trend

Input Stack:

window: Duration

expr: TimeSeriesExpr

⇨

Output Stack:

TimeSeriesExpr

Warning

Deprecated: Use :rolling-mean instead. This operator exists for backwards compatibility but new expressions should use the more flexible rolling operators.

Compute a moving average over a time-based window. This operator smooths time series data by averaging values within a sliding time window, which can help reduce noise and identify trends.

Parameters¶

expr: The time series expression to apply the moving average to
window: Time duration for the averaging window (e.g., 5m, PT20M)

Behavior¶

Window requirement: Emits NaN until the full window duration has data
NaN handling: Input NaN values are treated as zeros in the calculation
Window sizing: Window duration is rounded down to fit evenly within step size boundaries
Edge cases: If step size > window duration, the operation becomes a no-op

Window Sizing Rules¶

The effective window size depends on the relationship to step size:

Window	Step Size	Effective Window	Datapoints
5m	1m	5m	5
5m	2m	4m	2
2m	5m	No-op	1

Examples¶

5-minute moving average:

5 Minutes

:random,
PT5M,:trend

20-minute moving average:

20 Minutes

:random,
20m,:trend

Sample Calculation¶

Example showing how :trend processes data:

Input	2m,:trend	5m,:trend
0	NaN	NaN
1	0.5	NaN
-1	0.0	NaN
NaN	-0.5	NaN
0	0.0	0.0
1	0.5	0.2
2	1.5	0.4
1	1.5	0.8
1	1.0	1.0
0	0.5	1.0

Migration Guide¶

Replace :trend with :rolling-mean for similar functionality:

# Old (deprecated):
name,cpu,:eq,:sum,5m,:trend

# New (recommended):
name,cpu,:eq,:sum,5,:rolling-mean

:rolling-mean - Recommended replacement (uses datapoint count instead of time)
:rolling-max - Rolling maximum over datapoint windows
:rolling-min - Rolling minimum over datapoint windows
:sdes - More sophisticated smoothing with trend detection