Expected Learning – Week 1 Expected Learning by Eddy Tabone - June 14, 2021February 15, 20253 To view this content, you must be a member of Expected's Patreon at $1 or more Unlock with PatreonAlready a qualifying Patreon member? Refresh to access this content.
great idea. I look forward to this. One of my initial questions is why are is this called analytics for hockey and other sports, but sabermetrics for baseball? haha, just kidding. Actually, specifically for Buffalo I guess it should be called “Sabremetrics”, no? Seriously, what do you feel are some limitations to analytics in hockey right now? I know the data contained within them is static, but perhaps how one analyzes it might have a bias. A specific situation that occurred this past season had me thinking about this. Florida traded for Sam Bennett. Communicating with some analytic pundits that are Panthers fans, their initial reaction to the trade was “meh”. Sure, he had value, but wasn’t really expected to move the needle for them. But then a funny thing happened. He absolutely flourished with them. Now, I may be remembering this incorrectly, but Bennett’s “charts” actually improved once he joined his new team. (I could be wrong, so this may throw this whole argument out the window). So there is potentially a bias a team my have against acquiring a certain player if they have less than favourable analytics, when all said player may need is a different environment. Would you agree with this statement? Are there certain analytics that can be used to mitigate these situations…for better or for worse? Thanks…
So for the main question, I wouldn’t call it a limitation, but tracking data is the next frontier. The amount of information that will be able to be recorded is going to skyrocket, and we won’t know for a long time how much we will truly be able to get out of it (More on that later…) What’s funny about this example is at First Line we’re working with a client on a lineup-related project in basketball that is running into the same problem. Lineup interactions are a frontier that will take awhile before we *really* get a good idea how archetypes interact with each other and how players perform in and out of their archetype. Not only will player tracking need to be able to encompass the impact of all 12 players on the ice (in the hockey case) at once (right now we really mostly have off puck and then positioning frozen when an event happens with what little tracking data has been publicized), but they will also need to acquire a fair amount of games to get distinguishable clarity. And even with that, with everything in statistics, you’ll never know the true formula, just more accurate estimates when there are this many interactions. So with current evaluation, you are limited to how players produce individually and as part of a line. So in the Bennett example, you can assign identities to player types and look into how Bennett plays with different types of lines to get a general idea of what may happen, and then stick to what would happen over the course of a season instead of against different defenses. It’s both a pro and a con of hockey that positions aren’t really that different compared to other sports where this kind of analysis would be occurring. A pro because you can look at individual talent with results and not archetypes and not be too far off, but the con being that it’s going to probably take the longest to determine what those true archetypes are and are going to keep the range of expectations higher longer. Hope that all helps.
Thanks for the response, I appreciate it. I look forward to what the more advanced tracking data will provide. If it’ll be more games like Colorado-Vegas, then we will all benefit.