Spaces:
Running
Running
File size: 16,390 Bytes
0b3ef70 | 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72 73 74 75 76 77 78 79 80 81 82 83 84 85 86 87 88 89 90 91 92 93 94 95 96 97 98 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 120 121 122 123 124 125 126 127 128 129 130 131 132 133 134 135 136 137 138 139 140 141 142 143 144 145 146 147 148 149 150 151 152 153 154 155 156 157 158 159 160 161 162 163 164 165 166 167 168 169 170 171 172 173 174 175 176 177 178 179 180 181 182 183 184 185 186 187 188 189 190 191 192 193 194 195 196 197 198 199 200 201 202 203 204 205 206 207 208 209 210 211 212 213 214 215 216 217 218 219 220 221 222 223 224 225 226 227 228 229 230 231 232 233 234 235 236 237 238 239 240 241 242 243 244 245 246 247 248 249 250 251 252 253 254 255 256 257 258 259 260 261 262 263 264 265 266 267 268 269 270 271 272 273 274 275 276 277 278 279 280 281 282 283 284 285 286 287 288 289 290 291 292 293 294 295 296 297 298 299 300 301 302 303 304 305 306 307 308 309 310 311 312 313 314 315 316 317 318 319 320 321 322 323 324 325 326 327 328 329 330 331 332 333 334 335 336 337 338 339 340 341 342 343 344 345 346 347 348 349 350 351 352 353 354 355 356 357 358 359 360 361 362 363 364 365 366 367 368 369 370 371 372 373 374 375 376 377 378 379 380 381 382 383 384 385 386 387 388 389 390 391 392 393 394 395 396 397 398 399 400 401 402 403 404 405 406 407 408 409 410 411 412 413 414 415 416 417 418 419 420 421 422 423 424 425 426 427 428 429 430 431 432 433 434 435 436 437 438 439 440 441 442 443 444 445 446 447 448 449 450 | """
Step 5: Simulate Markov chains.
This script:
1. Loads a transition matrix
2. Runs N simulations from the initial state
3. Records sequence details
Input:
- Transition matrix CSV
- Number of simulations
Output:
- secuencias_simuladas.csv
"""
import pandas as pd
import numpy as np
from pathlib import Path
from typing import Dict, List, Tuple, Optional
from collections import Counter
from tqdm import tqdm
from .utils import load_config, ensure_output_dir, parse_tuple_string
# =============================================================================
# VALIDATION FUNCTIONS
# =============================================================================
def validate_matrix_file(matrix_path: Path) -> None:
"""Validate that the transition matrix file exists and is readable."""
if not matrix_path.exists():
raise FileNotFoundError(
f"Transition matrix file not found: {matrix_path}\n"
f"Please run Step 3 or Step 4 first to generate the transition matrix."
)
if matrix_path.stat().st_size == 0:
raise ValueError(f"Transition matrix file is empty: {matrix_path}")
def validate_matrix_structure(transition_matrix: pd.DataFrame, matrix_path: Path) -> None:
"""Validate the structure and content of the transition matrix."""
if transition_matrix.empty:
raise ValueError(
f"Transition matrix is empty after loading from: {matrix_path}"
)
# Check if matrix is square
if transition_matrix.shape[0] != transition_matrix.shape[1]:
raise ValueError(
f"Transition matrix must be square. "
f"Got shape {transition_matrix.shape} from: {matrix_path}"
)
# Check for NaN values
nan_count = transition_matrix.isna().sum().sum()
if nan_count > 0:
raise ValueError(
f"Transition matrix contains {nan_count} NaN values. "
f"Please check the matrix generation in Step 3/4."
)
# Check that all values are non-negative
if (transition_matrix.values < 0).any():
raise ValueError(
f"Transition matrix contains negative values. "
f"All probabilities must be >= 0."
)
def validate_probability_rows(transition_matrix: pd.DataFrame, tolerance: float = 0.01) -> None:
"""Validate that non-absorbing state rows sum to approximately 1."""
row_sums = transition_matrix.sum(axis=1)
# Identify rows that don't sum to ~1 (excluding absorbing states which have self-loops)
invalid_rows = []
for idx, row_sum in enumerate(row_sums):
state_name = transition_matrix.index[idx]
# Skip validation for absorbing states (they should sum to 1 via self-loop)
if 'ABSORCION' in str(state_name):
continue
if abs(row_sum - 1.0) > tolerance:
invalid_rows.append((state_name, row_sum))
if invalid_rows:
error_details = "\n".join([f" - {state}: sum={s:.4f}" for state, s in invalid_rows[:10]])
raise ValueError(
f"Found {len(invalid_rows)} rows with invalid probability sums (expected ~1.0):\n"
f"{error_details}"
f"{f'... and {len(invalid_rows) - 10} more' if len(invalid_rows) > 10 else ''}"
)
def validate_initial_state(state_to_idx: Dict, initial_state: Tuple) -> None:
"""Validate that the initial state exists in the matrix."""
if initial_state not in state_to_idx:
available_states = [s for s in state_to_idx.keys() if 'CORNER_START' in str(s)][:5]
raise ValueError(
f"Initial state {initial_state} not found in transition matrix.\n"
f"Available CORNER_START-related states: {available_states}\n"
f"Total states in matrix: {len(state_to_idx)}"
)
def validate_simulation_params(num_simulations: int, random_seed: int) -> None:
"""Validate simulation parameters."""
if num_simulations <= 0:
raise ValueError(f"num_simulations must be positive, got: {num_simulations}")
if num_simulations > 10_000_000:
raise ValueError(
f"num_simulations={num_simulations:,} is very large. "
f"Maximum recommended is 10,000,000 to avoid memory issues."
)
if random_seed < 0:
raise ValueError(f"random_seed must be non-negative, got: {random_seed}")
# =============================================================================
# SIMULATION FUNCTIONS
# =============================================================================
def is_absorption_state(state: Tuple) -> bool:
"""Check if state is absorption (terminal)."""
if isinstance(state, tuple) and len(state) > 0:
return state[0] == 'ABSORCION'
return False
def get_absorption_type(state: Tuple) -> Optional[str]:
"""Get absorption type from state."""
if isinstance(state, tuple) and len(state) > 1:
return state[1]
return None
def simulate_sequence(
transition_matrix: pd.DataFrame,
state_to_idx: Dict,
idx_to_state: Dict,
initial_state: Tuple,
rng: np.random.Generator,
max_steps: int = 1000
) -> Dict:
"""
Simulate a single sequence from initial state.
Returns:
Dictionary with sequence information
Raises:
ValueError: If initial state is not in the matrix (should be caught earlier)
"""
states = [initial_state]
# This should have been validated earlier, but check again for safety
if initial_state not in state_to_idx:
raise ValueError(
f"Initial state {initial_state} not found in state_to_idx. "
f"This should have been caught during validation."
)
current_state = initial_state
for step in range(max_steps):
current_idx = state_to_idx[current_state]
transition_probs = transition_matrix.iloc[current_idx, :].values
prob_sum = np.sum(transition_probs)
if prob_sum == 0:
# This is a dead-end state with no outgoing transitions
# This shouldn't happen in a well-formed matrix but we handle it
return {
'states': states,
'length': len(states) - 1,
'terminated': True,
'termination_reason': 'no_transitions',
'absorption_type': None
}
# Normalize probabilities if they don't sum to exactly 1 (floating point issues)
if abs(prob_sum - 1.0) > 1e-10:
transition_probs = transition_probs / prob_sum
# Sample next state
try:
next_idx = rng.choice(len(transition_probs), p=transition_probs)
except ValueError as e:
raise ValueError(
f"Failed to sample next state at step {step} from state {current_state}. "
f"Probabilities sum to {prob_sum}. Error: {e}"
)
next_state = idx_to_state[next_idx]
states.append(next_state)
if is_absorption_state(next_state):
return {
'states': states,
'length': len(states) - 1,
'terminated': True,
'termination_reason': 'absorption',
'absorption_type': get_absorption_type(next_state)
}
current_state = next_state
# Max steps reached - this indicates the chain didn't absorb
return {
'states': states,
'length': len(states) - 1,
'terminated': True,
'termination_reason': 'max_steps_reached',
'absorption_type': None
}
def run_simulations(
matrix_path: Path,
output_folder: Path,
num_simulations: int = 50000,
random_seed: int = 42
) -> Path:
"""
Main function to run Markov simulations.
Args:
matrix_path: Path to transition matrix CSV
output_folder: Output directory
num_simulations: Number of simulations
random_seed: Random seed for reproducibility
Returns:
Path to output CSV
"""
print(f"\n{'='*80}")
print("STEP 5: MARKOV CHAIN SIMULATIONS")
print(f" Simulations: {num_simulations:,}")
print(f" Random seed: {random_seed}")
print(f"{'='*80}")
# =========================================================================
# VALIDATION
# =========================================================================
print(f"\n🔍 Validating inputs...")
# Validate simulation parameters
validate_simulation_params(num_simulations, random_seed)
# Validate matrix file exists
validate_matrix_file(matrix_path)
# Load matrix
print(f"\n📂 Loading transition matrix from {matrix_path}...")
try:
transition_matrix = pd.read_csv(matrix_path, index_col=0)
except pd.errors.EmptyDataError:
raise ValueError(f"Transition matrix file is empty or malformed: {matrix_path}")
except pd.errors.ParserError as e:
raise ValueError(f"Failed to parse transition matrix CSV: {matrix_path}\nError: {e}")
# Validate matrix structure
validate_matrix_structure(transition_matrix, matrix_path)
# Parse states
print(f" Parsing state tuples...")
states = []
for i, s in enumerate(transition_matrix.index):
try:
parsed = parse_tuple_string(s)
states.append(parsed)
except Exception as e:
raise ValueError(
f"Failed to parse state at index {i}: '{s}'\nError: {e}"
)
state_to_idx = {state: i for i, state in enumerate(states)}
idx_to_state = {i: state for i, state in enumerate(states)}
print(f" ✅ Loaded {len(states)} states")
# Validate probability rows
validate_probability_rows(transition_matrix)
print(f" ✅ Probability rows validated")
# Initial state - always start from CORNER_START (the beginning of a corner sequence)
initial_state = ('CORNER_START', 'corner', 'atacante')
validate_initial_state(state_to_idx, initial_state)
# Run simulations
print(f"\n🎲 Running {num_simulations:,} simulations...")
rng = np.random.default_rng(random_seed)
sequences = []
for i in tqdm(range(num_simulations), desc=" Simulating"):
seq = simulate_sequence(
transition_matrix, state_to_idx, idx_to_state, initial_state, rng
)
seq['sequence_id'] = i + 1
sequences.append(seq)
print(f" ✅ Simulations complete")
# Analyze results
absorption_counts = Counter(s['absorption_type'] for s in sequences if s['absorption_type'])
termination_counts = Counter(s['termination_reason'] for s in sequences)
lengths = [s['length'] for s in sequences]
# Validate simulation results
max_steps_count = termination_counts.get('max_steps_reached', 0)
no_transitions_count = termination_counts.get('no_transitions', 0)
if max_steps_count > 0:
pct = max_steps_count / num_simulations * 100
print(f"\n⚠️ WARNING: {max_steps_count:,} sequences ({pct:.2f}%) reached max_steps without absorbing.")
if pct > 5:
raise ValueError(
f"Too many sequences ({pct:.1f}%) reached max_steps without absorbing. "
f"This indicates a problem with the transition matrix (e.g., missing absorption states)."
)
if no_transitions_count > 0:
pct = no_transitions_count / num_simulations * 100
print(f"\n⚠️ WARNING: {no_transitions_count:,} sequences ({pct:.2f}%) hit dead-end states.")
if pct > 1:
raise ValueError(
f"Too many sequences ({pct:.1f}%) hit dead-end states with no outgoing transitions. "
f"This indicates a problem with the transition matrix."
)
print(f"\n📊 Simulation statistics:")
print(f" Mean length: {np.mean(lengths):.2f}")
print(f" Median length: {np.median(lengths):.1f}")
print(f" Max length: {max(lengths)}")
print(f"\n📊 Termination reasons:")
for reason, count in termination_counts.most_common():
pct = count / num_simulations * 100
print(f" {reason}: {count:,} ({pct:.1f}%)")
print(f"\n📊 Absorption distribution:")
for abs_type, count in absorption_counts.most_common():
pct = count / num_simulations * 100
print(f" {abs_type}: {count:,} ({pct:.1f}%)")
# Count corners per sequence for statistics
# Each sequence starts with 1 corner (CORNER_START)
# If it ends in ABSORCION(corner), that's another corner won
corner_counts = Counter()
for seq in sequences:
# Start with 1 for the initial corner
num_corners = 1
# Add 1 if ended by winning another corner
if seq['absorption_type'] == 'corner':
num_corners += 1
corner_counts[num_corners] += 1
print(f"\n📊 Corners per sequence:")
for n_corners, count in sorted(corner_counts.items()):
pct = count / num_simulations * 100
label = "corner" if n_corners == 1 else "corners"
print(f" {n_corners} {label}: {count:,} ({pct:.1f}%)")
multi_corner_pct = sum(c for n, c in corner_counts.items() if n > 1) / num_simulations * 100
print(f" → Sequences ending with another corner: {multi_corner_pct:.1f}%")
# Prepare output DataFrame
rows = []
for seq in sequences:
states_list = seq['states']
# Count corners: 1 for initial + 1 if ended in corner absorption
num_corners = 1
if seq['absorption_type'] == 'corner':
num_corners += 1
# Get corner zone (first state after CORNER_START)
corner_zone = None
if len(states_list) > 1:
first_state = states_list[1]
if isinstance(first_state, tuple) and len(first_state) > 0:
corner_zone = first_state[0] if first_state[0] != 'ABSORCION' else None
# Count events
event_counts = Counter()
zones_visited = []
events_sequence = []
for state in states_list:
if isinstance(state, tuple) and len(state) >= 2:
if state[0] not in ['CORNER_START', 'ABSORCION']:
zones_visited.append(state[0])
event_counts[state[1]] += 1
events_sequence.append(state[1])
rows.append({
'sequence_id': seq['sequence_id'],
'num_corners': num_corners,
'num_events': seq['length'],
'absorption_event': seq['absorption_type'] or '',
'corner_zone': corner_zone,
'termination_reason': seq['termination_reason'],
'count_pass': event_counts.get('pass', 0),
'count_shot': event_counts.get('shot', 0),
'count_defensive_possession': event_counts.get('defensive_possession', 0),
'count_keeper_action': event_counts.get('keeper_action', 0),
'count_other_events': event_counts.get('other_events', 0),
'states_sequence': '|'.join(str(s) for s in states_list),
'zones_sequence': '|'.join(zones_visited),
'events_sequence': '|'.join(events_sequence),
})
df = pd.DataFrame(rows)
# Save
ensure_output_dir(output_folder)
output_path = output_folder / "secuencias_simuladas.csv"
df.to_csv(output_path, index=False)
print(f"\n ✅ Saved: {output_path} ({len(df):,} sequences)")
print(f"\n{'='*80}")
print("✅ STEP 5 COMPLETE")
print(f"{'='*80}")
return output_path
if __name__ == "__main__":
import argparse
parser = argparse.ArgumentParser(description="Run Markov simulations")
parser.add_argument("--matrix-path", required=True)
parser.add_argument("--output-folder", required=True)
parser.add_argument("--num-simulations", type=int, default=50000)
parser.add_argument("--random-seed", type=int, default=42)
args = parser.parse_args()
run_simulations(
matrix_path=Path(args.matrix_path),
output_folder=Path(args.output_folder),
num_simulations=args.num_simulations,
random_seed=args.random_seed
)
|