Arbitrage, middles and +EV

Lists of arbitrage, low holds, middles and positive expected value prices, computed from the prices we record, ready to check and act on.

Updated

These routes need Pro or above. Other keys get 403 Plan restriction ("This endpoint requires Pro access."). Prices move: check the price at the bookmaker before acting on any entry.

What each list is

  • Arbitrage. Two prices at two different bookmakers that cover both sides of one market, so backing both in the listed proportions returns a profit whichever side wins.
  • Low holds. The best price on each side of one market at two different bookmakers, where the two together carry almost no margin. Useful when you want tight markets, not only outright arbitrage.
  • Middles. Two bets on different lines of one market at two bookmakers, such as over 44.5 at one and under 47.5 at the other, so both win when the result lands in the gap.
  • Positive expected value (+EV). One bookmaker's price that is longer than the fair price of the outcome. An estimate, not a guarantee: a fair price is a model of the market, and prices move.

Every list is computed from the prices SockOdds records, for upcoming events only. Only bookmakers your key may read are used, so an entry never names a bookmaker you cannot see.

The routes

RouteWhat it returnsSorted by
GET /v1/opportunities/arbsArbitrage pairs.Best profit_pct first
GET /v1/opportunities/low-holdsPairs with a small combined margin.Lowest hold_pct first
GET /v1/opportunities/middlesPairs on two lines with a gap.Largest middle_profit_pct first
GET /v1/opportunities/evSingle prices above the fair price.Highest ev_pct first

All four take the same filters:

  • league: one league key, such as nba. Leave it out to list every league.
  • bookmakers: up to 20 bookmaker keys, comma separated. An entry is kept only when every price in it is at a named bookmaker.
  • markets: up to 20 market keys, comma separated.
  • min_edge: the lowest edge to list. On arbitrage it is profit_pct, on middles middle_profit_pct and on +EV ev_pct, all in percent. Low holds take max_hold instead, the highest hold_pct to list; min_edge on low holds, or max_hold anywhere else, gets 400.
  • max_age: the oldest price to use, in seconds since it was observed, from 60 to 2,700. The default is 600, the same age at which the odds routes stop calling a price available.
  • fair (+EV only): the fair-price methods to keep, comma separated: sharp_book, exchange, consensus.

Lists are 25 entries a page by default and up to 100 with limit. Follow links.next unchanged for the next page, as on every other list (pagination). For example, GET /v1/opportunities/arbs?league=nba&min_edge=1 lists NBA arbitrage paying at least 1% of the total stake.

Each call costs 10 credits, however many entries it returns; an empty answer costs nothing. See Plans and limits.

How fresh the lists are

The lists are recomputed about every four minutes. Each entry carries:

  • observed_at on every price: when that price was seen at the bookmaker.
  • price_age_seconds: the age of the oldest price in the entry when the response was built.
  • computed_at: when the entry was computed.
  • expires_at: when the entry is dropped unless a later pass finds it again.

An entry is listed only while every price in it is no older than max_age. Lower max_age for fewer, fresher entries. Even a fresh entry can be gone at the bookmaker by the time you get there, so read observed_at and check the price before you bet.

Reading a pair

Arbitrage, low holds and middles return pairs. Each pair has an id (stable while its two prices do not change), a type (arbitrage, low_hold or middle), the event and the market with its key, name, period and subject (null for a game market, the team or player otherwise), and two legs.

Each leg names its market_id, outcome_id, side, line (null when the market has no line), bookmaker, decimal_odds, implied_probability and observed_at. The ids are the same mk_ and oc_ ids the odds routes return, so you can look the outcome up in GET /v1/events/{event_id}/odds.

  • implied_sum is the two implied probabilities added. Below 1 is an arbitrage.
  • stake_pct is the share of the total stake to place on each leg, in percent, so both results return the same. It is null on a middle.
  • profit_pct (arbitrage and middles) is the profit in percent of the total stake when one leg wins. On a middle it is negative unless the pair is also an arbitrage.
  • hold_pct (low holds) is the combined margin of the two prices in percent: (implied_sum - 1) × 100, so 0 is break-even.
  • gap and middle_profit_pct (middles) are the distance between the two lines and the profit in percent of the total stake when both legs win.

Example response for GET /v1/opportunities/arbs?league=nba&limit=1 (sample prices, not real odds; the teams are invented):

{
  "data": [
    {
      "id": "op_3f9a1c7e5b2d4086",
      "type": "arbitrage",
      "event": {
        "id": "ev_7k2m9qx4hd3c",
        "name": "Harbour City Comets v Riverside Owls",
        "league": "nba",
        "scheduled_start": "2026-11-19T00:30:00.000Z"
      },
      "market": { "key": "total_points", "name": "Total points", "period": "full", "subject": null },
      "legs": [
        {
          "market_id": "mk_4n8r2kq7vd9x3hpc",
          "outcome_id": "oc_9tq3m7c2xk5w8z4e",
          "side": "over",
          "line": 221.5,
          "bookmaker": "sportsbet",
          "bookmaker_name": "Sportsbet",
          "decimal_odds": 2.02,
          "implied_probability": 0.49505,
          "stake_pct": 50.12,
          "observed_at": "2026-11-18T03:12:40.000Z"
        },
        {
          "market_id": "mk_4n8r2kq7vd9x3hpc",
          "outcome_id": "oc_2h6y8b4r9m3k7v5d",
          "side": "under",
          "line": 221.5,
          "bookmaker": "pinnacle",
          "bookmaker_name": "Pinnacle",
          "decimal_odds": 2.03,
          "implied_probability": 0.492611,
          "stake_pct": 49.88,
          "observed_at": "2026-11-18T03:13:05.000Z"
        }
      ],
      "implied_sum": 0.98766,
      "profit_pct": 1.249,
      "price_age_seconds": 150,
      "computed_at": "2026-11-18T03:14:00.000Z",
      "expires_at": "2026-11-18T03:18:00.000Z"
    }
  ],
  "meta": { "count": 1, "cursor": null },
  "links": { "next": null }
}

Here $100 split $50.12 on the over at 2.02 and $49.88 on the under at 2.03 returns about $101.25 whichever side wins, if both bets are accepted at those prices.

Reading a +EV entry

GET /v1/opportunities/ev returns one price per entry: the event, the market (with its id), the outcome (id, side, line), the bookmaker, its decimal_odds, implied_probability and observed_at, and:

  • fair: the fair price of the outcome, as probability and decimal_odds, and how it was made in method: - sharp_book: one sharp bookmaker's two-sided market with its margin removed, named in fair.bookmaker. - exchange: an exchange's back and lay prices, named in fair.bookmaker. - consensus: the margin-free prices of fair.books other bookmakers, combined, never including the bookmaker being priced. fair.devig says how each bookmaker's margin was removed.
  • ev_pct: the expected profit in percent of the stake, (decimal_odds × fair probability - 1) × 100.
  • kelly_pct: the full-Kelly stake in percent of bankroll. Many people stake a fraction of it.

One entry from GET /v1/opportunities/ev?league=nba&fair=sharp_book (sample prices, not real odds; the player and teams are invented):

{
  "id": "op_8c41e0b6a9d37f25",
  "type": "positive_ev",
  "event": {
    "id": "ev_7k2m9qx4hd3c",
    "name": "Harbour City Comets v Riverside Owls",
    "league": "nba",
    "scheduled_start": "2026-11-19T00:30:00.000Z"
  },
  "market": {
    "id": "mk_6w3d9r5k2t8h4n7b",
    "key": "player_points_over_under",
    "name": "Player points over/under",
    "period": "full",
    "subject": { "id": "pl_5r8k2m7x9c4d", "name": "Marcus Thornbury" }
  },
  "outcome": { "id": "oc_3k9v5h2w7r4m8c6t", "side": "over", "line": 24.5 },
  "bookmaker": "tab",
  "bookmaker_name": "TAB",
  "decimal_odds": 2.05,
  "implied_probability": 0.487805,
  "fair": {
    "probability": 0.510309,
    "decimal_odds": 1.96,
    "method": "sharp_book",
    "bookmaker": "pinnacle",
    "devig": null,
    "books": 1
  },
  "ev_pct": 4.613,
  "kelly_pct": 4.394,
  "observed_at": "2026-11-18T03:11:52.000Z",
  "price_age_seconds": 198,
  "computed_at": "2026-11-18T03:14:00.000Z",
  "expires_at": "2026-11-18T03:18:00.000Z"
}

Limits

  • Upcoming events only. An entry leaves the list when its event starts, when it expires, or when one of its prices gets older than max_age.
  • The lists are a few minutes behind the bookmakers at most, not live. For live prices read the odds routes or the stream.
  • +EV is an estimate, not a guarantee. A fair price is a model of the market, any single bet can lose, and prices move.
  • Bookmakers may limit or close accounts that take these prices often. Check every price at the bookmaker before acting.

To work the numbers out yourself from raw prices, see Check the market margin at the best prices. For fair prices on every market, closing line value and line moves, see Fair odds, closing line value and line moves. Every route and field is in the API reference.

Need help?FAQ · Email support

Search across the SockOdds documentation and catalogue.