Sunday, July 26, 2020
Statcast Lab: No Nulls Update
Back in 2016, we noticed that about 10% of batted balls had no tracking. If the lack of tracking was random, it would mean it is unbiased (*). But, most of the reasons we weren't tracking was for biased reasons: high popups or sharp grounders. And so if you looked at data like launch angles, having a bunch of low anglers or high anglers missing would lead to wrong conclusions, both at the player and at the league level. So, we implemented a stopgap reasonable solution: no nulls. Every untracked batted ball was given a speed and angle based on whether (a) the cause was biased or unbiased, (b) the stringer marked it as a GB, LD, FB, PU and (c) the actual outcome, single, double, triple, HR, error, out. We were going to enhance that process based on (d) who caught or fielded the ball and even more exciting (e) look at how the players moved to determine where the ball landed. (We tracked the players 98% of the time. It's the ball that was at 90%.) We only did a-c because it satisfied the immediate need, and working on d-e was pushed back in favor of other higher priority items.
(*) Not totally unbiased. For example, we have no tracking in London Games, but London Games had tons of scoring. So, we lack tracking for unbiased reasons, but based on outcomes, the lack of tracking ends up biasing the data. Technically. But, it's only two games.
Now we're in 2020 and the returns on Hawkeye testing in 2019 showed that the lack of tracking was low, and it was mostly unbiased. So, we decided for 2020 to not introduce the no-null solution. If we track it, we report it. If we don't track it, then we don't. This is the same solution we have always had for pitch tracking. It makes as much sense for pitch tracking as it does for 2020 hit tracking: low frequency and unbiased.
Since we've done that, we are revisiting the handling of 2015-2019. To be consistent with 2020, any lack of tracking in 2015-19 for unbiased reasons won't have the no-null solution implemented. And so, what will be left is that the no-nulls solution (i.e., fill-in data generated) will apply only for the 2015-19 data for untracked for biased reasons.
We're figuring out how to present this on Savant, so, stay tuned for how we'll handle it.
Recent comments
Older comments
Page 1 of 150 pages 1 2 3 > Last ›Complete Archive – By Category
Complete Archive – By Date
FORUM TOPICS
Jul 12 15:22 MarcelsApr 16 14:31 Pitch Count Estimators
Mar 12 16:30 Appendix to THE BOOK - THE GORY DETAILS
Jan 29 09:41 NFL Overtime Idea
Jan 22 14:48 Weighting Years for NFL Player Projections
Jan 21 09:18 positional runs in pythagenpat
Oct 20 15:57 DRS: FG vs. BB-Ref
Apr 12 09:43 What if baseball was like survivor? You are eliminated ...
Nov 24 09:57 Win Attribution to offense, pitching, and fielding at the game level (prototype method)
Jul 13 10:20 How to watch great past games without spoilers