Real-Time Streaming Moving VWAP

Volume-Weighted Average Price (VWAP) is the preeminent benchmark used by algorithmic execution desks at Tower Research and HRT to assess execution quality and minimize market impact.

Given a continuous chronological stream of trades, each characterized by a tuple $(timestamp, price, volume)$, design an online data structure MovingVWAP that maintains the rolling VWAP over a trailing time window of $W$ seconds: $\text{VWAP}_t = \frac{\sum_{i \in \mathcal{W}_t} P_i \cdot V_i}{\sum_{i \in \mathcal{W}_t} V_i}$ where $\mathcal{W}_t = \{i \mid t - W \le timestamp_i \le t\}$. If the total traded volume in the window is zero, return $0.0$.

Your Task

Implement the MovingVWAP class:

• __init__(window_seconds: float): Initializes the object with the time window parameter $W$.
• add_trade(timestamp: float, price: float, volume: int) -> None: Appends a new executed trade. Timestamps are guaranteed to be non-decreasing.
• get_vwap(current_timestamp: float) -> float: Evicts trades strictly older than $current\_timestamp - W$ and returns the active VWAP rounded to 4 decimal places.

Constraints

• $0.1 \le window\_seconds \le 86400.0$
• $0.0 \le timestamp \le 10^9$, strictly non-decreasing across calls.
• $0.01 \le price \le 10^5$
• $1 \le volume \le 10^7$
• Operations must achieve $\mathcal{O}(1)$ amortized time per call.

Examples

vwap = MovingVWAP(window_seconds=10.0)
vwap.add_trade(1.0, 100.0, 10)
vwap.add_trade(5.0, 102.0, 20)
# At t=6.0: (100*10 + 102*20) / (10 + 20) = 3040 / 30 = 101.3333
vwap.get_vwap(6.0) # returns 101.3333
vwap.add_trade(12.0, 105.0, 30)
# At t=12.0: window is [2.0, 12.0]. Trade at t=1.0 is evicted!
# Active trades: t=5.0 (102*20) and t=12.0 (105*30). Sum PV = 2040 + 3150 = 5190, Sum V = 50. VWAP = 103.8
vwap.get_vwap(12.0) # returns 103.8