|
||||
|
|
Talk Sports Discuss everything that is sports-related, like MLB, NFL, NHL, NBA, MLS, NASCAR, NCAA sports and teams, trades, coaches, bad calls etc. |
|
Thread Tools |
03-19-2020, 04:13 PM | #1 |
Banned
Join Date: Aug 2019
Location: Ban land in 3...2...
Posts: 2,943
|
WAR update
Baseball-Reference recently updated it's WAR calculations. Some changes to methodology, and some changes due to new data.
You can read about those changes here https://twitter.com/baseball_ref/sta...939884548?s=19 One of the common criticisms of WAR is that it changes. But change is good. Not bad. For example, one of the newest changes to WAR was the incorporation of missing play-by-play data. Specifically regarding things like caught stealing and double plays. That info was previously estimated. Now the actual numbers are known. This means that some players' WAR changed by as much as 7 for their career. Some people will look at that and respond with some version of "LOL. wAr SuCkS!". Obviously, that's not a very compelling argument. Especially since those people almost certainly had opinions on the players impacted by the new information and other changes. Opinions based, like WAR, on incomplete information. The difference is, WAR changes with new information. As, should be obvious, it should. Anyone who questions WAR because it changes with new information should, themselves, be questioned. Then, there are those who will say something like, "If we couldn't trust WAR then, why can we trust it now?". I'll cut those people some slack. That's actually a decent question. It also has a succinct answer: trust WAR because it works. Here's undeniable proof (See attachment) https://www.beyondtheboxscore.com/pl...ird-order-wins WAR doesn't know anything about how many wins a team, or a pitcher, had. It doesn't know about runs, or RBI, or saves, or batting average with RISP or anything like that. It only really knows about outs and what happened when it wasn't an out. And it does an excellent job modeing how many wins a team should have with certain amounts of outs and non-ours. Of course, it doesn't really even know about teams. It knows more about players. But it mostly knows about plays. Then adds up the plays for each player. Add up the players for each team. And there you go. An accurate model for team wins. It's not perfect. Because of course it's not perfect. No one wants it to be perfect. If it was perfect something would be wrong. K, Single, single, walk, Homer, K, K. Gives you an onfield team result much different than: Homer, K, single, single, K, walk, K. But for an individual batter those two sequences are equal. Either one could be better depending on was going on in the game at that time. But WAR doesn't know that. It, rightly, wasn't designed to know that. But what it was designed to do, it does very well. And now it does it a little bit better. Last edited by CBeisbol; 04-10-2020 at 04:23 AM. Reason: Edited out an unfortunate series of strike outs |
12-02-2020, 07:52 PM | #2 | ||
Banned
Join Date: Aug 2019
Location: Ban land in 3...2...
Posts: 2,943
|
Tom Tango wrote about the WAR framework
Quote:
This is probably in response to Bill James' comments from yesterday Quote:
Of course, if we check WAR we see that it is accurate (not 100%, but accurate) meaning that the assumptions and measurements are (no, not perfect) but good https://www.beyondtheboxscore.com/20...ird-order-wins |
||
12-02-2020, 09:42 PM | #3 | |
Hall Of Famer
Join Date: Apr 2009
Location: United States
Posts: 10,436
|
Quote:
You may recall my saying that I have recently been using WAR in my OOTPB game, and my statement that as long as one's WAR source is reliable and using its WAR calculation consistently, it is a useful tool for measuring and comparing player performances. Current and, hopefully, over time. My mistake in the past, and those of others perhaps, was looking for absolute WAR. It's a statistic that gains its relativity from comparison, not from absolute numbers. It is to be assumed that Baseball-Reference has, or will soon, go through its entire database and update its WAR ratings for every player. Once that is done, it will have relativity again. It was useful before and, if what you presented is true, will be even more useful in the future. Now, the fact is that if I looked up an historical player's WAR in Baseball-Reference and that in some other source (like OOTPB), I am likely to find significant variance. This is immaterial. One should choose a reliable source, or accept the variations between more than one reliable source, and look at WAR only within each source's framework.
__________________
- Bru |
|
12-03-2020, 11:51 PM | #4 | |
Banned
Join Date: Aug 2019
Location: Ban land in 3...2...
Posts: 2,943
|
Quote:
This is not my recommendation Understanding the differences between different forms of WAR and why they rate the same player differently, leads to a greater understanding of the player and their value/performance |
|
12-04-2020, 04:32 AM | #5 | |
Minors (Triple A)
Join Date: Feb 2018
Posts: 251
|
Quote:
If you want to compare players on a single overall value like WAR, it seems like the best you could do is just take the averages for the season across all situations, like how many runs a single adds on average. But maybe there is room for more situational values, because I think a player's value on one team or in one place in the batting order or whatever might be different if they were on a different team or place in the batting order. Maybe a single is more or less valuable for different teams, on average. For example, a weak offensive team might get less value from a single because there's likely less runners on base and less chance of the player who got the single to be batted in. Or a walk is more valuable on a team that hits more HRs. Or a more defensively capable player is less valuable to a team with better pitching that is already less likely than average to give up hits, compared to a team whose pitchers can't get a K. Or a player's batting line might be more valuable batting 1st instead of 4th. Even the field shape could benefit some players more than others, and could significantly affect their production. This kind of information could be useful to teams in finding players better suited to particular spot on their team than how they are currently be used in another team. So WAR is a nice general number for overall comparison, but there's a lot of potential for more specific situational comparison calculations. |
|
12-04-2020, 11:56 AM | #6 | |||
Banned
Join Date: Aug 2019
Location: Ban land in 3...2...
Posts: 2,943
|
Quote:
Quote:
Teams almost certainly do this. WAR specifically avoids doing this. The idea being to measure all players on an equal playing field. Quote:
Last edited by CBeisbol; 12-04-2020 at 04:58 PM. |
|||
12-04-2020, 02:23 PM | #7 |
All Star Starter
Join Date: Mar 2005
Location: Guarding The Line
Posts: 1,201
|
I am still an "old school" stats guy at heart, but over the past decade I have come to respect and use a few more stats. WHIP & OPS were the easiest to accept and slowly but surely WAR. It is now something I always look for both in and out of OOTP.
I remain a bit skeptical on modern fielding measures, at times when I look at dWAR for players who were solid fielders I am puzzled to see negative numbers,
__________________
"...If you want to look ahead to the bottom of the ninth, the Mets will be sending up Buddy Harrelson, Jerry Buchek , and Don Bosch, we'll be right back after this word from Rheingold Beer" The late great Lindsey Nelson |
12-07-2020, 12:26 PM | #8 | ||
Banned
Join Date: Aug 2019
Location: Ban land in 3...2...
Posts: 2,943
|
Quote:
Quote:
With hitting, largely, though it's catching up to defense, it's just looking at results. A single is a single. Whether it's a 100 MPH linev drive or 22 MPH squib off the end of the bar. Whereas with defense, it has to measure not only what happened, but what didn't happen. Sure, the SS made the play right at them, but how about the one 5 feet to their right? Measuring defense is kind of like measuring offense by saying not that was a single, but that was a swing on a 93 MPH pitch with 6" of vehicle run and 12" of horizontal drop in a ball 2" above the bottom of the zone and 6" inside from the center of the zone. What was the expected outcome versus what really happened? Defensive numbers for current players is pretty good and you can trust them. Though,, of course, they are not perfect. Defensive numbers for historical players are...not as good. The people who were keeping stats back then weren't, and didn't have the ability, to keep track of the right things |
||
Bookmarks |
Thread Tools | |
|
|