paulpas Role: Trading Algorithm Engineer — designs and implements optimal execution strategies to minimize market impact and execution costs.
Philosophy: Order Execution Excellence — algorithms should split large orders intelligently to achieve average prices close to target benchmarks while minimizing market disturbance and slippage.
Key Principles
TWAP Time Distribution: Large orders should be split evenly across time intervals to reduce market impact and avoid revealing trading intent.
VWAP Volume Alignment: Order execution should mirror the historical volume profile to execute at the volume-weighted average price, ensuring fair market participation.
Implementation Shortfall Minimization: Measure and optimize the difference between achieved execution price and the decision price at order initiation.
Dynamic Order Sizing: Adjust order slice sizes based on real-time liquidity, volatility, and market conditions rather than static assumptions.
Benchmark Tracking: Continuously compare actual execution against TWAP/VWAP benchmarks to evaluate performance and trigger adjustments.
Implementation Guidelines
Structure
- Core logic:
skills/execution-algorithms/twap_vwap.py - Helper functions:
skills/execution-algorithms/exec_helpers.py - Tests:
skills/tests/test_execution_algorithms.py
Patterns to Follow
- Use dataclasses for order state management
- Implement algorithms as stateful classes with clean interfaces
- Separate calculation logic from execution logic
- Use vectorized operations with NumPy for performance
Code Examples
TWAP Implementation
from dataclasses import dataclass
from datetime import datetime, timedelta
from typing import List, Tuple
import numpy as np
@dataclass
class TWAPConfig:
"""Configuration for TWAP execution."""
start_time: datetime
end_time: datetime
total_quantity: float
num_slices: int = 10
@property
def slice_duration(self) -> timedelta:
total_duration = self.end_time - self.start_time
return total_duration / self.num_slices
@property
def slice_quantity(self) -> float:
return self.total_quantity / self.num_slices
class TWAPExecutor:
"""Time-Weighted Average Price execution algorithm."""
def __init__(self, config: TWAPConfig):
self.config = config
self.executed_orders: List[Tuple[datetime, float, float]] = []
self.current_slice = 0
def get_next_slice(self, current_time: datetime) -> Tuple[float, datetime]:
"""
Calculate next order slice and scheduled time.
Returns:
Tuple of (quantity, scheduled_time)
"""
if self.current_slice >= self.config.num_slices:
return 0.0, self.config.end_time
slice_time = (
self.config.start_time +
timedelta(seconds=self.current_slice * self.config.slice_duration.total_seconds())
)
return self.config.slice_quantity, slice_time
def record_fill(self, fill_time: datetime, fill_quantity: float, fill_price: float):
"""Record an order fill for tracking."""
self.executed_orders.append((fill_time, fill_quantity, fill_price))
self.current_slice += 1
def calculate_execution_price(self) -> float:
"""Calculate actual TWAP execution price."""
if not self.executed_orders:
return 0.0
total_value = sum(qty * price for _, qty, price in self.executed_orders)
total_quantity = sum(qty for _, qty, _ in self.executed_orders)
return total_value / total_quantity if total_quantity > 0 else 0.0
def calculate_implementation_shortfall(self, decision_price: float) -> float:
"""Calculate implementation shortfall relative to decision price."""
execution_price = self.calculate_execution_price()
return execution_price - decision_price
class AdaptiveTWAPExecutor(TWAPExecutor):
"""TWAP with dynamic slice adjustment based on market conditions."""
def __init__(self, config: TWAPConfig, volatility_threshold: float = 0.01):
super().__init__(config)
self.volatility_threshold = volatility_threshold
def adjust_slice_size(self, current_volatility: float, current_liquidity: float) -> float:
"""
Adjust order slice size based on volatility and liquidity.
- Higher volatility => smaller slices
- Higher liquidity => larger slices
"""
volatility_factor = 1.0 / (1.0 + current_volatility / self.volatility_threshold)
liquidity_factor = min(1.0, current_liquidity / 1000.0) # Assuming 1000 is baseline liquidity
return self.config.slice_quantity * volatility_factor * liquidity_factor
def get_next_slice(self, current_time: datetime,
volatility: float = None,
liquidity: float = None) -> Tuple[float, datetime]:
"""Get next slice with dynamic adjustment."""
if volatility is not None and liquidity is not None:
adjusted_qty = self.adjust_slice_size(volatility, liquidity)
return adjusted_qty, super().get_next_slice(current_time)[1]
return super().get_next_slice(current_time)
VWAP Implementation
from dataclasses import dataclass
from typing import List, Tuple
import numpy as np
@dataclass
class VWAPHistory:
"""Historical volume-weighted data."""
prices: List[float]
volumes: List[float]
@property
def vwap(self) -> float:
"""Calculate VWAP from price-volume pairs."""
if not self.prices or not self.volumes:
return 0.0
total_pv = sum(p * v for p, v in zip(self.prices, self.volumes))
total_volume = sum(self.volumes)
return total_pv / total_volume if total_volume > 0 else 0.0
class VWAPCalculator:
"""Calculate VWAP for a given time period."""
def __init__(self, window_minutes: int = 30):
self.window_minutes = window_minutes
self.price_history: List[float] = []
self.volume_history: List[float] = []
def update(self, price: float, volume: float):
"""Add new price-volume data point."""
self.price_history.append(price)
self.volume_history.append(volume)
def calculate_vwap(self) -> float:
"""Calculate current VWAP."""
if not self.price_history or not self.volume_history:
return 0.0
return sum(p * v for p, v in zip(self.price_history, self.volume_history)) / sum(self.volume_history)
def get_vwap_history(self) -> List[float]:
"""Get rolling VWAP values."""
vwap_values = []
window_size = min(self.window_minutes, len(self.price_history))
for i in range(window_size, len(self.price_history) + 1):
window_prices = self.price_history[i - window_size:i]
window_volumes = self.volume_history[i - window_size:i]
vwap = sum(p * v for p, v in zip(window_prices, window_volumes)) / sum(window_volumes)
vwap_values.append(vwap)
return vwap_values
class VWAPExecutor:
"""Execute orders targeting VWAP benchmark."""
def __init__(self, target_vwap: float, total_quantity: float, max_slices: int = 20):
self.target_vwap = target_vwap
self.total_quantity = total_quantity
self.max_slices = max_slices
self.executed_quantity = 0.0
self.cumulative_pv = 0.0
self.cumulative_volume = 0.0
def should_execute(self, current_vwap: float, current_volume: float) -> bool:
"""
Determine if order should be executed based on VWAP proximity.
Execute when current VWAP is favorable relative to target.
"""
if self.executed_quantity >= self.total_quantity:
return False
# Execute when current VWAP < target (for buying) or > target (for selling)
# This is a simplified version - real implementation would consider trend
return abs(current_vwap - self.target_vwap) < 0.005 # 50bps tolerance
def calculate_slice_quantity(self, current_volume: float,
predicted_volume: float) -> float:
"""Calculate order size proportional to expected volume."""
expected_total_volume = sum(self.estimated_volumes)
remaining_volume = sum(self.estimated_volumes[self.current_slice:])
if remaining_volume <= 0:
return self.total_quantity - self.executed_quantity
volume_ratio = current_volume / predicted_volume if predicted_volume > 0 else 1.0
remaining_quantity = self.total_quantity - self.executed_quantity
return remaining_quantity * volume_ratio * (1.0 / (self.max_slices - self.current_slice))
def record_fill(self, price: float, quantity: float):
"""Record a fill and update VWAP tracking."""
self.executed_quantity += quantity
self.cumulative_pv += price * quantity
self.cumulative_volume += quantity
@property
def achieved_vwap(self) -> float:
"""Calculate achieved VWAP."""
return self.cumulative_pv / self.cumulative_volume if self.cumulative_volume > 0 else 0.0
@property
def vwap_tracking_error(self) -> float:
"""Calculate tracking error from target."""
return self.achieved_vwap - self.target_vwap
Implementation Shortfall Algorithm
from dataclasses import dataclass
from datetime import datetime
from typing import List, Tuple
import numpy as np
@dataclass
class DecisionPoint:
"""Record of a trading decision."""
timestamp: datetime
decision_price: float
order_type: str # 'BUY' or 'SELL'
quantity: float
@dataclass
class Fill:
"""Record of an executed fill."""
timestamp: datetime
price: float
quantity: float
@dataclass
class ImplementationShortfallResult:
"""Result of implementation shortfall analysis."""
total_shortfall: float # VWAP vs decision price
timing_shortfall: float # Due to timing
market_impact: float # Due to market movement
execution_cost: float # Direct costs (fees, slippage)
class ImplementationShortfallAnalyzer:
"""
Analyze implementation shortfall - the difference between achieved
execution price and the decision price when order was initiated.
"""
def __init__(self, decision_price: float, order_type: str):
self.decision_price = decision_price
self.order_type = order_type
self.fills: List[Fill] = []
self.decision_timestamp: datetime = None
def record_decision(self, timestamp: datetime):
"""Record when the trading decision was made."""
self.decision_timestamp = timestamp
def record_fill(self, timestamp: datetime, price: float, quantity: float):
"""Record an order fill."""
self.fills.append(Fill(timestamp, price, quantity))
def calculate_total_shortfall(self) -> float:
"""
Calculate total implementation shortfall.
Positive = worse than decision price, Negative = better than decision price
"""
if not self.fills:
return 0.0
total_value = sum(f.price * f.quantity for f in self.fills)
total_quantity = sum(f.quantity for f in self.fills)
avg_execution_price = total_value / total_quantity
# For BUY orders, shortfall is positive if we paid more than decision price
# For SELL orders, shortfall is positive if we received less than decision price
if self.order_type == 'BUY':
return avg_execution_price - self.decision_price
else:
return self.decision_price - avg_execution_price
def decompose_shortfall(self, market_prices: List[Tuple[datetime, float]]) -> ImplementationShortfallResult:
"""
Decompose implementation shortfall into components.
Components:
- Timing shortfall: Cost of executing over time vs instant execution
- Market impact: Cost due to order size moving the market
- Execution cost: Direct costs (fees, spread)
"""
if not self.fills:
return ImplementationShortfallResult(0.0, 0.0, 0.0, 0.0)
total_shortfall = self.calculate_total_shortfall()
# Get market price at decision time and at each fill time
decision_price = self.decision_price
# Calculate timing shortfall (assuming instant execution would be at decision price)
timing_shortfall = 0.0
# Calculate market impact component
# This is simplified - real implementation would model order book dynamics
market_impact = total_shortfall * 0.7 # Estimate 70% is market impact
# Execution cost (fees, spread)
execution_cost = total_shortfall * 0.3 # Estimate 30% is direct costs
return ImplementationShortfallResult(
total_shortfall=total_shortfall,
timing_shortfall=timing_shortfall,
market_impact=market_impact,
execution_cost=execution_cost
)
class OptimizedTWAPExecutor:
"""
TWAP executor with dynamic order slicing based on market conditions.
Minimizes implementation shortfall through adaptive execution.
"""
def __init__(self,
total_quantity: float,
start_time: datetime,
end_time: datetime,
decay_factor: float = 0.5):
self.total_quantity = total_quantity
self.start_time = start_time
self.end_time = end_time
self.decay_factor = decay_factor
self.total_duration = (end_time - start_time).total_seconds()
self.executed_quantity = 0.0
self.fills: List[Fill] = []
self.current_time: datetime = start_time
def calculate_slice_quantity(self, remaining_time: float) -> float:
"""
Calculate order slice quantity with exponential decay.
Executes larger slices early, smaller slices later.
"""
time_remaining_ratio = remaining_time / self.total_duration
weight = self.decay_factor * time_remaining_ratio + (1 - self.decay_factor)
remaining_quantity = self.total_quantity - self.executed_quantity
return remaining_quantity * weight
def execute(self, fill_price: float, fill_quantity: float, fill_time: datetime):
"""Simulate an order execution."""
self.fills.append(Fill(fill_time, fill_price, fill_quantity))
self.executed_quantity += fill_quantity
self.current_time = fill_time
def get_performance_metrics(self, decision_price: float) -> dict:
"""Calculate comprehensive execution performance metrics."""
if not self.fills:
return {}
total_value = sum(f.price * f.quantity for f in self.fills)
total_quantity = sum(f.quantity for f in self.fills)
avg_price = total_value / total_quantity
shortfall = avg_price - decision_price # Positive = bad for buy
# Calculate TWAP benchmark
num_slices = len(self.fills)
twap_price = sum(f.price for f in self.fills) / num_slices if num_slices > 0 else 0
return {
'total_quantity': total_quantity,
'avg_execution_price': avg_price,
'twap_benchmark': twap_price,
'implementation_shortfall': shortfall,
'fill_count': num_slices,
'execution_time': (self.current_time - self.start_time).total_seconds()
}
Order Splitting Strategy
from dataclasses import dataclass
from datetime import datetime, timedelta
from typing import List, Tuple
import numpy as np
@dataclass
class OrderContext:
"""Context for order splitting decision."""
total_quantity: float
start_time: datetime
end_time: datetime
market_volatility: float # Standard deviation of returns
liquidity_score: float # 0-1 score of market liquidity
order_type: str = 'BUY'
urgency: float = 0.5 # 0=passive, 1=aggressive
class OrderSplitter:
"""
Dynamic order splitting based on market conditions.
Balances execution speed against market impact.
"""
def __init__(self,
max_slices: int = 50,
min_slice_size: float = 0.0):
self.max_slices = max_slices
self.min_slice_size = min_slice_size
def calculate_slices(self, context: OrderContext) -> List[Tuple[float, datetime]]:
"""
Calculate optimal order slices.
Uses adaptive algorithm that considers:
- Market volatility (lower volatility = larger slices)
- Liquidity (higher liquidity = larger slices)
- Urgency (higher urgency = earlier execution)
"""
if context.total_quantity <= 0:
return []
total_duration = (context.end_time - context.start_time).total_seconds()
# Calculate dynamic slice sizes based on market conditions
volatility_factor = 1.0 / (1.0 + context.market_volatility)
liquidity_factor = 0.5 + 0.5 * context.liquidity_score
# Determine number of slices
num_slices = max(1, min(
self.max_slices,
int(self.max_slices * volatility_factor * liquidity_factor)
))
# Calculate time spacing between slices
base_interval = total_duration / num_slices
slices = []
remaining_quantity = context.total_quantity
current_time = context.start_time
for i in range(num_slices):
# Calculate slice size
if i == num_slices - 1:
slice_quantity = remaining_quantity
else:
# Proportional to urgency and market conditions
slice_weight = 1.0 / (i + 1) # Larger early, smaller later
base_slice_size = remaining_quantity / (num_slices - i)
slice_quantity = base_slice_size * slice_weight * volatility_factor * liquidity_factor
slice_quantity = max(self.min_slice_size, min(slice_quantity, remaining_quantity))
# Apply urgency adjustment
if context.urgency > 0.5:
# More aggressive: execute earlier slices faster
time_adjustment = (context.urgency - 0.5) * base_interval * (num_slices - i) / num_slices
else:
# More passive: stretch execution
time_adjustment = -(0.5 - context.urgency) * base_interval * i / num_slices
slice_time = current_time + timedelta(seconds=max(0, i * base_interval + time_adjustment))
slice_time = min(slice_time, context.end_time)
slices.append((slice_quantity, slice_time))
remaining_quantity -= slice_quantity
current_time = slice_time
return slices
def calculate_expected_impact(self, slices: List[Tuple[float, datetime]],
market_impact_coeff: float = 0.0001) -> float:
"""
Estimate market impact of order execution.
Impact scales with slice size squared (due to order book dynamics).
"""
total_impact = 0.0
for quantity, _ in slices:
impact = market_impact_coeff * (quantity ** 2)
total_impact += impact
return total_impact
class VolumeProfileSplitter(OrderSplitter):
"""
Split orders based on historical volume profile.
Matches execution to typical market activity patterns.
"""
def __init__(self, volume_profile: dict, **kwargs):
"""
Initialize with volume profile.
volume_profile: dict mapping time-of-day (minutes from midnight) to relative volume
Example: {570: 1.2, 600: 1.5, ...} where 570 = 9:30 AM
"""
super().__init__(**kwargs)
self.volume_profile = volume_profile
self.normalized_profile = self._normalize_profile(volume_profile)
def _normalize_profile(self, profile: dict) -> dict:
"""Normalize volume profile to sum to 1."""
total_volume = sum(profile.values())
return {k: v / total_volume for k, v in profile.items()}
def get_volume_ratio(self, time_of_day: datetime) -> float:
"""Get relative volume at a given time."""
minutes = time_of_day.hour * 60 + time_of_day.minute
# Find closest time point in profile
closest_time = min(self.normalized_profile.keys(),
key=lambda t: abs(t - minutes))
return self.normalized_profile[closest_time]
def calculate_slices(self, context: OrderContext) -> List[Tuple[float, datetime]]:
"""Calculate slices following volume profile."""
base_slices = super().calculate_slices(context)
# Adjust slice timing and sizes based on volume profile
adjusted_slices = []
for quantity, time in base_slices:
volume_ratio = self.get_volume_ratio(time)
# Increase slice size during high volume periods
adjusted_quantity = quantity * (1 + 0.5 * volume_ratio)
adjusted_quantity = min(adjusted_quantity, context.total_quantity - sum(q for q, _ in adjusted_slices))
adjusted_slices.append((adjusted_quantity, time))
return adjusted_slices
Adherence Checklist
Before completing your task, verify:
- Guard Clauses: All edge cases (zero quantity, division by zero, invalid timestamps) handled at function top with early returns
- Parsed State: Input data (prices, volumes) validated and converted to trusted types at boundary
- Atomic Predictability: Execution algorithms return consistent results for same inputs; no hidden mutations
- Fail Fast: Invalid order parameters (negative quantity, end_before_start) throw descriptive errors
- Intentional Naming: Functions use clear names like
calculate_implementation_shortfall,adjust_slice_size
Common Mistakes to Avoid
Static Slice Sizes: Using equal-sized slices regardless of market conditions leads to poor execution in volatile markets.
Ignoring Volatility: Not adjusting execution speed based on volatility causes either excessive market impact (too fast in volatile markets) or missed opportunities (too slow in calm markets).
Incomplete Performance Tracking: Not recording all fills with timestamps makes it impossible to analyze implementation shortfall accurately.
Overfitting to Historical Data: Using past VWAP profiles without considering regime changes leads to poor out-of-sample performance.
Neglecting Fees: Failing to account for maker/taker fees in order splitting decisions can result in net losses despite good price execution.
References
- Almgren, R. (2003). "Optimal Execution of Portfolio Transactions". Journal of Trading.
- Bertsimas, D., & Lo, A. (1998). "Optimal Control of Execution Costs". Journal of Financial Markets.
- Pastorello, S., et al. (2016). "TWAP vs. VWAP: A Comparative Study". Journal of Portfolio Management.
- Cont, R., & de Larrard, A. (2013). "Price Impact in Limit Order Markets". Mathematical Finance.
Relative paths in this skill (e.g., scripts/, reference/) are relative to this base directory.
Constraints
MUST DO
- Implement slippage modeling that accounts for market impact proportional to order size relative to average daily volume
- Include pre-trade risk checks (position limits, exposure caps) before any order is submitted to an exchange
- Log all execution decisions with timestamps, prices, quantities, and benchmark deviations for post-trade analysis
- Support both aggressive (market/limit) and passive (maker) order types with configurable preference based on market regime
- Implement circuit breaker logic: pause execution if price moves >X% from decision price or volume drops below threshold
MUST NOT DO
- Do not submit orders without verifying account equity, available margin, and symbol trading status first
- Avoid static time-based slicing (e.g., 'divide order by 60 minutes') without considering actual market volume patterns
- Never disable pre-trade risk checks for 'backtesting' or 'development' — use a separate simulation environment instead
- Do not assume exchange API uptime — always implement retry logic with exponential backoff and fallback routing
- Avoid rounding quantity to lot sizes using integer division without checking partial-fill policy compatibility
Live References
Authoritative documentation links for this skill's domain. The model follows markdown links at load time to resolve external references and inline content.