Aren't we talking about a .305 lifetime hitter? One that did not hit the homers that May's and Mantle did? I think .335 is approoriate, maybe generous. Take that year out and look at his other seasons a real good hitter. Not a .390 hitter.

This is usually called a player’s true talent level, which is basically what all projection systems are to estimate. The cool thing about WIS is that we could use both prior and following seasons to do this, since we’re dealing primarily with historical seasons.

The most basic decent projection system is called Marcels and that uses three key components:

• player’s previous stats from the 3 prior seasons, weighted 5/4/3 (so for a 2020 projection you’d weight 2019 stats as 5/12ths, 2018 as 4/12ths, and 2017 as 3/12ths)
• regression to the mean
• age adjustment

http://tangotiger.net/marcel/

For WIS, we could add in the 3 following seasons also as 5/4/3 weighting, for example (so 6 seasons total would be used).

You can continue to add complexity to deal with things like luck, park factors, era, etc. but this would be the gist of what you’d want to do.

An interesting note, one that may not really meani anything but one I simply found interesting after reading all of this, is that Brett’s 1980 season average performance on this site is .364.

A point that everybody here neglects to mention is that "1980 George Brett", for purposes of WIS, is not "George Brett". 1980, regardless of the factors at play, is an isolated outcome that is not influenced by the rest of his career. The WIS player (for a thought exercise, consider him nameless, so your preconceived notions about George Brett are irrelevant) should be modeled as a .390 hitter, in which case a .454 season is very reasonable within a normal distribution of outcomes.

Also, the simulation already adjusts for seasonal and park effects, which is a level of granularity we need to accept. We can calculate BABIP but we don't analyze all of the variables that the figure is a derivative of. We don't have the season's at-bat by at-bat data regarding opposing pitchers, situations, weather, etc, let alone exit velo or launch angle or defensive positioning. We have to make some assumptions in any simulation, so this is what we are left with. He hit .390, which is the most accurate representation of his "ability" in that single season than any other statistical metric we can come up with.

There are Career Leagues where you can draft players with stats adjuster and normalized for their entire careers, in which case you can have a "George Brett" that probably performs along with the real-life player's "true ability". That sim player won't hit .454, and maybe he tops out around .390, but we should look at "1980 George Brett" as its own entity, regardless of whether or not that season is an outlier caused by extraneous effects other than ability.

This has been something I’ve always wondered why they haven’t fixed/changed contrarian. Particularly with pitchers who play in places like Coors or Wrigley or any number of hitter/HR friendly parks who get double penalties when they get out back into those stadiums. Also makes you wonder how good 94/95 Maddux would be if you normalized him for AFC.

Posted by contrarian23 on 4/29/2016 10:28:00 AM (view original):
Here's the same table, with a column that shows the variance between his hypothesized actual ability, and his actual batting average:

Season	Brett's Age	Ability	Real Average	Variance
1974	21	.280	.282	.002
1975	22	.290	.308	.018
1976	23	.300	.333	.033
1977	24	.310	.312	.002
1978	25	.320	.294	-.026
1979	26	.330	.329	-.001
1980	27	.335	.390	.055
1981	28	.330	.314	-.016
1982	29	.325	.301	-.024
1983	30	.320	.310	-.010
1984	31	.315	.284	-.031
1985	32	.310	.335	.025
1986	33	.305	.290	-.015
1987	34	.300	.290	-.010
1988	35	.295	.306	.011
1989	36	.290	.282	-.008
1990	37	.285	.329	.044
1991	38	.280	.255	-.025
1992	39	.275	.285	.010
1993	40	.270	.266	-.004
Career		.303	.305	.001

As expected, over a 20 year career he sometimes hit better than expected, sometimes less so, and on balance those things more or less cancelled out over the long haul. But he had 4 seasons where he had very strong positive variance, outhitting his expectation by 25 points or more. Unsurprisingly, those were his 3 batting titles and his great 1985. He also had four seasons where he had very strong negative variance, hitting under his expectation by 24 points or more. These included 1978, 1982, 1984, 1991. In each case, these variances are within the bounds of what can be generated in a simple Excel model.

I’m a couple years late to this discussion, but I found this chart very intriguing. When creating his “true ability” in the first column, did you use some kind of formula/structure or just kinda play with some numbers based with common themes in player progression with age?

Yeah park effects definitely need to be stripped out in this context

Interesting point re: park effects Contrarian. I mistakenly thought that the normalized "#" stats were not only adjusted for the season, but for the home park as well. That was a leap of faith I should not have taken! I stand corrected and as always, value your insights and opinion. The double-boosted production in your Bichette example is pretty frustrating, and I agree is something that should be corrected.

I think we'll respectfully "agree to disagree" about the philosophical aspect of the original question; I don't necessarily expect the sim to align with real-life outcomes (yes, I see how it's counter-intuitive to NOT expect that from a simulation), so I don't have as much of an issue with 1980 George Brett having a range of outcomes centered on .390. I completely get where you're coming from, but I enjoy some of the more "unrealistic" aspects of the sim that arise from expanded variance.

This has got me thinking now...would certain players in the sim be prone to more variance across simulations? Off the top of my head, would a .300 hitter with a high k rate (higher BABIP) show more variance than a .300 hitter with a low k rate, at the very least driven by pitching and park differentials in one league vs another? I can think of multiple reasons why this may NOT be true, but now the wheels are turning and I have something to chew on all afternoon...

You guys make my head hurt..................

To try and relieve it a bit I think I'll go "read" the Amazing Mets thread........... especially pages 145 and 146.

Posted by chargingryno on 4/26/2020 11:13:00 PM (view original):
An interesting note, one that may not really meani anything but one I simply found interesting after reading all of this, is that Brett’s 1980 season average performance on this site is .364.

I don't think chargingryno's comment here should be lost in the shuffle.

It does not, of course, tell us what the RL George Brett's talent level is. But using the interesting distinction between George Brett in 1980 (the real person) and 1980 George Brett (the single entity as far as the WIS database is concerned) it is fascinating that the .390 input does not result in a .390 average of his 65 uses the mean is .364 - BUT, and this is the most interesting part: the highest performance was .445 - there is your outlyer performance that contrarian23 rightly has railed against treating as a reasonable outcome BUT BUT BUT....wait for it....

The lowest performance was .363 !! How is that even possible ? THAT is a weirder statistical outcome than the .390 average or even the .454 average !

So, this player has been utilized 65 times with an average result of .364 but 63 of those times he hit exactly or just about exactly .364, once .363 and once .445 to end up with .364 as a mean.

Have I missed something? This WIS database players IS a .364 hitter. Period. Whether the real Brett was or not remains an open question.

The lowest performance was .363 !! How is that even possible ? THAT is a weirder statistical outcome than the .390 average or even the .454 average !

italyprof, best and worst seasons in Performance History are based on Runs Created. So while that "worst" 1980 Brett had a decent slash line of .364/.416/.531, he played in only 97 games and thus created only 85 runs.

In prog leagues and multi-season themes you can look at a player's past seasons and order the results by, say, most home runs or stolen bases or best batting average. Unfortunately you can't do that in Performance History. I wish you could.

Ah, live and learn. Thanks crazystengel. I will leave my misguided and misinformed comment up because your answer which it leads to will provide others with this information which I may not be the only one to not have known. And it is great to know that Runs Created is the criteria used here at WIS. I never knew that.

Oddly enough, given that crazystengel informed in the posts just above this that the Performance History is based on Runs Created, I had re-logged on just now and come back to this thread because I wanted to post something based on Runs Created !

Well, yes and no. So I am pretty sure that many here are familiar with Bill James' system of Win Shares, based on Runs Created with analogous tools for gauging pitching and fielding contributions. In any case, I won't describe the system here, which is complicated - he takes whole books to explain it - but it creates a levle playing field for judging and comparing players' contributions and performances across different baseball eras, with winning and losing teams, correcting for ballpark effects etc. So a player whose performance was the same level of contribution to their team in 1908 with a losing team as the contribution of a player in 2000 with a winning team will be shown to result in the same number of Win Shares as a measure of player performance.

I thought, in other words, that we might approach this same problem using a different methodology (in academia we get extra points for using the word "methodology" once a day. Points. They still don't pay us anything).

So, back to Brett - According to Bill James' book "Win Shares" his 1980 season meant that that year Brett contributed 36 Win Shares to his team with his performance. That means that Brett's performance was responsible for the equivalent of 12 entire victories for the Royals that year ! A hell of a year.

BUT....1980 was NOT Brett's best season ! His best season was in 1985 when he contributed 37 Win Shares, one-third of a victory more than in 1980. Brett's Win Shares over that whole time period: 1975-1989 are as follows:

25, 33, 29, 23, 33, 36, 14, 27, 24, 14, 37, 19, 15, 26, 17.

What we see here I think are two things that are clear: 1) Brett DID have several years that were in the same category of achievement as his 1980 performance - 1976 (33), 1977 (29), 1979 (33), 1980 (36), 1985 (37) and two or three other seasons that were not as good but still very good, in the mid to high twenties in Win Shares. BUT 2) He is an especially inconsistent player. This may be why contrarian23 chose Brett's 1980 performance to discuss these larger philosophical questions - because 1980 is not how he performed most of the time.

By the way, Mike Schmidt had 37 Win Shares in 1980. Rickey Henderson had 34 that same year.

But James' system suggests that Brett's 1980 performance, if we focus on the underlying quality of what he did and not on the epiphenomenal stats themselves, was well within the same league as many of his other seasonal performances as I note in point 1) above. His inconsistency is most apparent when we see that his Win Shares for the 1984 and 1986 seasons are far below the standards of his best season which was not when he hit .390 but instead was in 1985 when he hit .335 with 30 homers. That year, 1985, Rickey Henderson had 38 Win Shares, Tim Raines 36. So we still don't have anything beyond the believable in terms of underlying performance. Just a great year comparable to other great players' great seasons.

So, what ABOUT THAT .406 season of Ted Williams?

Here are Ted Williams' Win Shares for the decade 1940-1949 with the war years when he was not in baseball showing X:

30, 42, 46, X, X, X, 49, 44, 39, 40 - Williams then had two other seasons over 30 Win Shares (the equivalent of winning 10 entire games single-handedly for your team): 1951 - 34 Win Shares and 1957, 38.

What I see here are two things: 1) Ted Williams WAS a consistent player. That 1941 season was no anomaly, and the .406 batting average is just an epiphenemon, a surface expression of the undelying ability of the player who consistently performed at that level year after year. And 2) 1941 WAS NOT TED WILLIAMS' BEST SEASON ! His best season was 1946.

Let's look at two players we might think of as having some simliarities to Brett, though both had less power than he had, Boggs and Gwynn.

Tony Gwynn had two 30+ seasons in the 1980s: 1984 (.351) and 1989 (.336). The year he hit .370 - 1987, James' system credits him with 29 Win Shares, fewer than Raines (34), Boggs (32), Trammel (35), Jack Clark (33), Ozzie Smith (33) - of course Win Shares takes fielding contributions into account, so it is not a comment on the offensive contribution of Ozzie relative to Tony Gwynn here - and Daryl Strawberry (30).

In the 1990s, Gwynn had only one 30 plus season: 1997 when he hit .372. His .394 batting average in 1994 is of course limited in its contribution by the shortened strike season, he got only 17 Win Shares for his performance for that partial season. As far as I can tell only Jeff Bagwell got 30 Win Shares, 30 exactly, for 1994. 28.06 of which were for hitting.

Boggs: Wade Boggs first season is 1982. His line for the 1982-1989 period: 15, 34, 28, 31, 37,32,31,29, so remarkably consistent, compared to either Brett or Gwynn. For the 1990s Boggs gets 24 and 25 in 1990 and 1991, then hits 20 only once more in 1993. So we see the talent and performance decline, though years of 18 in 1994, 18 in 1995, 15 in 1992 and 1996 are by no means bad seasons.

Finally, to compare with Ted Williams, let's look at Stan Musial - I don't think there is anyone else remotely comparable that we could use:

Musial's first full season is 1942. He misses 1945 (X). From 1942-49 his line is: 28, 39, 38, X,44, 25, 46, 40. Pretty consistent and way above Brett's performances in his career, comparable only to Williams (well and Babe Ruth, Barry Bonds etc. but those are different kinds of players, Musial and Williams are very similar in a lot of things).

In the 1950s: 1950-59: 32, 39, 37, 33, 30, 29, 26, 30, 21, 8. So up through 1957 Musial is consistently at the quality of a great season of 30 or more. Any great season of his has to be taken as well within the reasonable, and his amazing 1948 year (46 Win Shares) came in the context of having had 5 seasons of playing out of 6 where he had had 38-44 Win Shares, with one war year missing.