can be null; BetMGM never publishes handle percentages. Line and odds fields can be absent or null when unavailable. Check available_metrics for the metrics the book publishes, and handle missing/null values when reading individual markets.
{
"success": true,
"data": [
{
"available_metrics": ["bets_pct", "handle_pct"],
"book": "circa",
"ts": "2026-04-16T13:03:21.071966+00:00",
"timestamp": 1776344602.36,
"spread": {
"away_line": -1.5,
"home_line": 1.5,
"handle_pct": { "away": 0.37, "home": 0.63 },
"bets_pct": { "away": 0.27, "home": 0.73 }
},
"total": {
"line": 8,
"handle_pct": { "over": 0.43, "under": 0.57 },
"bets_pct": { "over": 0.55, "under": 0.45 }
},
"moneyline": {
"outcomes": 2,
"away_odds": 104,
"home_odds": -126,
"handle_pct": { "away": 0.35, "home": 0.65 },
"bets_pct": { "away": 0.35, "home": 0.65 }
}
}
],
"meta": {
"event_id": "mlb_guardians_orioles_2026-04-16",
"total": 1,
"books": ["circa"],
"limit": 100,
"has_more": false,
"next_cursor": "",
"oldest": "2026-04-16T13:03:22.360588312Z",
"newest": "2026-04-16T13:03:22.360588312Z",
"updated_at": "2026-04-16T19:28:50.525875452Z"
}
}Data is collected every ~5 minutes and retained for 48 hours via a Valkey sorted set (splits_history:{event_id}) scored by Unix timestamp.
Full History
curl "https://api.sharpapi.io/api/v1/splits/history?event_id=nba_thunder_timberwolves_2026-03-15" \
-H "X-API-Key: YOUR_API_KEY"Interpreting SplitsPermalink for this section
| Signal | What it means |
|---|---|
| Bet % high, Handle % low | Public side — lots of small bets |
| Bet % low, Handle % high | Sharp side — fewer but larger bets |
| DK and Circa agree | Market consensus — both public and sharp aligned |
| DK and Circa diverge | Sharp-public split — Circa (sharp) disagrees with DK (public) |
Sportsbook splits come from three books only — DraftKings and BetMGM (recreational) and Circa (sharp-adjacent). The only other rows are consensus rows (see Data Sources). No true sharp book publishes splits. Use splits as one signal alongside line movement and +EV analysis, not in isolation.