tag:blogger.com,1999:blog-8199936405783540709.post5319433208381613462..comments2023-03-14T04:00:21.712-04:00Comments on Rink Stats: Win probabilities metric 1.0Anonymoushttp://www.blogger.com/profile/06654463214049859090noreply@blogger.comBlogger6125tag:blogger.com,1999:blog-8199936405783540709.post-179760019472565052014-04-15T10:41:50.470-04:002014-04-15T10:41:50.470-04:00Oh I see what you mean. In some sense my model alr...Oh I see what you mean. In some sense my model already has this implicitly built into it. I average over all teams (and over a few years) to get the probability of winning at each intermission, given the current score. The number of NHL.com are just averaging over all possible scores and then breaking it out by team and season. I've intentionally tried to avoid including team-specific statistics in my metric because I want to keep it so that you can compare the metric between games with totally different teams. I have gotten a lot of feedback from people wanting to see inclusion of Vegas odds, so if I made a separate version of the metric based on that, I could probably include some of this win% stuff.Anonymoushttps://www.blogger.com/profile/06654463214049859090noreply@blogger.comtag:blogger.com,1999:blog-8199936405783540709.post-22072783554513955672014-04-15T10:26:20.713-04:002014-04-15T10:26:20.713-04:00The nhl calls it win%, but i think it's simila...The nhl calls it win%, but i think it's similar to probability of winning: <br />http://www.nhl.com/ice/teamstats.htm?fetchKey=20142ALLSAAAll&sort=winPctAfterLead1P&viewName=recordWhenLeading<br />Phileohttps://www.blogger.com/profile/03489790780839086228noreply@blogger.comtag:blogger.com,1999:blog-8199936405783540709.post-49888656611053505212014-04-15T10:02:12.795-04:002014-04-15T10:02:12.795-04:00Thanks! I've actually never seen that before. ...Thanks! I've actually never seen that before. Do you have a link to where that statistic is?Anonymoushttps://www.blogger.com/profile/06654463214049859090noreply@blogger.comtag:blogger.com,1999:blog-8199936405783540709.post-7115241735866755032014-04-15T01:12:16.512-04:002014-04-15T01:12:16.512-04:00Very interesting stuff.
nhl.com has Probability of...Very interesting stuff.<br />nhl.com has Probability of winning when leading/trailing after 1/2 periods - Is it possible to incorporate this into your model?Phileohttps://www.blogger.com/profile/03489790780839086228noreply@blogger.comtag:blogger.com,1999:blog-8199936405783540709.post-71909903647798550962014-04-14T00:42:43.188-04:002014-04-14T00:42:43.188-04:00Thanks! I keep going back and forth on whether to ...Thanks! I keep going back and forth on whether to add the Vegas line into it at the beginning of each game. The advantage to the way I have it now is that you can compare the metric between games, and you can average across games to get a sense of how well a team is playing overall. If I added in the Vegas odds then I'd lose this because good teams would be favorites more often and this would create a feedback loop in the metric which would make them look better and better. And bad teams would be in a loop the other direction, they'd look bad in the metric and their Vegas odds would look bad making them look even worse on the metric. I've considered having two versions of the metric for this exact reason though.Anonymoushttps://www.blogger.com/profile/06654463214049859090noreply@blogger.comtag:blogger.com,1999:blog-8199936405783540709.post-22570117115530063692014-04-14T00:27:13.287-04:002014-04-14T00:27:13.287-04:00This is really cool. Do you think you should imple...This is really cool. Do you think you should implement the Vegas line in there somehow? That way each game doesn't start right at 50/50. Just a thought Anonymoushttps://www.blogger.com/profile/02583309335354795728noreply@blogger.com