Kevin Appier's 1993 season is either just as good or almost as good as Ted Williams' 1941 and 1942 seasons and better than his 1947, 1949, and 1957 seasons. The only season of Ted's that is better than Kevin's 1993 season is Ted's 1946 season which is at 10.6 WAR.
Registered User
Neither do I, but we know about how many runs above average he was and a 179 ERA+ for 238 innings is a great year-basically what Verlander did last year. Verlander's 8.6 WAR is believable. The difference by the way is the leverage index wher Verlander is at 0.9 and Appier 1.1. I am sure those are rounded so they look more extreme but Appier is at least at 1.05 and Verlander under 9.5. If we switch those, Appier's WAR drops to 8.6. So its a philisophical issue. Does a starter deserve a higher leverage index because his team probably gave him less run support? I agree, no leverage index for starters. It is almost all team dependent. If a starter gets a lot of blowout wins he gets scored worse even though in a neutralized setting he probably pitches the same and gets a 10% lower LI or more.Originally Posted by Ubiquitous
Last edited by brett; 05-06-2012 at 06:46 AM.
Registered User
Right, Appier is getting a double boost for the runs-to-wins issue and the LI. When/if corrected, his 1993 season will be nowhere nearly as close to Williams best seasons, for example.
1885 1886 1926 1931 1934 1942 1944 1946 1964 1967 1982 2006 2011
1887 1888 1928 1930 1943 1968 1985 1987 2004
1996 2000 2001 2002 2005 2009 2012
The Top 100 Pitchers In MLB History
The Top 100 Position Players In MLB History
Registered User
[QUOTE=csh19792001;2008905]Does any stat out there adjust for the average quality of the opposition a pitcher faced and/or their run support?
...
I wanted to re-examine the Ned Garver 1950 season because it had come up some years ago on another baseball forum. This time I wanted to apply my defense metric to TEAM and LEAGUE, by position for the A.L. 1950 season, just to see how the numbers all shook out at the end of the exercise. I wanted to see if pitching runs, batting runs and defense runs might balance with a team's record at season end.]Ned Garver had a .451 wpct. 129-157 lifetime. Yet, a 112 ERA+. In 1950, he went 13-18 on a team that lost 96 games and LED the league in ERA+!! Garver pitched 65% of his career innings and had the worst run support in baseball history.
Should we adjust for run support, defensive support, and the average strength of the opposing team a pitcher faces??
1. St. Louis Browns Defense, by positon, versus American League, my defense metric and run conversion:
Position.....Browns ..... League ... Runs
C......... .930........... .936... -2.45
1B....... .946........... .971... -5.68
2B....... .956........... .959... -1.04
3B....... .941........... .961... -4.60
SS....... .936........... .964...-10.47
OF....... .968........... .952... +6.05
OF collectively weighted, +/- for aggregated OF rating. Total Brown vs. AL = -18.19 DR
Equivalence:
On a team level, much more than at individual positions, team pitching staff K's vs League can impact position player opportunity to field batted balls in play and turning them into outs If a staff has high K's then the surplus to the avarage [+] can be added to the raw RUNS, to adjust.
The Browns pitching staff recorded FEWER K's than League [448-570] = -122 K's. The metric, responding to each data entry adjustment for K's, values each K at .11 defense runs; so the Browns get a -122 * .11 adjustment = from -18.19 DR to -31.61 DR.
Batting, the Brown produced 684 Runs, -98 Batting Runs below League average.
Pitching, I personally favor WHIP as the core of any pitching evaluation. I realize it lacks the neat distillation of FIP; but it just strikes me that it is as tidy as one can get in distilling the battle between pitch and bat right down to the inning. The Browns pitching staff had a 1.670 WHIP vs. a League average of 1.541. Run conversion from WHIP elements into runs scored was about .37 in 1950; so the Brown discrepancy, applied to innings played [1,365.1] = 176.1 * .37 = 65.2 runs expected surrendered above average.
Now adding the elements, each happening to be deficits in the 1950 Browns case:
Batting -98 Runs
WHIP +65.2 Runs surrendered
Defense = -31.61 Defense Runs
Taking a 10-to-1 Runs-to-Wins view, we have team with a cumulative deficit of 194.76 runs, or =19.48 wins. The Browns played 154 games in 1950; so half [77] minus 19.48 = 57.52 estimated "wins." The Browns record in 1950 was 58-96.
How does this involve Garver?
Well, a look at his "line" in 1950 is unimpressive, except for his 260 IP. Yep, he was a workhorse. However, despite a 5 run scoring climate, he looks like a pitcher always working his way into [and out of] trouble:
260 IP with 264 hits surrendered
260 IP with 108 BB issued
260 IP with 85 K.
Now, for his WHIP. At 1.431 he is [1.670-1.431] = .239 WHIP elements * 260 IP = 62.14 WHIP elements @ .37 run expectation rate = 23 pitching runs better than his staff. Compared to League WHIP [1.541] he his .11 * 260 = 28.6 * .37 = 10.58 runs better than League in run expectancy over his 260 IP.
In this context, we see a pitcher rising somewhat above a miserable situation; but the astounding 146 ERA+, I believe, is a fluke wrought by taking a statistical element out of its intended purpose and arriving at a misleading result.
Garver had significantly better seasons than 1950, with no such distorted "greatness" making them suspect.
Last edited by leewileyfan; 05-06-2012 at 05:03 PM.
Registered User
To put my post [above] into its intended context, if I take the WHIP comps to team and league, and apply them in runs saved as applied to Garver's 260 IP, I get this picture:
1. Garver's 1.451 WHIP better than League WHIP [1.541] distilled to Garver's 260 IP, has Garver 10.58 runs better than League, which reduces to 150.29 ER [League] in 260 innings. The League ERA, full season, all teams, is 4.58. If we credit Garver with 10.58 runs better than League, then 132.28 - 10.58 = 121.66 ER over 260 IP = 121.66/28.89 [9 inning games] = 4.21.
2. If we isolate Garver's WHIP 1.451 vs. team [1.670], we have Garver 23 runs better than team [260 innings], the team ERA [5.20] distilled to 260 IP = 150.29 ER -23 = 127.29/28.89 = 4.41.
Either way, substituting Garver [4.21] for League [4.58] or team [4.41] vs [5.04], we get a different perspective on Garver's performance from that 146 ERA+. I do believe the alternative approach is far more credible from the standpoint of what fans look for in making player comparisons to whatever time-frame they are examining.
Registered User
ERA+ depends on how a pitcher pitches with men on base though. Who looks better in your system, Ryan or Seaver? Seaver pitched relatively better than average with runners on base. Ryan pitched relatively about normal or a little worse with runners on base.
http://www.baseball-fever.com/showth...ver&highlight=
Registered User
While I'm sorry that Sean is having a rough time, it's a great opportunity to learn about WAR and about the way people think of it. In addition to this discussion, there is a good one on Tango's blog with Sean participating. There's a link from the blog page: insidethebook.com/ee
(apologies for not giving the full link--no cut and paste function.)
I realized that to get something right (or plausible, anyway), you have to get it ALL right (opa), because, as ubi said, it all meshes together. I think that's too much to expect from anyone at this point, but it is fascinating to see how the problems are identified and addressed. I am so impressed by his efforts--and results.
Registered User
Is this true? I didn't think he forced all the players' wins on a team to equal actual OR projected wins. Is there a "pool" of wins for a team or not?
You can't have an infinite amount of wins for a team.
Registered User
Brett, it's not really a system, more a personal approach. Comparing Nolan Ryan straight up with Tom Seaver is pretty straightforward, in that they are contemporaries and that they weren't world travelers, being traded like bubble-gum cards or shuffled back and forth between leagues.
Trying NOT to look at W-L, I selected eight seasons for side=by-side comps using WHIP and run conversions as explained above. Both were young and entering their respective peak periods. One was essentially A.L.[Ryan]; the other was entirely N.L. [Seaver] for the seasons selected. The seasons are 1971 through 1978.
The exercise gave me a renewed respect for Tom Seaver.
Season.....Pitcher.....WHIP.....LG WHIP.....IP.....BFP.....Pitcher.....WHIP.....LG WHIP.....IP........BFP
1971........Ryan........1.586.....1.296.....152.0. .. 705.....Seaver.....0.946.....1.296.....286.1.....1 103
1972......................1.137......1.231.....284 .0...1154..................1.150......1.290.....26 2.0....1060
1973......................1.227......1.370.....326 .0...1355...................0.976.....1.333.....29 0.0.....1147
1974......................1.272......1.329.....332 .2...1392...................1.161.....1.357.....23 6.0......956
1975......................1.434......1.362.....198 .0....864....................1.088....1.360......2 80.1....1115
1976......................1.322......1.318.....284 .1...1196...................1.063.....1.320......2 71.0...1079
1977......................1.344.......1.377....299 .0...1272...................1.014....1.368......26 1.0....1031
1978......................1.411......1.351.....234 .2...1008...................1.182.....1.317.....25 9.2.....1075
Applying the same step-by-step process as I did for the Garver post:
1.Subtract League WHIP from pitcher WHIP [i.e. Seaver, 1971 = 0.946 - 1.296 = -0.35].
2. Multiply that quotient by pitcher IP during the season [-0.035*286.1 = -100.135 [this is WHIP "elements" ... BB + H]
3. I took the liberty of using .36 as a reasonable conversion of WHIP "elements" into runs.
4. -100.135 * .36 = -36.05 runs ... Seaver was -36.05 runs below average in 1971.
Doing this for each season for Ryan and Seaver, the final tally over the eight seasons is:
Ryan = -984 runs
Seaver = -201.4 runs
Using the 10 runs = 1 run model, Ryan would be expected to be 9.84 wins above average [.500]. Over the eight seasons, Ryan was 132-121. For 253 decisions, .500 would yield 126.5 wins. In actuality, Ryan came in at 5.5 wins above .500.
Applying the same to Seaver, his 201.4 runs would yield an expectation of +20.1 wins above .500. Over the eight seasons, Seaver was 165-89. A .500 rate for 254 decisions = 127 wins. Seaver came in at 38 wins above that level.
With Ryan-Seaver it's not even close. The numbers show an economy of effort by a pitcher who has mastery over his control. Speaking ONLY for myself, I like WHIP and BFP to look into the core of pitcher evaluation. If you smush the two elements together, you get an insight into offensive prowess against individual pitchers.
For example, if you multiply WHIP by IP and divide that number by BFP, it presents in a form resembling batting average[actually more like On-Base %]; and it gives a pretty good insight into how truly dominant a pitcher really is:
1971: Ryan .342; Seaver .245
1972: Ryan .280; Seaver .284
1973: Ryan .295; Seaver .247
1974: Ryan .304; Seaver .287
1975: Ryan .329; Seaver .273
1976: Ryan .314; Seaver .267
1977: Ryan .316; Seaver .257
1978: Ryan .328; Seaver .285
I looked into ERA+ and see nothing in the basic definition and purpose that would credit it with any special power to differentiate pitcher effectiveness with/without men on base.
Last edited by leewileyfan; 05-07-2012 at 09:22 AM.
Registered User
I don't know anything about the specifics of WAR, but I would assume it must "fit" players wins on a team to the team's actual total. this was one of the basic premises of Bill James' Win Share system, and it's a good one. I have to think that WAR does something similar.
Registered User
Here's a question. Since starters give up about 10% more runs than the league rate (I think they average about a 92 ERA+), should they be rated versus the run setting for the innings they pitched (which will tend to be a little higher). Say innings 1-5 produce a 90 ERA+ for example should a guy be rated against that for his first 6 innings of a game. Then relievers get rated against a 120 ERA+ for the last 3 innings.
WAR isn't tied to wins like how they are for Win Shares but they are still connected.
Replacement level is set at .320 so a replacement level team gets about 52 wins. The average team thus gets about 29 WAR and the whole league gets 875 WAR. Fangraphs sets theirs at .275 I believe so they have 1100 WAR to spread around. Last year MLB players recorded 872 WAR. 3 shy of the 875 amount but two teams didn't play a full schedule . They each didn't play a game. The rest is pretty much a rounding error.
Registered User
I wouldn't tinker with pitcher data based on such supposition, even IF numerically logical. The reason, for me, is that such tinkering can only result in the tail wagging the dog, tacking into realms of liimited expectation and performance.
What I would rather do is distill each pitcher down to some format that features WHIP and runs allowed per BFP. I've noodled around with this and there is a clear bias in favor of effective closers:
Pitcher A: Starter; IP 220; Runs [all] 71; WHIP 1.225; LG WHIP 1.350; BFP 917
Pitcher B: Closer; IP 80; Runs[all] 19; WHIP .960; LG WHIP: 1.350; BFP 320
The starter, using WHIP has 1.225-1.350 = -.0125*200IP = -25 WHIP elements to his credit, above average, which = about 9 runs.
The closer, using WHIP, has .960-1.350 = -.39 * 80IP = -31.2 WHIP elements above average, which = about 11 runs.
If I reduce the two to runs/bfp, I get:
Starter: 71/917 = .0774, which [having done tons of these] is a HoF rate;
Closer: 19/320 = .0594, pretty typical for a very good closer.
That bias seems fixed across the entire span of 1901-2011 when considering starters and definable closer roles, even non-closer relief roles when the reliever is effective.
Maybe it need no tweaking at all, although I could see a per outing factor levied on closers to align their results with other pitchers. A 40% levy on BFP would give the closer a denominator of 192 and with 19 runs allowed, his rating would be at .099. I might look fair; but I can hear closer fans screaming about the injustice.
All it needs is a clear statement that lower ratings are expected from closers, and leave it at that.
Registered User
Registered User
Appier did have a historically great season. He had a 179 ERA+ and 238 innings which is virtually equal to what Verlander did last year, or Peavy a few years back or just about anyone. And he went 18-8 for a team that would project to win just 44% of their games given average pitching. The concept that a run saved is worth more than a run produced, especially when pitchers save multiple runs concentrated into single games is legitimate. The idea of giving a player a leverage advantage because his team never scored enough to not be "close" is a problem, and if we eliminate that he has 8.6 WAR which is about right for what was a really fabulous season.
Registered User
Please bear with me, guys: As I write this, I am trying to work logically through my WHIP fixation while playing Devil's Advocate with myself, and trying to explore the true value of Kevin Appier's pitching in 1993.
1. Appier posted a WHIP of 1.106 vs. a League average WHIP of 1.418, a nice starting point.
2. Breaking it down, 1.106-1.418 = -0.312 [W+H for each IP] * 238.2 IP = = -74.32 base elements of scoring potential.
3. The League WHIP, applied to Appier's IP [238.2] = 1.418*238.2 = 337.76 base elements of scoring potential.
4. IF we simply did the subtraction [337.76-74.32 = 263.44] AND multiplied 263.44 by the actual rate at which the League turned OB opportunities into runs scored [.372], we would project Appier to have surrendered [263.44*.372 = 98 runs]. The League would have yielded [337.76*.372 = 125.65 runs. Not bad. 126 - 98 runs = 28 runs to the good for Kevin Appier.
5. However, if we go to the records for 1993, we see that total runs [earned + unearned] yielded by Appier were actually only 74. Therefore, we have a pitcher at 52 runs better than League.
There is another considerable dynamic in play here that compounds the benefit of WHIP; and that element has a big effect on the RATE at which opportunity [W+H/IP] is converted into runs scored. The League rate is .372. Appier's rate is .281. Appier did have a decent defense behind him but nothing outstanding and very little of full-time at any position.
Appier yielded only 8 HR's in 238.2 IP [.302 per 9 inning game], while League was at .914 HR per 9 inning game. In the equivalent of 26.47 games [Appier's 238.2 IP], League would have yielded over 24 HRs as opposed to Appiers' 8. Factor in man-on-base possibilities and the numbers are reconciled perfectly.
So, if Appier was 52 runs better than League; and if we convert that into wins [5.2], then Appier [18-8] had 26 decisions. Average [.500] would have made him 13-13. Add 5.2 wins to average, and we get 18.2 - 7.8, pretty darn close to Appier's actual record.
Good season. On a plane with TSW? Hardly. I get uncomfortable when numerically precise formulators attempt to evaluate a mixed stew of pitching with anything else, except maybe defense, of which pitching in an integral part.
Example: Say an average position player, batting, "creates" runs at a rate of .125 per PA and has 600 PA in a season. He creates 75-76 runs. If we say another is +52 in run creation in the same scenario, then THAT player creates 127-128 runs in 600 PA; and we can distill that further to .2125, all-star stuff, indeed. However, it doesn't approach a 170 RC season, not even close. [Besides, the position player production is daily; pitcher production is rationed].
Back around 1963 or so, I was right in the midst of settling a brand new family [to add one more within the year] in a new home. Part of that was establishing a lawn. The head groundskeeper where I worked and I had hit it off very well and often talked about "whatever." When it came to horticulture the guy was a whiz. When I asked him about the latest Scott's products [with +2 or +5 in the name], he urged but the stuff you need for each job, one at a time. The "super-stuff" you pay more for pretends to to it all ... BUT none of it is done right."
Sometimes with baseball stats, I believe the long way 'round is worth the effort to produce logical results, even if they lack labels.
Last edited by leewileyfan; 05-08-2012 at 06:36 PM.
Fangraph's WAR, which uses a lower replacement level, puts Appier's 1993 season at 7.4 WAR.
Registered User
If you remove the leverage, and simply count runs "saved" as runs added I get 7.5. Fangraphs may set it lower and at .275 that would be another 0.3 WAR but somehow a lot.most players score LOWER in Fangraphs than BBRef.
Registered User
Fangraphs also ignores any BABIP "skill" that existed, sequencing, holding runners, GIDPs, XHB prevention, wild pitches, pitcher defense. So even with a lower replacement level, if he did well in these areas, that would explain why his brWAR is still higher.
1885 1886 1926 1931 1934 1942 1944 1946 1964 1967 1982 2006 2011
1887 1888 1928 1930 1943 1968 1985 1987 2004
1996 2000 2001 2002 2005 2009 2012
The Top 100 Pitchers In MLB History
The Top 100 Position Players In MLB History
Ted Williams gets 140 WAR with Fangraphs and 119 WAR with BRef.
Ty Cobb gets 164 WAR with FG and 144 with BR.
Willie Mays gets 163 WAR with FG and 151 with BR.
Greg Maddux gets 121 WAR with FG and 104 with BR.
FG has the top positional player in 2011 as Ellsbury at 9.4. BRef puts him at 7.8. The top 30 positional players in 2011 racked up 192 WAR according to FG and 178 according to BR.
Registered User
I get all that, maybe the hard way, by taking Appier's 52 runs better than average and creating two teams, playing 34 games against each other - 34 being the number of Appier starts:
Team A = League 4.71 average runs per game, envisioned as Appier's team, scoring just that many runs per game: [4.71*34 games = 160.14 runs].
Team B = Appier's opponents, held to 52 fewer runs [160.14-52] = 108.14 runs.
Then do the Pythago sum of squares thingee to get a W-L projection for each team. Turns out Appier's team will win .687 of the games, or have a W-L record of 23.35 wins and 10.65 losses, exactly. Getting back to baseball, that's 23-11 as I round it. Looking at it this way, Appier looks to be +6 wins above .500 [average]; BUT Appier was involved in only 26 decisions, going 18-8, which in fact = .692 - a better percentage than the .687 predicted by Pythago.
That 11-23 reverse record for the virtual team against Appier starts sure looks like replacement level; but even with a higher than predicted W-L percentage, Appier, on paper, looks to be a far cry from +10 anything above replacement [he would have been at 21 wins].
Facts get in the way:
-Appier STARTED 34 games; but he averages 7 IP per start.
-Appier was involved in only 26 decisions, going 18-8. Thirty-four has little to do with the actualities of his final record[s] for the season.
238.2 IP falls short enough from 34*9 [306 IP] to leave lots of wiggle room for extra "win shares" or W above replacement to get dealt to others.
I can see the logic of +5 wins, purely on the logic of comps to average. Maybe it's just me, but WAR muddies the waters too much for me to buy into it, even if it turns out tidy arithmetic results.
Registered User
Sorry I was looking at Baseball Gauge. I think they actually use a higher replacement level.
Baseball Gauge uses .380 for offense, .410 for starting pitching, and .480 for relievers. Which I believe comes out to 52 wins.
Last edited by Ubiquitous; 05-09-2012 at 04:45 AM.
Registered User
So I noticed something's up with BBRef WAR this morning. Jake Peavy is showing some new columns and his wins per run is lower than 1 to 10. They removed Rrep and it looks like they are using the leverage index for relievers and at the point they enter the game. It's not showing for Appier but maybe they are starting with active pitchers.
http://www.baseball-reference.com/pl...eavyja01.shtml
Last edited by brett; 05-09-2012 at 07:12 AM.
Posting Permissions
Reply With Quote
Bookmarks