A Data Science Central Community
Yesterday's expert analysis here decided that Boston has a slight advantage in spite of the finals going to game 7 in Vancouver. We decided to dig a little deeper into the data and try out a quick decision tree analysis using RapidMiner.
We are trying to find out if the data reveals any interesting trends from all the games Vancouver has played at home. Is there a particularly interesting trend whenever they have lost at home? They have lost only 11 out of 40 games (and only 3 of 11 playoff games) at home.
Boston's on-road record is a little more upbeat for their fans and hence the conclusion of a slight advantage mentioned earlier. The question we are addressing is what factors influence their away wins?
Our data set for Canucks included all their home games this season - about 41 samples. The decision tree appears quite uncomplicated, but revealing. To win comfortably, Boston has to resort to real aggressive offense and score no less than 4 goals (see decision tree below). Anything less, i.e. a conservative style of play might not work.
Analysis of a similar dataset for Boston (i.e. all their AWAY games from this season) shows a slightly more complicated decision tree - see below, but indicates the same strategic route. The key variable is conversion rate (goals scored/shots taken). Any rate higher than 9% will ensure that they can take the game away from Vancouver.
In this series Boston's conversion rate has been hovering around 9.3%, so we are tempted to give this last game and the cup to Boston. A few more hours will show how good this model is!