Hi,
Each Machine Learning job generates different types of results, and the anomaly scores from each of these result types are displayed in the various components of the Anomaly Explorer view (Top Influencers, Anomaly Timeline, anomaly charts and table).
The Overall lane at the top of the Anomaly timeline displays the maximum anomaly scores of the 'bucket influencer' result type. The bucket influencer results are an aggregation of each of the 'influencers' you chose when creating the multi metric job, selected in the 'Key fields' section of the Multi Metric job wizard.
In addition to any 'field' influencers you selected when creating the job, there is a built-in bucket influencer, called 'bucket_time'. This bucket influencer is the aggregation of all the anomaly records in each time bucket.
The calculation behind the bucket_influencer anomaly score displayed in the Overall timeline is more complex than just a simple average of all the individual record anomaly scores, but will have a contribution from the influencer scores in each bucket.
More information on the scores used in the various components of the Anomaly Explorer window can be found in this post, and there is information on the bucket_influencer result type in the docs here.
Hope that helps
Pete