---

name: GTO Solver Integration description: Integrate PioSolver data with the poker training system

GTO Solver Integration Skill

Overview

Integrate PioSolver-solved hands with the Smarter.Poker training and AI systems.

Database Tables

solved_spots_gold (Postflop Engine)

Copy-- Query structure
SELECT * FROM solved_spots_gold 
WHERE street = 'Flop'
  AND stack_depth = 40
  AND game_type = 'MTT'
  AND topology = '6-Max'
  AND mode = 'ICM';

Key fields:

• scenario_hash - Unique identifier
• board_cards - Array like ['As', '7d', '2h']
• strategy_matrix - JSONB with all 1,326 hands
• macro_metrics - Range advantage, SPR, etc.

memory_charts_gold (Preflop Engine)

Copy-- Query structure
SELECT chart_grid FROM memory_charts_gold
WHERE chart_name = 'UTG_Open_100bb_9Max';

Chart grid format:

Copy{
  "AA": {"action": "Raise", "freq": 1.0, "size": "3bb"},
  "AKs": {"action": "Raise", "freq": 1.0, "size": "3bb"},
  "72o": {"action": "Fold", "freq": 1.0}
}

Strategy Matrix Format

Copy{
  "AhKd": {
    "best_action": "Raise",
    "max_ev": 10.5,
    "ev_loss": 0,
    "actions": {
      "Raise": {"ev": 10.5, "freq": 1.0, "size": "66%"},
      "Call": {"ev": 8.2, "freq": 0.0},
      "Fold": {"ev": 0.0, "freq": 0.0}
    },
    "is_mixed": false
  }
}

Helper Functions

Get Horse Action (applies personality bias)

CopySELECT get_horse_action(
  1,  -- author_id
  '{"Raise": {"freq": 0.45}, "Call": {"freq": 0.35}, "Fold": {"freq": 0.20}}'::jsonb
);
-- Returns: 'Raise', 'Call', or 'Fold' based on personality

Get Horse Personality

CopySELECT get_horse_personality(1);
-- Returns full JSONB with gto_adherence, mixed_strategy_bias, etc.

Training Integration

Scenario Selection

1. Query solved_spots_gold for matching conditions
2. Extract hero's hand from strategy_matrix
3. Present options to user
4. Compare to GTO frequencies
5. Calculate EV loss for feedback

Example Flow

Copy// 1. Get solved spot
const { data: spot } = await supabase
  .from('solved_spots_gold')
  .select('strategy_matrix, macro_metrics')
  .eq('scenario_hash', scenarioHash)
  .single();

// 2. Get hero's hand strategy
const heroStrategy = spot.strategy_matrix[heroHand];

// 3. Calculate EV loss if user picks wrong
const evLoss = heroStrategy.max_ev - heroStrategy.actions[userAction].ev;

Ingestion Script

Located at: /scripts/ingest_god_mode.py

• Processes PioSolver CSV exports
• Uploads to Supabase
• Creates scenario hashes

Best Practices

1. Always check is_mixed before presenting single action
2. Show frequencies for mixed spots (e.g., "Raise 45%, Call 55%")
3. Use macro_metrics for context (range advantage, SPR)
4. Cache frequently-used spots for performance