This would be a fun data science project. A lot of different testable hypotheses in here.

I made an app if anyone wants to give it a try! . You can rank your favorite national parks and fill in a few other details about when you went, for how long, etc. For now it only collects data but will eventually host fun analyses like consensus rankings.

Join millions who have switched to Grok.

Nice job Cuse team 🍊.

RT @hedgertronic: Heart rate data using @WHOOP from my first ever big league spring training game. 4:35: manager walks out from dugout to….

@RebootMotion @jameshbuffi @justmohsen Anyways, that’s a lot of info already and most of it was more conceptual than technical advice. Maybe this should’ve been a blog post and not a super-long tweet thread.

@RebootMotion @jameshbuffi @justmohsen Another really cool field I’ve been inspired by here is motor control - there are coordination measurement and motor variability features that can be generated from biomechanics data. Here’s a recent one that describes a process for getting at coordination

@RebootMotion @jameshbuffi @justmohsen The Phillies very recently co-authored a paper on in-game grip strength changes, which I hadn’t seen tested before. Think it needs survivorship bias corrections and more data, but there could reasonably be something there

@RebootMotion @jameshbuffi Ok, a few last quick hits - studies show jerk (derivative of acceleration) as potentially useful Just note jerk is 3 derivations away from mocap/CV data and only 1 from IMU. Shoutout to WHOOP colleague @justmohsen on this paper.

There’s been an initial public venture from @RebootMotion / @jameshbuffi:

Anyways, we don’t have to rely on video now that Hawkeye is everywhere. The big “if” is whether park effects/biases can be corrected well enough. At least I hadn't figured out a great way there.

a change in mechanics/pitch traits and vice versa. A good test of this: Second - I believe it’s still more predictive than using measures of load at the same timeframe.

There’s a legitimate counter to this dynamic change idea though - the possibility that it’s all just leakage. For ex. is the change already evidence that an injury occurred? Maybe! 2 thoughts - I think an injury prevents play or impairs performance, and that doesn’t necessitate.

That’s why I’m skeptical of any hope for a long-term injury forecast. These change / variation values can’t be included in year-to-year predictions. This paper is only a commentary but shares a similar message

Maybe the most valuable thing here is the evidence it supplies for this concept of change from an individual’s baseline as an indication of fatigue. I think that’s a key way to think about it - increased variation and/or deviation from norms will be your best predictors.

Before all this in-game Hawkeye data, all people had was video, and I loved this paper when I found it There’s so many cool takeaways here and even good places to build upon (partial pooling would go a long way). I had like a hour+ call with the author AJ

Pitch tracking isn’t very new though, and it’s clear metrics themselves or changes to an individual’s metrics can only go so far towards understanding fatigue. Biomechanics data almost certainly offers more value.

The Twins built off this idea here but focused less on dynamic fatigue detection and more on injury risk year-over-year. This is a good one for calibrating expectations and relative variable importance.

Let’s start with This is a good jumping-off point for the kinds of science I’d pursue to identify fatigue. Nothing crazy technical here but it’s nice to see there are potentially meaningful trends prior to a UCL.

I’ll still preface this advice by saying I think there are tons of issues remaining when it comes to productionalizing an injury risk model so I intend all to come as only a means to instigate conversations with training staff-there’s nothing wrong with just having a conversation.

Ok ok, so workload should be much less of a factor in decision-making. What should? While I still think injury prediction is not very productive, there are a few really interesting offshoots here. I guess this veers more into a monitoring topic.