tangotiger

While not an objective, it's an eventuality. All the pieces are there, and it's just a matter of merging them properly. I'd like to say this will be done in the off-season. But we have so many other things to do as well.

Tomorrow morning as well.

Tomorrow. Literally, tomorrow morning.

That is a good question. We have to try to understand WHY we are doing park factors. And it's to (try) to put players on the same baseline. If we look at 2023 (for illustrative purposes), we can see that Yankee Stadium helps HR hitters and hurts players who get non-HR hits: https://baseballsavant.mlb.com/leaderboard/statcast-park-factors?type=year&year=2023&batSide=&stat=index_wOBA&condition=All&rolling=&sort=12&sortDir=desc This is also true at Dodger Stadium. Indeed, if you look at Dodger Stadium year by year, you will see that is consistent: https://baseballsavant.mlb.com/leaderboard/statcast-venue?venueId=22 Now, let's say you have two batters who are equivalent away from Dodger Stadium, one who is a power hitter and one who is a bigtime singles hitter. But overall, they are equivalent. At Dodger Stadium, the power hitter will likely be more helped than the singles hitter. Or, for illustrative purposes, let's say this is true. So the question is thusly: do we want to apply a SINGLE park factor, the same park factor, for EVERY hitter? Or do we want to apply an individualized park factor? Well, it depends what you are trying to answer. In terms of helping the team win, then you want a single park factor. One batter was able to leverage Dodger Stadium and the other one was not. It's not their fault that they had to play 81 games there. But, that's indeed what happened. And to explain what happened, the one who was able to leverage Dodger Stadium will get the wins. But, if you try to trade for a player, or try to place all the players in a neutral park (or have them play 5 games at each of the 30 parks), then you'd want an individualized park adjustment. This is actually far more important for pitchers than batters in terms of asking the question, and yet 99% of the time, the question is asked about batters.

It really depends if your purpose is to model the PLAY or the PLAYER. Most of my time is spent on trying to understand players, and so, we need to tease out those things that cause a large amount of random variation. This is similar to removing fieldable batted balls and only using FIP. Because FIP describes the player much better when you exclude fieldable balls. Of course, if you want to explain the PLAY, they you want to include fieldable balls. Which do you prefer? And you can't say "both", not if you want ONE metric. If you want two, then fine. But that's two separate questions and so two distinct metrics. UPDATE: Here's a good post on the subject: http://tangotiger.com/index.php/site/article/spray-angle-overfits-xwoba

That is kind of you to say thank you! I used to spend alot of time at https://battersbox.ca/ (I still have admin access there.) I'm from Montreal, for those that don't know. I came here because every now and then I google my name and Google directed me to a 425 (!) page thread. So, you can thank or blame Sergey and Larry.

Similar to the positional adjustments, feel free to ask your questions. Try to frame your question in the form of a question: don't include your opinion in the question, or preamble it with your opinion. I'm not here to "defend" anything, just to explain it.

I'll create a separate thread for batting.

Thank you kindly for your support!

All good questions, and I haven't looked at his plays specifically. 1B is actually more difficult to model because of the option to run or throw and the dependence on the pitcher covering. That said, Christian Walker always comes out on top every season, so it must be doing something right.

I last played in the early 90s. Yes, I did fine. I don't sell articles, I work at MLB.

You mean "clutch fielding" or something? If you mean clutch hitting, then let's table that for a different thread.

See my above note on wall-balls.

It definitely includes positioning. In the first play, Varsho went the wrong way initially. So, this is a GOOD example of the system working. The second play is a wall-play and is more problematic. Right now, we aren't looking for how high a ball is off the wall, just if it's "playable". That has more to do with tracking limitations.

Correct, outfield only. Infield has more variables to consider, so it's not as easy to show in this manner.

If you follow me on my blog, you will see I am currently delving into the bat swing.

Go to my last post about Varsho. We are simply evaluating each play, one at a time. Each play gets a plus/minus. And we add those up. It's really as simple as it sounds.

We make available EVERY SINGLE PLAY, with video, right here: https://baseballsavant.mlb.com/savant-player/daulton-varsho-662139?stats=statcast-r-fielding-mlb Go about half-way down, hover over any of the dots to see the Catch Prob. Then click the dot to see the video. We are extremely transparent here. So, feel free to click those, and then link back here any play you think is questionable. Yes, "lost in sun" is a real thing, and I know Varsho had at least one. Unfortunately, we are not tagging "lost in sun" or "hit the catwalk" in our play events. If he has two of them, then he gets minus 2 OAA, if they were otherwise routine plays.

It is mostly the latter part the "simpler" explanation. The model is very intuitive, and simply asks the question you are asking: how much time does the fielder need to get to the ball, and how much time does the fielder need to throw (or run) to the target base. And at the moment that the fielder/ball intersect, where is the batter/runner, and how much time will that player need to get to the target base. And so, we simply ask: who will win the race? It's EXACTLY the way you evaluate fielders while watching the play develop. And so, the model is matching reality. There are some nuances to 1B and 2B and SS and 3B that requires a bit extra work to make sure that we're all aligned to what "average" is. The twist and turns of the body, the running in and throwing across the body etc. But that's a secondary effect to the simpler intuitive time-distance model.

When it comes to Statcast data, this is not an issue. We are measuring the players using actual data, not inferred data. Do you have an example that you are thinking about?

It's easier to separate infield from outfield. And it's also better if we differentiate between UZR/DRS and OAA. So, speaking specifically about outfield: UZR/DRS compare the performance of players to the average for that position OAA compares the performance to the average outfielder, regardless of position As a result, OAA will DIRECTLY allow you to compare a LF to a CF to a RF without ANY adjustments. The average CF for example is close to +5 runs above the average outfielder. The average LF is -5 runs, and the average RF is 0 runs. Naturally, in UZR/DRS, the average of each is 0. And so, you CANNOT DIRECTLY compare players at each position. You need a "positional adjustment" to fix this issue. The positional adjustments that I offered 15 years ago was that the average CF is about +6 runs above the average outfielder, while the LF and RF were -3 runs each. As you can see, we have more refined data in the Statcast age, more accurate data. At the same time, those adjustments were based on data from 15 to 25 years ago. It's just as possible the talent level of players has indeed changed. That's the main problem with the positional adjustments: they require constant monitoring. We don't have that problem with OAA, because everyone is always compared to the same single standard for that season.

OAA is mine. OAA uses actual ball location and actual player location.

It is indeed me. Here's proof: http://tangotiger.net/files/bluejaysmessageboard.txt

I saw discussion regarding positional adjustments in a thread. I'll just start this one so we can focus the discussion. Feel free to ask your questions, and I'll see how I can help...