Please explain WAR to me

[topper09er]

But isn't WAR relative to that season?

No its not, its ALWAYS relative to the numbers of runs you produced/saved and then how many runs a win was worth that season. For example the old 1880s guys who threw 500+ innings have huge WAR totals, but you need to understand the context that they allowed them to get that high.

 

Silver King accumulated 16.5 WAR in 584.2 IP in the American Association in 1882. This breaks down to 0.0282 WAR/IP. Pedro in 2000 had 10.1 WAR in 217 IP for 0.0465 WAR/IP, or 1.65 times better

[rluzinski]

Managers don't allow guys to throw more innings like they did before the early 80's. Wouldn't that artificially deflate WAR for modern pitchers?

It doesn't artificially inflate anything. WAR is simply a framework:

(WINS - WINS_replacement level) x playing time

Like I said before, it's more of a value stat than a "who is more skilled" stat. If your point is that Pedro was way more dominant in his prime than Niekro, no one will argue with you.

So why do

people continue to use a player's WAR to assess who is having a better season

or who is the better player? That happens all the time.

If we are comparing two players roughly the same PA/IP, it is like an unregressed projection that ignores anything but the current year's projection, so I am not that fond of it being used that way either. It is not a a real projection nor is it really value, as it is context neutral (over a career, context evens out almost completely).

Calibri;mso-fareast-theme-font:minor-latin;mso-bidi-font-family:"Times New Roman";

mso-bidi-theme-font:minor-bidi;color:black;mso-ansi-language:EN-US;mso-fareast-language:

EN-US;mso-bidi-language:AR-SA">Three nobody really seems to understand

precisely how it derives it's outcome. We seem to know what goes in but not

sure what measure of each and such. Hell Rluz once said he wasn't exactly sure

how it was derived and he's one of the most astute stats guy here.

I don't recall when I made that comment but I linked to the fangraph's very in depth explanation of how it is derived. One of the beauties of WAR is that it ISN'T a black box.

Calibri;mso-fareast-theme-font:minor-latin;mso-bidi-font-family:"Times New Roman";

mso-bidi-theme-font:minor-bidi;color:black;mso-ansi-language:EN-US;mso-fareast-language:

EN-US;mso-bidi-language:AR-SA">My final issue is more based on my skepticism of

the very concept that one stat can be all encompassing than on WAR itself. I

understand the desire for an easy to read, easy to understand number that

encompasses everything we need to know about a player or when comparing two

players.

Even the greatest proponents of WAR would not suggest that it renders all other stats irrelevant. Nor should they pretend that there isn't room to ammend/improve the current incarnations of it. As previously mentioned, Fangraphs and Baseball Reference uses two different version. I think the concept of wins (or runs) above replacement times playing time is a valuable framework to explore, however.

[Baldkin]

Johnny Damon is going to get 3,000 hits more than likely, does that make him a better player than Ted Williams?

 

You'd be crazy to think that, the same goes for WAR. It can tell you who is having a better season, who might be breaking out, and how a guy might perform in the future. It's also very useful for comparing guys across era's and looking at player peaks. `

[The Truth]

Thanks for the explanations. So obviously WAR isn't a very effective means of comparing the All-Time greats.

[topper09er]

Thanks for the explanations. So obviously WAR isn't a very effective means of comparing the All-Time greats.

Actually it is because it adjusts for the scoring era, numbers of runs required for a win. Just remember if you are not comparing guys with a similar number of PAs then you should look at WAR/PA to get a better idea of who was more valuable over their career.

 

It is also good to identify huge seasons, for example without using defense here are the seasons of 10 WAR from offense (hitting + base running + position) alone in the last 50 years:

Bonds 04 (12.1)

Bonds 02 (11.9)

Bonds 01 (13.0)

Giambi 01 (10.0)

Yount 82 (10.7)

Carew 77 (10.0)

Morgan 75 (10.3)

Allen 72 (10.0)

Aaron 63 (10.1)

Mantle 61 (10.9)

 

If you had to pick the most valuable offensive years in the last 50 you might not pick those. Remember, this is most VALUABLE, not best.

[danzig6767]

Niekro pitched for a lot of bad teams, but he was not a very good pitcher. The only reason that he was able to accumulate the stats that he did was that he started 40 some games a year in his prime, never got hurt, and pitched until he was 50. The whole 'best knuckleballer ever' thing doesn't play for me either. Wilbur Wood was better in the 70's, but he got drilled by a line drive in the leg, basically ending his career.

Martinez on the other hand was basically a late 60's Bob Gibson clone smack in the middle of the roid era.

This is off topic from the WAR discussion and i don't know enough War to comment on it, but Pedro in his prime years with Boston easily was the best pitcher i've ever seen pitch and i'd have to think those years were among the greatest ever when you factor in the roids factor. He was ridiculously dominant, those 1999 and 2000 seasons were incredible. What a joy it much have been to be a Boston fan during those years, every Pedro start being near must see viewing.

[Patrick425]

I don't fully understand how WAR is calculated, even after reading through this thread (my head is about to explode!). However, I understand it's a cumulative stat (like hits, HRs, RBIs, etc) and so you really can't compare different players career WAR because some of it is a function of how long the player played. Someone brought up the Pete Rose/Ty Cobb example to illustrate this.

 

However, you could take Pete Rose's and Ty Cobb's Hits per AB (basically batting average) to get a more accurate comparison of talents. Couldn't you do the same with WAR if it's truly cumulative? Couldn't you just divide it by number or years played or games, or innings played?

[CheezWizHed]

But isn't WAR relative to that season?

No its not, its ALWAYS relative to the numbers of runs you produced/saved and then how many runs a win was worth that season. For example the old 1880s guys who threw 500+ innings have huge WAR totals, but you need to understand the context that they allowed them to get that high.

I guess I was figuring that the replacement level would fluctuate from year to year based on playing conditions (changes in baseballs, bats, general player conditioning), thus WAR attempts to show how far one particular player stands out from his peers (assuming all have equal chances) in that particular year.

 

For example, if you grabbed Babe Ruth with a time machine and dropped him in the Brewer's lineup, he probably wouldn't be very good (at least not like he was), since general training methods have improved the overall level of play and Ruth would be "conditioned" by lower standard methods (mainly drinking and smoking). But he still was a great player for that era.

 

But that was a good point about the number of IP or number of PA to accumulate that WAR. I hadn't thought about that.

[topper09er]

I don't fully understand how WAR is calculated

Conceptually it is simple, you convert everything a player a does into runs, add them up and then determine how many runs it takes for a win in that season/ballpark.

 

The way this is done is much more complicated. In batting, the run value of each event has been determined. These are called the linear weights which are attached to each event (single, double, sac bunt, strikeout, etc) to weigh them relatively. They are determined based on empirical data and differ depending on the run scoring environment. There are 24 base-out states, 8 possible ways the bases can be occupied and 3 different amounts of outs. Every base-out state has an expected number of runs that will score from that point until the end of the inning based on historical data that actually happened in games. You can see it here.

 

Whatever the result of the at bat, you just subtract runs expected before the AB (so if you are batting with the bases loaded, 0 out you already expect 2.417 runs to score that inning) from what is expected after (so if you then get a single, score 2 runs and leave R2,R1, 0 out, you get 1.573 - 2.417 + 2 = 1.156). If instead you strike out you get 1.65-2.417 = -0.767. With no one on and 2 outs a single is worth 0.251-0.117 = 0.134 runs. Then all the base-out states are weighed based on how often they occur and you wind up with the runs added/subtracted above/below average for each hitting event, seen here. (Based on 1999-2002 data, a little higher scoring than we have right now). Note a HR is about 3 times more valuable than a single, not 4 times as implied by the slugging percentage.

 

In summary:

BB = 0.33, 1B = 0.474, 2B = 0.764, 3B = 1.063, HR = 1.409, K = -0.31, other out = -0.29, SB = 0.195, CS = -0.456, etc...

 

If you add up every PA for the season and the number is positive the player has produced that many runs above average on the year.

[topper09er]

The other thing that seems to be misunderstood a lot is the concept of a replacement player. In reality, WAR is really wins above a certain minimum winning percentage. I don't remember exactly but early in Moneyball they talk about if you have a roster of entirely players being paid league minimum, basically a AAA team, you would still win a certain number of games based on the long season and randomness of baseball. In football, this level would be 0, as demonstrated by the Lions and Buccaneers of the late 70s, but in baseball it would be around .333 or so. The worst winning percentage of all-time in the modern game was the 1962 Mets who went 40-120 (.250). At BBref they use 52-110 (.320).