Strategy Grid

While /lookup answers "what should this player do with their hand", /strategy_grid answers the same question for every possible starting hand at once: one request returns the full grid — the 169 starting-hand classes (13×13 pairs / suited / offsuit) with their strategies, EVs, and presentation menus, down to the individual combos.

Examples on this page use a No Limit Hold'em cash game.

Endpoint

Endpoint: POST /strategy_grid

curl --location '<glue-URL>/strategy_grid' \
--header 'Content-Type: application/json' \
--header 'Authorization: <Auth-TOKEN>' \
--data '{ ... }'

The actual <glue-URL> for your environment (dev / stg / prod) and the <Auth-TOKEN> are provided separately — the same ones you use for /lookup. As with /lookup, engine soft failures can be reported inside an HTTP 200 body via the error / error_code response fields.

Request

The request body is the same as for /lookup: request_id, the hand object, and request_profile. Two things change in practice:

• No hole_cards. The grid computes every starting hand, so no player carries hole cards.
• The grid is computed for the player whose turn it is at the end of actions.entries (returned as currentPlayer in the response).

request_profile works exactly as on /lookup — the same single field applies the same predefined configuration, with the same override rules, and its effects carry over to the grid: probabilities are thresholded, results are deterministic, every action carries its ev_rank label, and each hand gets the reduced presentation menu. See B2C Features for the full parameter walkthrough.

The same 8-handed spot used on the B2C Features page (straddle on seat 2, 50-chip ante, preflop, seat 3 to act):

curl --location '<glue-URL>/strategy_grid' \
--header 'Content-Type: application/json' \
--header 'Authorization: <Auth-TOKEN>' \
--data '{
  "request_id": "grid-demo",
  "hand": {
    "gameuuid": "3b-0.5_8p_game",
    "game_type": "nlhe",
    "game_mode_code": "normal",
    "players": [
      { "seat_no": 0, "stack": 46720 },
      { "seat_no": 1, "stack": 32580 },
      { "seat_no": 2, "stack": 27000 },
      { "seat_no": 3, "stack": 40368 },
      { "seat_no": 4, "stack": 105250 },
      { "seat_no": 5, "stack": 9700 },
      { "seat_no": 6, "stack": 35440 },
      { "seat_no": 7, "stack": 13550 }
    ],
    "big_blind": 100,
    "ante": 50,
    "dealer_seat": 7,
    "sb_seat": 0,
    "bb_seat": 1,
    "straddle_seat": 2,
    "actions": { "entries": [] }
  },
  "request_profile": "<client_profile>"
}'

Response

{
  "hands": [ <Hand-class> × 169 ],
  "actionLabels": [ <Action-label> ],
  "currentPlayer": <acting-seat>,
  "numCombos": 1326,
  "numBoardCards": 0,
  "overallReach": 1.0,
  "supportScore": <0-to-1>
}

The response also carries diagnostic members not listed here.

Action labels

actionLabels describes the spot's action menu once, so the per-hand data can stay compact. Each entry:

{
  "name": "raise 5.0 BB",
  "amount": 500,
  "amountBb": 5.0,
  "potRatio": 0.3157894736842105,
  "strategy": 0.10491631517209216
}

In the example spot, the range-level strategy reads: fold 76.9%, call 0.02%, raise to 5 bb 10.5%, raise to 7 bb 12.6% — the acting player opens about a quarter of all hands.

Hand classes

Each hands[] entry describes one starting-hand class:

{
  "handName": "TT",
  "reach": 1.0,
  "ev": 5.747151374816894,
  "actions": [
    { "actionName": "raise5.0", "probability": 0.5339687181136681, "ev": 5.747151374816895 },
    { "actionName": "raise7.0", "probability": 0.466031281886332, "ev": 5.747151374816895 }
  ],
  "reduced_actions": [
    { "actionName": "fold", "probability": 0.0, "ev": 0.0,
      "ev_rank": "EVR_2", "reduced_action_sizing": 0.0 },
    { "actionName": "call", "probability": 0.0, "ev": 4.7471513748168945,
      "ev_rank": "EVR_3", "reduced_action_sizing": 1.0 },
    { "actionName": "raise5.0", "probability": 0.5274196517442273, "ev": 5.7471513748168945,
      "ev_rank": "EVR_4", "reduced_action_sizing": 2.5 },
    { ... }
  ],
  "combos": [
    {
      "combo": "TcTd",
      "reach": 1.0,
      "actions": [
        {
          "actionName": "fold",
          "probability": 0.0,
          "ev": 0.0,
          "ev_rank": "EVR_2"
        },
        {
          "actionName": "raise5.0",
          "probability": 0.533968718113668,
          "ev": 5.7471513748168945,
          "ev_rank": "EVR_4"
        },
        { ... }
      ],
      "reduced_actions": [ ... ]
    },
    { ... }
  ],
  "supportScore": 1.0
}

The ev_rank labels follow the same scale as on /lookup — see Reading the Response on the B2C Features page.

Reading the example: TT opens 100% of the time, mixing a raise to 5 bb (53.4%) and a raise to 7 bb (46.6%), both labeled EVR_4 and worth +5.75 bb — while a class like 72o returns a single entry, fold at 100%. The compact class-level actions list carries just what the hand plays; the reduced_actions and combos carry the full detail when you need it.

Notes

• All B2C features apply per hand — strategy_min_threshold trims the class-level actions, reduced_actions_tier shapes each hand's presentation menu, show_labels provides the ev_rank labels — and request_profile follows the same silent-no-match rule as on /lookup: a misspelled profile name applies no defaults and raises no error.
• The grid response can be large (169 classes × combos). Request it per decision point, not per player action batch.