Riftcast Developer API

Base URL

https://riftcast.gg/v1

All responses are JSON. Every charged response includes an X-Credits-Remaining header.

Authentication

Pass your key on every request, via either header:

Authorization: Bearer YOUR_KEY
# or
X-Api-Key: YOUR_KEY

Create and manage keys, and see your balance, on your account page. A key is shown only once at creation. Every account gets 20 free credits per month — claim them, or subscribe to a credit plan / top up, on your account page.

Credits & cost

Calls are billed from a prepaid credit pool shared across all your keys. Discovery and usage calls are free; data and predictions cost credits:

*Odds calls are charged only when the series actually has captured odds for the requested scope — a "no odds yet" 404 is always free.

Failed calls are never charged: a 404 (unknown id), 400 (bad request) or 500 (prediction error) costs nothing. When your balance is too low you get 402 Payment Required — top up and retry.

Every account gets 20 free credits per month. For more, subscribe to a credit plan (best value per credit) or buy one-time top-ups on your account page.

Fair use: credits are per account. Creating multiple accounts to farm the free allowance isn't allowed — using the API from more than one account on the same IP address will get those accounts banned.

Discovery free

Fetch the valid IDs to use in stats and prediction calls. These change rarely — cache them.

• GET /v1/teams — all teams (id, name, region, logo)
• GET /v1/players — all players (id, name, role, team)
• GET /v1/champions — all champions
• GET /v1/tournaments — all tournaments

curl https://riftcast.gg/v1/teams -H "Authorization: Bearer YOUR_KEY"

{
  "count": 420,
  "teams": [
    { "id": 279, "name": "100 Thieves", "region": null, "logoUrl": "https://..." },
    ...
  ]
}

Stats 1 credit

• GET /v1/teams/{id} — record, win rate, GPM, side win rates, champion pool
• GET /v1/players/{id} — KDA, CS/min, gold share, kill participation, champion pool
• GET /v1/tournaments/{id} — side win rates, first-objective rates, leaders

curl https://riftcast.gg/v1/teams/279 -H "Authorization: Bearer YOUR_KEY"

{
  "id": 279,
  "name": "100 Thieves",
  "record": "5W - 1L",
  "winRate": 83.3,
  "stats": { "gpm": 2000, "csPerMin": 32.2, "kda": 5.68, ... },
  "sideWinRate": { "blue": 1.0, "red": 0.75 },
  "championPool": [ ... ]
}

Odds 1–2 credits

The latest captured market odds for the next scheduled series between two teams — the same market line riftcast.gg surfaces. Snapshots refresh periodically pre-match; capturedAtUtc tells you how fresh the line is, and isClosingLine marks the final pre-kickoff snapshot. You're charged only when odds exist for the requested scope — a "no odds yet" 404 is free.

GET /v1/odds

curl "https://riftcast.gg/v1/odds?team1Id=13&team2Id=3&markets=all" \
  -H "Authorization: Bearer YOUR_KEY"

{
  "match": {
    "esportsMatchId": "...", "league": "MSI", "bestOf": 5,
    "scheduledUtc": "2026-07-04T08:00:00",
    "team1": { "id": 13, "name": "Bilibili Gaming" },
    "team2": { "id": 3, "name": "T1" }
  },
  "capturedAtUtc": "2026-07-03T12:04:11",
  "isClosingLine": false,
  "markets": {
    "seriesWinner": [
      { "team": { "id": 13, "name": "Bilibili Gaming" }, "decimalOdds": 2.20 },
      { "team": { "id": 3, "name": "T1" }, "decimalOdds": 1.61 }
    ],
    "handicap": [
      { "line": -1.5, "odds": [
        { "team": { "id": 3, "name": "T1" }, "line": -1.5, "decimalOdds": 2.30 },
        { "team": { "id": 13, "name": "Bilibili Gaming" }, "line": 1.5, "decimalOdds": 1.57 } ] }
    ],
    "totalMaps": [ { "line": 4.5, "over": 2.75, "under": 1.40 } ],
    "correctScore": [
      { "team": { "id": 3, "name": "T1" }, "score": "3-0", "decimalOdds": 4.33 },
      ...
    ]
  }
}

Markets the line-setter hasn't published come back null. With markets=winner only seriesWinner is returned. Coverage isn't universal — smaller leagues often have no listed odds, in which case the call is a free 404.

Predictions

Win probabilities for any matchup. Provide team IDs (from /v1/teams) — names also work but IDs are unambiguous. Optionally pass decimal oddsBlue / oddsRed to get expected value (EV) back.

POST /v1/predict

curl -X POST https://riftcast.gg/v1/predict \
  -H "Authorization: Bearer YOUR_KEY" \
  -H "Content-Type: application/json" \
  -d '{"blueTeamId":279,"redTeamId":110,"oddsBlue":2.0,"oddsRed":1.8,"model":"consensus","tournament":"LCS 2026 Summer"}'

{
  "blueTeam": { "id": 279, "name": "100 Thieves" },
  "redTeam":  { "id": 110, "name": "3BL Esports" },
  "model": "consensus",
  "blueWinProbability": 0.8105,
  "redWinProbability": 0.1895,
  "calibratedBlueWinProbability": 0.7462,
  "calibratedRedWinProbability": 0.2538,
  "winner": "blue",
  "odds": { "blue": 2.0, "red": 1.8, "blueEv": 0.621, "redEv": -0.659 },
  "models": [
    { "model": "fasttree", "blueWinProbability": 0.79, "redWinProbability": 0.21, "calibratedBlueWinProbability": 0.74, "calibratedRedWinProbability": 0.26, "winner": "blue" },
    { "model": "lightgbm", "blueWinProbability": 0.83, "redWinProbability": 0.17, "calibratedBlueWinProbability": 0.76, "calibratedRedWinProbability": 0.24, "winner": "blue" },
    { "model": "pcasweep", "blueWinProbability": 0.81, "redWinProbability": 0.19, "calibratedBlueWinProbability": 0.75, "calibratedRedWinProbability": 0.25, "winner": "blue" }
  ]
}

A consensus call runs all three base models and returns them in models alongside the blended top-level probability — one call, the full picture.

POST /v1/predict-draft

As /v1/predict, plus bluePicks and redPicks — lists of { "role": "TOP", "champion": "Aatrox" } (roles: TOP, JUNGLE, MID, BOT, SUPPORT). Champion names come from /v1/champions.

How to read the probabilities

Raw vs calibrated. blueWinProbability / redWinProbability are the models' raw output. calibratedBlueWinProbability / calibratedRedWinProbability map that raw confidence through the model's realized accuracy at the same confidence level (measured on graded predictions) — this calibrated number is what riftcast.gg displays. When a model has been overconfident recently, calibrated will sit well below raw. Until enough graded history exists, calibrated equals raw.

Series semantics. The probability is the series-winner estimate used across riftcast — it is not a single-map number to compound with a best-of-N formula. Maps in a series are not independent (sides alternate, drafts adapt), so riftcast deliberately avoids iid Bo3/Bo5 amplification and instead grades and calibrates the raw number against realized series outcomes.

Sides. Both prediction endpoints are side-neutral by design: map sides are unknown pre-match and alternate across a series, so the engine always averages the two side orientations — swapping blueTeam and redTeam returns complementary probabilities (within a small numerical tolerance). The blue/red labels orient the response; they are not a claim about which map side a team will play. /v1/predict-draft differs by being draft-aware — each team's picks stay attached to that team — not by being side-aware.

Consensus. The API's consensus is the plain average of the three base models. The site's displayed Consensus additionally weights models by their recent accuracy, so the two can differ by a point or two.

Errors

Errors are JSON: { "error": "code", "message": "..." }.