Vertex AI | Model Monitoring | Deviation Score

[kaushikpredx]

I was trying to understand the deviation score logic behind the below given categorical variable distribution . As per the documentation the score is calculated with L-infinity distance (https://cloud.google.com/vertex-ai/docs/model-monitoring/overview#:~:text=distance%20score%20is-,cal...) . I tried to manually calculate it gave me something around 0.1 but as you see in the screeshot it says 0.8 

I have given the number of distribution below

Training Distribution (P) : False 711 True 125
Prediction Distribution (Q): False 2802 True 540