Horsehide Harry
<B>Mr. Free Trade</B>
Not setting the bar very high are you?
Matt Kemp
- Thread starter jsebe10
- Start date
Not setting the bar very high are you?
goldfly
<B>if my thought dreams could be seen</B>
nice
bravesfanMatt
Steve Harvey'd
good lord, I need to proof read when posting on my phone.. What the hell "contestant"..
The point is to see how the eye test compares.
I'm not sure I understand the point, either. Statcast measures everything you would be looking for when watching yourself, it just does it much more accurately. It will tell you where a player is positioned, where the ball is hit, how hard the ball is hit, how long it is in the air, the distance the player covers, the speed with which the player covers that distance, and on and on.
How you interpret and use the Statcast data is the question, not whether the Statcast measurements are accurate.
nsacpi
Expects Yuge Games
It would be useful to have alternative facts to compare with the data
Luckily for us, we are graced by the presence of several regular posters who ONLY rely on alternative facts. We are up to our ears in alternative facts here at Chop Country!
nsacpi
Expects Yuge Games
we are fortunate indeed...blessed I would say
Not exactly.
The algorithms which power Statcast simply can't accurately account for external factors like the exact weather conditions in the park the moment the ball is put in play, crowd noise levels (sound off the bat), or why the player is positioned where he is. Obviously, on a case-by-base basis, you can make inferences, but in the aggregate the data is accurate to a point but not to an end.
So, to thethe's comment, an 'eye test' (or more comprehensive analysis) would in fact be a useful cross-reference because it extracts the sterility of a Statcast measurement and infuses it with actual, useful, information from the game itself. Did the player spend a significant portion of the previous inning on base? Was it late and close? Was it sundown at Turner Field?
I love Statcast. It's great, but it's almost more of a party trick at this point in its inception. When tools are employed (such as HRMs, for example) to better capture the conditions the player is actually facing on the field then it becomes more of a complete science. Right now, it's not.
I certainly don't trust an eye test to accurately account for things like weather and how tired a player is. I'm not sure how that information is all that useful; when taking aggregate data, that stuff will all average out.
Player positioning is an area where I think there can still be some work done, but I still don't trust an eye test to more accurately measure defensive capability than Statcast data. And I don't see how it's a party trick; it gives clear, concrete, accurate data. Again, it's up to teams as to how they want to interpret it and utilize it, but it measures what it measures extremely well.
Again, the data measures what it measures. It doesn't measure a player's motivations or the context of the situation because it wasn't designed to do that. If you think it would be enjoyable to watch games and track all these hits, then by all means, do it. But it would still be infinitely easier to just take the Statcast data and alter it accordingly.
It may not be a perfect tool to objectively determine the exact defensive value of a player, but it does what it does very well, and certainly better than a fan at home can track that data.
It makes no sense to attempt to do what Statcast already does because you don't feel 100% confident in it. This data is available to us, might as well just use it.
That's why I proposed the community do it together. Maybe I'm a nerd but I thought it would be fun to compare our findings with what is on statcast
GovClintonTyree
<B>Vencer a Los Doyers</B>
Reality is that I may be wrong, I'm skeptical, but not beyond convincing. I'm thinking about it.
The other reality is that I personally insulted him because he insults others. Constantly. It's uncalled for. He's very prescient, that's not my beef.
As Hawk pointed out later in the thread, there are things not accounted for in those numbers. Other similar have pointedly out that the interpretation of the data, how it's used, is important. For instance, folks were hypothetically suggesting each hit that falls would be worth a run. Well, where is that written and derived? Sounds like a shot in the dark to me.
I'm a pre-advanced stats guy, a little older than many of you, so my lens is a little different from yours. For years defense was kind of a "good, average, bad" toggle. I'm certainly willing to concede that the search for the science is worth pursuing. And I've learned a lot from you guys. Just skeptical, is all.
That would lead to a lot of sadness when balls aren't hit to Ender
Don't get so hung up on how an 'eye test' is measured - just consider it a broad evaluation of factors that are (currently) metrically unquantifiable. I get that you cringe at the concept of a subjective 'eyeballing', but that's not what I'm getting at.
And, sure, it gives accurate data within a specific parameter. I implied as much. But you claimed that it would 'measure everything you would be looking for when watching yourself [...] it just does it much more accurately' ... and I don't find that to be at all true.
nsacpi
Expects Yuge Games
i think the Statcast data can be disaggregated by situation such as score. But this is also the case with all sorts of data. There are situations where you would evaluate a stolen base attempt very differently for example. I have no objections to someone spending their time watching all the stolen base attempts so they can take into account score, field conditions, etc. Just doesn't strike me as a productive use of time.
You don't think the state of the game is available to these algorithms?
Given enough time to write the tool, I am 100% confident an analyst could write a tool that calculates the optimal play for every player to make in every single situation based on game state tables and average run outcomes tables, and then grade the actions of defenders and base runners accordingly.
For example, air-mailing the cutoff guy and throwing the ball home with a near zero chance of getting the runner at home while allowing the batter to advance from first to second is a bad play unless that runner is the winning run. It is relatively trivial to calculate how many runs that misplay cost the team, and ding the player accordingly.
There is no interpretation needed. If action A produced an expected run value 0.1 runs above action B, then action A was the correct action to take. If teams make 10 such correct decisions over the course of a game, they will score/prevent, on average, 1 more run.
I realize this data-driven view of baseball is jarring for those who have separated sports from brains their whole lives, but this is where it's heading. You can either take advantage of the vast amount of info available and educate yourself, or you can dig your heels in and be willingly ignorant as the rest of the baseball world passes you by.
zitothebrave
Connoisseur of Minors
As someone who's a CPA you should know enough about stats to realize that's insignificant. If we look at the cubs last year, their average run differential per game was 1.6. It was the biggest in baseball by a landslide. Most teams in the playoffs were around 0.5. So a handful of PAs in a blowout game won't skew the overall data. That's why advanced stats work the best in baseball. You have such a massive quantity of data. On an individual player basis there are a few issues, but when you start looking at larger collections, then you get a pretty danged accurate picture.