VWAP execution algorithm development

VWAP (Volume-Weighted Average Price) Execution Algorithm Development

VWAP execution algorithm is a smarter version of TWAP: instead of distributing volume evenly over time, volume is distributed proportionally to historical trading volume. More orders during high-activity hours, fewer during quiet periods. Goal: execute order at price close to market VWAP over the period.

Difference Between VWAP and TWAP

TWAP divides volume evenly over time. VWAP divides volume proportionally to typical market volume in each time interval.

Example for BTC/USDT (24-hour session): volume during 14:00–16:00 UTC (Europe-America overlap) significantly higher than 02:00–04:00 UTC (minimum activity). VWAP algorithm accounts for this.

Intraday Volume Profile Prediction

def build_volume_profile_intraday(historical_df, n_buckets=48):
    """
    Build average volume for each 30-minute interval of the day
    based on historical data (last 30 days)
    """
    historical_df['time_bucket'] = historical_df.index.time
    avg_volume = historical_df.groupby('time_bucket')['volume'].mean()
    
    # Normalize to unity (weights should sum to 1)
    weights = avg_volume / avg_volume.sum()
    return weights

Execution Algorithm

class VWAPExecutor:
    def __init__(self, symbol, total_qty, duration_hours, exchange):
        self.total_qty = total_qty
        self.volume_weights = self.load_volume_profile(symbol, duration_hours)
        # slice_sizes[i] = qty for i-th interval
        self.slice_sizes = [w * total_qty for w in self.volume_weights]
    
    async def execute_interval(self, interval_idx):
        target_qty = self.slice_sizes[interval_idx]
        # Adapt if volume over past intervals differed from forecast
        actual_volume = await self.get_market_volume(interval_idx)
        expected_volume = self.expected_volumes[interval_idx]
        
        if actual_volume > expected_volume * 1.5:
            # Market more active — increase order
            target_qty *= (actual_volume / expected_volume)
        
        await self.place_order(target_qty)

Participation of Volume (POV)

Participation Rate — alternative approach: execute X% of current market volume.

target_qty_per_interval = market_volume × participation_rate (e.g., 10%)

POV guarantees minimal market impact (no more than 10% of market), but doesn't guarantee execution by deadline on low volume.

Real-Time Adaptation

If current execution lags plan (market moves unfavorably) — algorithm can:

• Execute remaining volume more aggressively
• Temporarily switch to market orders
• Extend time horizon (if allowable)

Benchmark and Reporting

Full execution report on completion: execution timeline, average fill price vs VWAP, slippage per interval.

Stack: Python (asyncio + CCXT), PostgreSQL for execution logs, Grafana for real-time progress visualization.