The .holt() function basically samples the series data from the beginning and then forecasts the future trend of series data using different optional parameters. We can also use it for anomaly detection. It has different parameters such as alpha, which ranges from 0 to 1; if we increase the alpha value, the new series will closely follow the original series, and if we lower the alpha value, it will make the series smoother.

Then comes the beta value, which also ranges from 0 to 1, and increasing it will make the rising/falling line longer, while decreasing the beta value will make this quickly learn the new trends. After beta, we have gamma, which also ranges from 0 to 1, and if we increase it, it will give more importance to the recent reasons; if we lower it, then it will give less importance to the recent series data that is similar to giving more importance to the historical data.

Then, we can provide the season such as the duration to pick for repeating, such as 1w for one week. If we provide the season, then we can also set the last parameter, which is a sample to set the number of seasons to pick for data sampling. The following expression shows a simple .holt() function implementation:

.es(index=metricbeat*, metric=count:event.duration).label('event duration').holt(0.1,1), .static(50, "Threshold Value").color('red')

In the preceding expression, we have added the .holt() function after the .es() function to see the trend; the following screenshot shows the actual chart created using the preceding expression:

In the preceding screenshot, we can see the current trend in the form of a smooth representation of the chart.