mahmoud20138 Skill: Execution Algo Trading | Domain: trading | Category: execution | Level: advanced Tags:
trading,execution,twap,vwap,iceberg,algo
Execution Algorithm Trading Skill
Overview
Implements institutional-grade execution algorithms used by buy-side desks to minimise market impact and transaction costs when executing large orders.
Python Module
xtrading/skills/execution_algo.py
Stack
- numpy — Numerical computations, binomial tree, random generation
- pandas — VWAP calculation, fill data management
- scipy — Statistical computations
1. TWAP Executor
from datetime import datetime, timedelta
from xtrading.skills.execution_algo import TWAPExecutor
now = datetime.now()
twap = TWAPExecutor(
symbol="EURUSD",
total_qty=100_000,
side="buy",
start_time=now,
end_time=now + timedelta(hours=4),
n_slices=20,
randomise_size=True, # add ±15% size variation
randomise_time=True, # add ±10% timing jitter
)
schedule = twap.build_schedule()
# schedule.n_slices = 20
# schedule.estimated_cost_bps ≈ 3.0
# schedule.slices[i].target_time, .quantity, .order_type
# Simulate against historical prices
import pandas as pd
prices = pd.Series(...) # mid prices with DatetimeIndex
result = twap.simulate_execution(prices, spread_bps=2.0)
# {"avg_fill_price": 1.1005, "twap_benchmark": 1.1003,
# "vs_benchmark_bps": 1.8, "total_cost_bps": 3.8}
2. VWAP Executor
from xtrading.skills.execution_algo import VWAPExecutor
import numpy as np
# Custom intraday volume profile
profile = np.array([0.10, 0.08, 0.06, 0.05, 0.04, 0.04,
0.04, 0.04, 0.05, 0.06, 0.07, 0.08,
0.09, 0.10, 0.10, 0.10])
vwap = VWAPExecutor(
symbol="XAUUSD",
total_qty=50_000,
side="sell",
start_time=now,
end_time=now + timedelta(hours=8),
volume_profile=profile,
participation_cap=0.15, # max 15% of any interval
)
schedule = vwap.build_schedule(avg_interval_volume=10_000)
# schedule.participation_rate ≈ 0.031 (3.1% of daily volume)
# Calculate realised VWAP
import pandas as pd
vwap_price = vwap.calculate_volume_weighted_price(prices, volumes)
3. Implementation Shortfall (Almgren-Chriss)
from xtrading.skills.execution_algo import ISOptimiser
opt = ISOptimiser(
total_qty=100_000,
T_hours=4.0, # execute over 4 hours
sigma=0.02, # daily vol
eta=2.5e-6, # temporary impact coefficient
gamma=1e-7, # permanent impact coefficient
risk_aversion=1e-6, # λ: 0 = minimise IS only
)
result = opt.optimal_trajectory(n_intervals=10)
# {
# "urgency_factor_kappa": 0.000123,
# "expected_shortfall_bps": 4.2,
# "schedule": [
# {"interval": 1, "qty_to_trade": 8234, "remaining_inventory": 91766, "completion_pct": 8.2},
# ...
# ]
# }
# Efficient frontier (trade-off: urgency vs cost)
frontier = opt.efficient_frontier() # DataFrame
4. POV (Percentage of Volume)
from xtrading.skills.execution_algo import POVExecutor
pov = POVExecutor(
symbol="GBPUSD",
total_qty=200_000,
side="buy",
target_pov=0.10, # 10% of market volume
min_slice_qty=1_000,
max_slice_qty=20_000,
)
# Called on each new market volume observation
slice_order = pov.on_market_volume(
interval_volume=50_000,
mid_price=1.2650,
spread_bps=1.5,
)
# Summary after execution
summary = pov.execution_summary()
# {"n_fills": 15, "avg_fill_price": 1.2652, "vs_vwap_bps": 1.2, "completion_pct": 85.3}
5. Iceberg Order
from xtrading.skills.execution_algo import IcebergOrder
ice = IcebergOrder(
symbol="XAUUSD",
total_qty=10_000,
display_qty=500, # only 500 oz visible at a time
side="buy",
limit_price=2000.0,
randomise_display=True, # vary display size ±10%
)
# Process fills
status = ice.on_fill(fill_qty=500, fill_price=2000.5)
# {"filled": 500, "total_filled": 500, "remaining": 9500,
# "visible": 487, "reserve": 9013, "refreshed": True, "complete": False}
print(ice.summary)
6. Slippage & Market Impact Analysis
from xtrading.skills.execution_algo import SlippageAnalyser, MarketImpactModel
# Post-trade slippage decomposition
decomp = SlippageAnalyser.decompose(
arrival_price=1.1000,
avg_fill_price=1.1012,
vwap_benchmark=1.1008,
twap_benchmark=1.1005,
side="buy",
spread_bps=2.0,
)
# {"implementation_shortfall_bps": 10.9, "market_impact_bps": 8.9,
# "spread_cost_bps": 2.0, "grade": "B (Acceptable)"}
# Pre-trade impact estimate
impact = SlippageAnalyser.estimate_market_impact(
qty=50_000, adv=2_000_000, price=1.1000,
volatility_daily=0.008, side="buy", model="sqrt"
)
# {"impact_bps": 5.3, "participation_rate": 0.025}
# Full cost model (Almgren-Chriss components)
model = MarketImpactModel(sigma=0.01, adv=1_000_000,
bid_ask_spread=0.0002, price=1.1000)
cost = model.total_cost(qty=100_000, execution_time_hours=2.0)
# {"permanent_impact_bps": 2.8, "transient_impact_bps": 1.9,
# "spread_cost_bps": 0.9, "total_cost_bps": 5.6}
7. TCA Report
from xtrading.skills.execution_algo import TCAReport
import pandas as pd
fills = pd.DataFrame({
"timestamp": [...],
"qty": [1000] * 20,
"fill_price": [...],
"mid_price": [...],
})
report = TCAReport(
symbol="EURUSD", side="buy", fills=fills,
arrival_price=1.1000, vwap=1.1005, twap=1.1003,
algorithm="VWAP", benchmark="vwap"
)
d = report.to_dict()
# {
# "total_cost_bps": 4.2,
# "quality_score": 91.6,
# "grade": "A (Good)",
# "implementation_shortfall_bps": 2.3,
# }
Decision Framework
| Order Size (% ADV) | Recommended Algorithm | Typical Cost (bps) |
|---|---|---|
| < 1% | Market / Limit | 1–2 |
| 1–5% | TWAP (1–2h) | 2–4 |
| 5–15% | VWAP (full day) | 4–8 |
| 15–30% | IS + POV | 8–15 |
| > 30% | Iceberg + multi-day | 15–30 |
Usage Conventions
- qty — shares, lots, or contracts (consistent units throughout)
- adv — average daily volume in the same units as qty
- sigma — daily volatility as decimal (0.01 = 1%)
- spread_bps — round-trip spread cost, not half-spread
- TCA benchmark — use VWAP for passive strategies, arrival for aggressive