EndpointsExperience Requests

Create counterfactual evaluation

POST/v2/counterfactual-evals

Queues paired replays of baseline episodes with a candidate policy. Returns before comparisons finish.

Authorization

BearerAuth
AuthorizationBearer <token>

Bearer token issued by Softmax

In: header

Request Body

application/json

candidate_policy_version_id*Candidate Policy Version Id

Candidate policy version to evaluate.

Formatuuid
baseline_policy_version_id*Baseline Policy Version Id

Baseline policy version to replay and compare against.

Formatuuid
league_id?|

League that supplies evaluation settings. Required when source is tournament.

source?CounterfactualEvalSource

Workflow creating the evaluation.

Default"tournament"

Value in

  • "tournament"
  • "experience_request"
n?|

Number of paired comparisons, or null to use the configured default.

idempotency_key*Idempotency Key

Key used to deduplicate repeated create requests.

Length5 <= length <= 200
curl -X POST "https://softmax.com/api/observatory/v2/counterfactual-evals" \  -H "Content-Type: application/json" \  -d '{    "candidate_policy_version_id": "b63b910e-ed55-43c0-be2c-6fd9d66eb022",    "baseline_policy_version_id": "353a49c0-b329-40ab-bec7-26c418899e74",    "idempotency_key": "string"  }'
{  "id": "string",  "candidate_policy_version_id": "b63b910e-ed55-43c0-be2c-6fd9d66eb022",  "baseline_policy_version_id": "353a49c0-b329-40ab-bec7-26c418899e74",  "candidate_policy_label": "string",  "baseline_policy_label": "string",  "candidate_player_name": "string",  "candidate_owner_name": "string",  "baseline_player_name": "string",  "baseline_owner_name": "string",  "league_id": "string",  "league_name": "string",  "source": "tournament",  "n_requested": 0,  "n_paired": 0,  "status": "pending",  "skip_reason": "string",  "error": "string",  "verdict": "progression",  "net_score": 0,  "wins": 0,  "losses": 0,  "neutrals": 0,  "mean_score_delta": 0,  "experience_request_id": "256083de-09b1-42d4-acb9-be39dd53448f",  "submission_id": "945dfedc-ad7f-4af7-b20c-06082ffa3cb3",  "league_policy_membership_id": "93c17082-599d-4779-9bb4-07972e51d754",  "idempotency_key": "string",  "created_by_user_id": "string",  "started_at": "2019-08-24T14:15:22Z",  "completed_at": "2019-08-24T14:15:22Z",  "created_at": "2019-08-24T14:15:22Z",  "pairs": [    {      "job_index": 0,      "baseline_episode_request_id": "string",      "candidate_episode_request_id": "string",      "candidate_episode_id": "442ebfae-0775-4701-b3ea-a6e05152d140",      "seed": 0,      "probe_slot": 0,      "opponent_policy_version_ids": [        "string"      ],      "coworld_id": "string",      "variant_id": "string",      "baseline_score": 0,      "candidate_score": 0,      "score_delta": 0,      "outcome": 0    }  ]}