• 24-25 season,  predictions

    Making score predictions

    A group of four of us make a prediction on every Cardiff Devils home game. We used to put £1 in a pot each for every prediction, and the winner would collect the sum at that point. Given that was 4 predictions per game and you might play around 35 games with Cup and Playoffs, you would think someone would be winning regularly. That was not the case – the pot would regularly get beyond £40 and the end of the season would arrive with no winner so we would share it out. Now it’s likely that we are just bad at making predictions, but it also highlights an issue…

  • 23-24 season,  stats

    Nothing really matters

    In their book Hockey Analytics, Stephen Shea and Christopher Baker make the point that ice hockey is a complex sport for analytics. Baseball is largely decided by one-on-one matchups which simplifies comparison, basketball is high scoring which gives a greater base for analysis, football has limited substitutions and is largely played at full strength. Hockey is relatively low-scoring (although thankfully higher than football), has continuous substitution, frequent penalty periods and is played as a team. And it’s on ice. This makes it difficult statistically to extract meaningful measures, for example, possession is not as closely correlated to scoring as it is in basketball, say. Which means, hey why bother? Sometimes…

  • closewin,  league

    Close Win percentage

    During the last NHL playoffs, I worked on a predictive model which I’ll cover in another post (it was, – surprise – not very predictive). One of the factors I tried to capture in that, which I thought might be especially relevant in the playoffs was, for want of a better word – grit. I interpreted this as the percentage of games won in overtime. On reflection I think grit is possibly an important factor in playoff hockey, if we think of it as the ability to scrape a win in close games. It can reflect team spirit, belief, squad depth and other factors which can be difficult to measure…

  • 24-25 season,  league

    Room for improvement

    Looking at the data for the EIHL 23-24 season, I thought I’d see if there were areas that the data alone suggests could be improved for each team. A couple of caveats first – this is very broad stroke data, because the EIHL don’t have an API so I had to enter it all manually, and the data won’t tell us lots of pertinent information (for example the tragedy for the Panthers). Also I’m not basing it off watching the teams (although I have watched all of them several times), just what a raw look at the data tells us. No shouting now, it’s just for fun. So working from…