OggyCodes12/OpenEnvHackathon
0
1#!/usr/bin/env python32"""3QUICK START GUIDE - QuantitativeTrading-v14 5Copy & paste examples to get started immediately.6"""7 8# ============================================================================9# EXAMPLE 1: Simplest Possible Usage (5 lines of code)10# ============================================================================11 12from env import TradingEnvironment, MarketDataLoader13from tasks import TaskFactory14from agents import RandomAgent15 16task = TaskFactory.create_easy_task()17env = TradingEnvironment(task.config, MarketDataLoader())18agent = RandomAgent(task.config.symbols)19state, _ = env.reset(seed=42)20 21for _ in range(50):22 action = agent.act(state, {})23 state, reward, done, info = env.step(action)24 if done:25 break26 27print(f"Final portfolio: ${info['portfolio_value']:.2f}")28 29 30# ============================================================================31# EXAMPLE 2: Full Training Run with Evaluation32# ============================================================================33 34from env import TradingEnvironment, MarketDataLoader, Config35from tasks import TaskFactory, Evaluator36from agents import RandomAgent, MomentumAgent, TechnicalAgent37 38def run_agent_on_task(agent, task, num_episodes=3):39 """Run agent and return metrics."""40 results = []41 42 for episode in range(num_episodes):43 loader = MarketDataLoader()44 env = TradingEnvironment(task.config, loader)45 state, _ = env.reset(seed=42 + episode)46 47 portfolio_values = [env.get_portfolio_value()]48 returns = []49 done = False50 51 while not done:52 action = agent.act(state, {})53 state, reward, done, info = env.step(action)54 portfolio_values.append(env.get_portfolio_value())55 returns.append(info['metrics']['step_return'])56 57 metrics = Evaluator.evaluate_episode(portfolio_values, returns, [])58 results.append(metrics)59 60 return results61 62# Test all agents on medium task63task = TaskFactory.create_medium_task()64 65agents = {66 'Random': RandomAgent(task.config.symbols),67 'Momentum': MomentumAgent(task.config.symbols),68 'Technical': TechnicalAgent(task.config.symbols),69}70 71for agent_name, agent in agents.items():72 results = run_agent_on_task(agent, task, num_episodes=3)73 avg_return = sum(r.total_return for r in results) / len(results)74 avg_sharpe = sum(r.sharpe_ratio for r in results) / len(results)75 print(f"{agent_name:10}: Return={avg_return*100:6.2f}%, Sharpe={avg_sharpe:6.2f}")76 77 78# ============================================================================79# EXAMPLE 3: Custom Agent Implementation80# ============================================================================81 82from agents import BaseAgent83from env import Action, OrderType84 85class SimpleMovingAverageAgent(BaseAgent):86 """Buy on short MA above long MA, sell on cross below."""87 88 def __init__(self, symbols):89 super().__init__(symbols, "SMAAgent")90 self.prev_signal = None91 92 def act(self, state, info):93 """Generate action based on MA crossover."""94 symbol = self.symbols[0]95 96 # In a real implementation, extract MA values from state97 # For now, use a simple heuristic98 step_return = info.get('metrics', {}).get('step_return', 0)99 100 if step_return > 0.01 and not self.prev_signal:101 action = Action(symbol=symbol, order_type=OrderType.BUY, quantity=20.0)102 self.prev_signal = True103 elif step_return < -0.01 and self.prev_signal:104 action = Action(symbol=symbol, order_type=OrderType.SELL, quantity=20.0)105 self.prev_signal = False106 else:107 action = Action(symbol=symbol, order_type=OrderType.HOLD, quantity=0.0)108 109 return action110 111 def reset(self):112 self.prev_signal = None113 114 115# ============================================================================116# EXAMPLE 4: Custom Configuration and Task117# ============================================================================118 119from env import Config, TradingEnvironment, MarketDataLoader120 121# Create custom config122custom_config = Config(123 symbols=['ASSET1', 'ASSET2'],124 initial_cash=50000.0,125 max_positions=2,126 max_position_size=30.0,127 slippage_bps=2.0,128 commission_bps=2.0,129 min_order_size=1.0,130 max_leverage=2.5,131 episode_length=150,132 lookback_window=75,133)134 135# Create environment with custom config136loader = MarketDataLoader()137env = TradingEnvironment(custom_config, loader)138 139# Use with any agent140agent = RandomAgent(custom_config.symbols)141state, _ = env.reset(seed=42)142 143total_reward = 0144for _ in range(100):145 action = agent.act(state, {})146 state, reward, done, info = env.step(action)147 total_reward += reward148 if done:149 break150 151print(f"Total Reward: {total_reward:.2f}, Portfolio: ${info['portfolio_value']:.2f}")152 153 154# ============================================================================155# EXAMPLE 5: Advanced: Loading Real Data and Analysis156# ============================================================================157 158from env import MarketDataLoader, TradingEnvironment, Config159from tasks import TaskFactory, Evaluator160import json161 162# Load custom data163loader = MarketDataLoader()164 165# Generate synthetic data for multiple assets166for symbol in ['ASSET1', 'ASSET2', 'ASSET3']:167 loader.generate_synthetic_data(168 symbol, 169 num_bars=2000, 170 volatility=0.02,171 trend=0.0001172 )173 174# Get task and environment175task = TaskFactory.create_medium_task()176env = TradingEnvironment(task.config, loader)177 178# Run episode179state, info = env.reset(seed=42)180agent = RandomAgent(task.config.symbols)181 182metrics_history = []183portfolio_values = [env.get_portfolio_value()]184returns_list = []185 186done = False187while not done:188 action = agent.act(state, info)189 state, reward, done, info = env.step(action)190 191 portfolio_values.append(env.get_portfolio_value())192 returns_list.append(info['metrics']['step_return'])193 metrics_history.append(info['metrics'])194 195# Evaluate196final_metrics = Evaluator.evaluate_episode(197 portfolio_values, 198 returns_list, 199 []200)201 202print(f"\nEpisode Analysis:")203print(f" Duration: {len(portfolio_values)-1} steps")204print(f" Return: {final_metrics.total_return*100:.2f}%")205print(f" Sharpe: {final_metrics.sharpe_ratio:.2f}")206print(f" Max Drawdown: {final_metrics.max_drawdown*100:.2f}%")207print(f" Win Rate: {final_metrics.win_rate*100:.2f}%")208print(f" Score: {final_metrics.score():.1f}/100")209 210# Save results211results = {212 'portfolio_values': portfolio_values,213 'returns': returns_list,214 'final_metrics': {215 'total_return': float(final_metrics.total_return),216 'sharpe_ratio': float(final_metrics.sharpe_ratio),217 'max_drawdown': float(final_metrics.max_drawdown),218 'score': float(final_metrics.score()),219 }220}221 222with open('episode_results.json', 'w') as f:223 json.dump(results, f, indent=2)224 225 226# ============================================================================227# EXAMPLE 6: Parallel Testing with Multiple Seeds228# ============================================================================229 230from env import TradingEnvironment, MarketDataLoader231from tasks import TaskFactory232from agents import MomentumAgent233import numpy as np234 235task = TaskFactory.create_hard_task()236seeds = [42, 123, 456, 789, 999]237results = []238 239for seed in seeds:240 loader = MarketDataLoader()241 env = TradingEnvironment(task.config, loader)242 agent = MomentumAgent(task.config.symbols)243 244 state, _ = env.reset(seed=seed)245 portfolio_values = [env.get_portfolio_value()]246 247 done = False248 while not done:249 action = agent.act(state, {})250 state, reward, done, info = env.step(action)251 portfolio_values.append(env.get_portfolio_value())252 253 final_return = (portfolio_values[-1] - 100000) / 100000254 results.append(final_return)255 256print(f"\nRobustness Analysis (5 seeds):")257print(f" Mean Return: {np.mean(results)*100:.2f}%")258print(f" Std Dev: {np.std(results)*100:.2f}%")259print(f" Min: {np.min(results)*100:.2f}%")260print(f" Max: {np.max(results)*100:.2f}%")261 262 263# ============================================================================264# EXAMPLE 7: Integration with RL Libraries (Pseudocode)265# ============================================================================266 267"""268# Using with Stable-Baselines3 (install: pip install stable-baselines3)269 270from stable_baselines3 import PPO271from env import TradingEnvironment, MarketDataLoader272from tasks import TaskFactory273import gymnasium as gym274from gymnasium import spaces275import numpy as np276 277class GymWrapper(gym.Env):278 '''Wrap TradingEnvironment for gym compatibility'''279 280 def __init__(self, task):281 self.env = TradingEnvironment(task.config, MarketDataLoader())282 self.observation_space = spaces.Box(283 low=-np.inf, high=np.inf, 284 shape=(64,), dtype=np.float32285 )286 self.action_space = spaces.Discrete(287 len(self.env.config.symbols) * 3 # symbol x [BUY, SELL, HOLD]288 )289 290 def reset(self, seed=None):291 state, _ = self.env.reset(seed=seed)292 return state, {}293 294 def step(self, action_idx):295 # Convert action index to Action object296 num_symbols = len(self.env.config.symbols)297 symbol_idx = action_idx // 3298 order_type_idx = action_idx % 3299 300 from env import Action, OrderType301 symbol = self.env.config.symbols[symbol_idx]302 order_types = [OrderType.BUY, OrderType.SELL, OrderType.HOLD]303 304 action = Action(305 symbol=symbol,306 order_type=order_types[order_type_idx],307 quantity=25.0308 )309 310 state, reward, done, _, info = self.env.step(action)311 return state, reward, done, False, info312 313# Train agent314task = TaskFactory.create_medium_task()315env = GymWrapper(task)316 317model = PPO("MlpPolicy", env, verbose=1)318model.learn(total_timesteps=10000)319 320# Test trained model321obs, _ = env.reset()322for _ in range(250):323 action, _ = model.predict(obs)324 obs, reward, done, _, _ = env.step(action)325 if done:326 break327"""328 329 330# ============================================================================331# EXAMPLE 8: Debugging and Logging332# ============================================================================333 334import logging335 336# Setup detailed logging337logging.basicConfig(338 level=logging.INFO,339 format='%(asctime)s - %(name)s - %(levelname)s - %(message)s'340)341 342from env import TradingEnvironment, MarketDataLoader343from tasks import TaskFactory344from agents import RandomAgent345 346task = TaskFactory.create_easy_task()347loader = MarketDataLoader()348env = TradingEnvironment(task.config, loader)349agent = RandomAgent(task.config.symbols)350 351state, info = env.reset(seed=42)352 353print("Starting episode debug run...")354for step in range(20):355 action = agent.act(state, info)356 state, reward, done, info = env.step(action)357 358 print(f"\nStep {step}:")359 print(f" Action: {action.symbol} {action.order_type.name} x{action.quantity}")360 print(f" Reward: {reward:.4f}")361 print(f" Portfolio: ${info['portfolio_value']:.2f}")362 print(f" Positions: {list(info['positions'].keys())}")363 364 if done:365 print(f"\nEpisode ended at step {step}")366 break367 368 369# ============================================================================370# TESTING CHECKLIST371# ============================================================================372 373"""374Before deploying your agent:375 376✓ Run quick tests:377 python test_quick.py378 379✓ Verify imports:380 from env import *381 from agents import *382 from tasks import *383 384✓ Test all difficulty levels:385 - TaskFactory.create_easy_task()386 - TaskFactory.create_medium_task()387 - TaskFactory.create_hard_task()388 389✓ Test all baseline agents:390 - RandomAgent391 - MomentumAgent392 - TechnicalAgent393 394✓ Test with multiple seeds:395 env.reset(seed=42)396 env.reset(seed=123)397 env.reset(seed=456)398 399✓ Run full episode:400 state, _ = env.reset()401 done = False402 while not done:403 action = agent.act(state, info)404 state, reward, done, info = env.step(action)405 406✓ Check metrics computation:407 metrics = Evaluator.evaluate_episode(...)408 score = metrics.score()409 410✓ Verify no errors:411 python -c "import env; import agents; import tasks; print('✓ OK')"412"""413 414 415if __name__ == "__main__":416 print("QuantitativeTrading-v1 - Quick Start Examples")417 print("Copy-paste any example to get started!")418 print("\nAll examples are self-contained and runnable.")419