How to Implement a Sliding Window Average in Python

Compute the average of the most recent N values in a stream using a bounded deque, efficiently updating the total as new values arrive.

Easy Python 3.9+ Aug 9, 2026 Data pipelines & processing 15 views 0 copies

deque sliding-window streaming statistics

Python code

26 lines

Python 3.9+

from collections import deque


class SlidingWindowAverage:
    def __init__(self, window_size):
        self.window_size = window_size
        self.window = deque(maxlen=window_size)
        self.total = 0

    def add(self, value):
        if len(self.window) == self.window_size:
            self.total -= self.window[0]
        self.window.append(value)
        self.total += value

    def average(self):
        if not self.window:
            return 0.0
        return self.total / len(self.window)


if __name__ == "__main__":
    swa = SlidingWindowAverage(3)
    for value in [10, 20, 30, 40, 50]:
        swa.add(value)
        print(f"Added {value:>2}, average: {swa.average():.1f}")

Output

stdout

Added 10, average: 10.0
Added 20, average: 15.0
Added 30, average: 20.0
Added 40, average: 30.0
Added 50, average: 40.0

How it works

The deque(maxlen=window_size) automatically discards the oldest item when the window is full, keeping only the most recent N values. Each add call adjusts self.total by subtracting the evicted value (if any) and adding the new value, so the average is computed in O(1) time instead of summing the window each time. The average method guards against an empty window by returning 0.0. This pattern is ideal for streaming data where you need a rolling statistic without storing the entire history.

Common mistakes

Forgetting to subtract the evicted element from `self.total` before appending
Using a plain list and slicing `[-window_size:]`, which is O(N) per call
Not handling the empty window case, causing a ZeroDivisionError
Confusing the deque's maxlen behavior with manual padding logic

Variations

Use `statistics.mean` on a `collections.deque` slice if performance is less critical
Implement a ring buffer with a fixed-size list and modulo indexing for lower-level control

Real-world use cases

Tracking the average latency of recent HTTP requests to a service in a metrics agent.
Computing rolling CPU or memory usage averages in a system monitoring script.
Averaging the latest N sensor readings to smooth out noise in an IoT data pipeline.

How to Implement a Sliding Window Average in Python

Python code

Output

How it works

Common mistakes

Variations

Real-world use cases

More from Data pipelines & processing

Tutorials

Quizzes

Python code

Output

How it works

Common mistakes

Variations

Real-world use cases

More from Data pipelines & processing

Keep learning

Tutorials

Quizzes