Joe Saunders’s Deceptive Winning Percentage

Arizona Interim GM Jerry DiPoto, via Nick Piecoro of AZCentral.com

“He’s a pedigreed major league pitcher who has accomplished quite a bit. I believe it’s a .630 winning percentage in his major league career. We’re getting a pitcher for our major league club who comes in and delivers a message to our guys that this is about winning now and winning in the future.”

Jerry DiPoto has been largely panned for this statement on why he traded Dan Haren for a package including Joe Saunders and prospects, particularly among those in the sabermetric community. Of course, we know that pitcher wins don’t tell the whole story; after all, how many times have we seen a pitcher have a win blown by his defense or the lineup, and how many times have we seen a pitcher bailed out by spectacular run support?

However, with enough games, these things tend to even themselves out, and wins begin to track true talent. Looking just at starters with at least 690 innings pitched, winning percentage and ERA+ correlate strongly, with r^2 = .500. ERA+ is a simple statistic to use here, but I feel that it works given the sample, and the park adjustments come into play as well. Obviously, correlation doesn’t imply causation, but for the most part, pitchers that receive a win in most of the games that they start are better.

This is a pretty simple conclusion to make, but it does, in part, justify the statements made by Jerry DiPoto. Joe Saunders has a 54-32 career record, good for a .628 winning percentage, good for 16th all time among pitchers with as many innings pitched – 692 – as Saunders. Some pitchers with similar winning percentages? Juan Marichal, Dwight Gooden, Roy Oswalt, Rich Harden, and Justin Verlander – a pretty solid group. If Saunders’s true talent was indeed akin to these pitchers, this trade would not be so widely panned.

Saunders, however, simply isn’t that good. The correlation between winning percentage and ERA+ isn’t perfect, and Saunders is a great example of that.

We see Saunders is well below the ERA+ that we would expect given how many of his games he has won. That’s because Saunders has spent his career playing for good teams, for the most part, and also with decent defenses. What about his counterpart in this trade, Dan Haren? Haren’s 120 ERA+ is well above what we’d expect from a player who has won 55% of his decisions, although it’s not terribly out of the ordinary – Erik Bedard, Cole Hamels, and Kevin Appier have similar careers.

One of the largest differences between winning percentage and ERA+ comes from Haren’s former Diamondbacks teammate, Brandon Webb. Webb’s 142 ERA+ is nothing short of elite, ranking fourth among players with at least 690 IP and 95% of their appearances as starts, right between Roger Clemens and Randy Johnson. However, Webb has only won 58.4% of his decisions, well below the 65.8% mark from Clemens and the 64.6% from Johnson. Why? The Diamondbacks haven’t been as good as a team as the Angels or most of Clemens and Johnson’s teams. Outside of the 2007 playoff season, Arizona has had 3 seasons with win totals in the 70s, one in the 80s, and this year’s team is on pace to win 61 games.

Simply put, in many situations, DiPoto would be right to go after the pitcher with the higher winning percentage. I’m not sure that he actually used winning percentage in his evaluation – that may just be a media talking point. Regardless, this is just another example of how we need context with all of these stats, and particularly with statistics like wins. With context, it’s clear that Saunders’ winning percentage doesn’t mean he’s an elite pitcher, and Dan Haren and Brandon Webb are good to great pitchers despite poor luck in the win column.



Print This Post



Jack Moore's work can be seen at VICE Sports and anywhere else you're willing to pay him to write. Buy his e-book.


Sort by:   newest | oldest | most voted
D4P
Guest
D4P
5 years 11 months ago

Certain old-school GMs (e.g. Ken Williams, Brian Sabean, Ned Colletti, etc.) who are completely clueless with respect to how to construct an offense nevertheless seem to have had more success (whether through talent or luck) at putting together a good pitching staff.

Which raises a question:

Do old-school pitching stats (e.g. win-loss records and ERA) do a better job of predicting pitching value than old-school batting stats (e.g. BA and RBI) do at predicting batting value?

fjkagreklg
Guest
fjkagreklg
5 years 11 months ago

no

BaconSlayer09
Member
5 years 11 months ago

I’m pretty sure ERA is the best out of all the stats you mentioned.

BaconSlayer09
Member
5 years 11 months ago

I also wouldn’t say Williams is clueless at offense construction. He values things like OBP and he really likes power given that U.S. Cellular Field is a small park.

He has made some stupid moves, but for the most part, the White Sox have had pretty good offensive teams in his tenure.

Jeff Francoeur
Guest
Jeff Francoeur
5 years 11 months ago

Certain new-school GMs (Billy Beane, Andrew Friedman, Jack Zduriencik, Omar Minaya) always seem to have great pitching staffs (whether through talent or luck). Are new-school stats better than old-school stats?
Rays: 4th Best ERA (Tied)
Mets: 6th Best ERA
A’s: 7th Best ERA
Mariners: 9th Best ERA

Jeff Francoeur
Guest
Jeff Francoeur
5 years 11 months ago

I also wonder if budgets have any effect…

Dylan
Guest
Dylan
5 years 11 months ago

Completely irrelevant to what he was asking, he wasn’t saying old-school stats are better than new-school stats.

Mets Fan
Guest
Mets Fan
5 years 11 months ago

Omar is far from being a new-school GM. He’s about as old-school as it gets.

The Typical Idiot Fan
Guest
5 years 11 months ago

Good team defense doesn’t hurt either.

Doug
Guest
Doug
5 years 11 months ago

Inclusion of Omar = Sarcasm

Right?

Dylan
Guest
Dylan
5 years 11 months ago

I would definitely say yes. I think win-loss CAN be some kind of indicator of talent, in that, if offensive production evens out over time, it can be an indicator of how often a guy pitches well, which is important for a guy who isn’t playing every day. If you’re someone like Jamie Moyer, who has two atrocious starts this year (total of 4 innings, 15 runs in those starts), it is possible for win-loss to be a decent indicator of how often you do well in games you start. Obviously even then there would be better indicators, but it isn’t the worst stat in the world.

Unlike the hitting stats, which give you nothing beyond what their face value, which doesn’t tell us much.

Jason B
Guest
Jason B
5 years 11 months ago

“Obviously even then there would be better indicators, but it isn’t the worst stat in the world.”

That’s certainly true. But it’s also true that there are so, so many better indicators of pitcher quality than wins and losses. Why drive the Chevelle when you can step up to the Cadillac, so to speak?

Eric M. Van
Guest
Eric M. Van
5 years 11 months ago

Lou Gorman was a GM who excelled at evaluating pitching and was terrible with offense, and I always suspected this was part of the reason.

CircleChange11
Guest
CircleChange11
5 years 11 months ago

I’m glad to hear someone here say something positive about pitching wins and pitcher quality given a sufficient sable size.

I feel the same sort of way about RBIs. When a pitcher averages 17 wins over 6 seasons or a batter drives in 100 runs for 7 consecutive seasons, those dumb ol counting stats tell you something about the players quality.

It is doubtful that an average player plays on such great teams consistently enough to rack up those numbers.

I agree that there are better metrics, but the stats do tell you something about player quality.

For example, last year the CWS were among the lowest scoring teams in MLB with 4.4 runs per game. A really good pitcher is going to notch a good amount of wins, even with that run support.

twinsfan
Member
twinsfan
5 years 11 months ago

Indeed they do tell you something. They tell you that you’re going to need to look at the better metrics in the end anyway in order to provide enough context for those ‘dumb ol counting stats” to be really useful.

Neil
Member
Neil
5 years 11 months ago

“When a pitcher averages 17 wins over 6 seasons or a batter drives in 100 runs for 7 consecutive seasons, those dumb ol counting stats tell you something about the players quality.”

And what do they tell us, exactly?

Between 1986 and 1997, Joe Carter hit 98 or more RBI in eleven seasons out of twelve, including 6 consecutive 100+ seasons. And in those seasons, he also posted a wRC+ of 110 and a wOBA of about .338.

Taking a quick look at the last 3 seasons on the Leaders page, here are some of the most comparable players in terms of wOBA: Lyle Overbay, Chone Figgins, Jorge Cantu – all of them a solid, marginally above-average bat over the whole period.

So… a string of 100+ RBI seasons tells us that someone is at least an average player?

Dylan
Guest
Dylan
5 years 11 months ago

How did you just go from 98 to 100, and from marginally above-average to average? But anyway, yes, being a marginally above-average player isn’t exactly something to scoff at.

CircleChange11
Guest
CircleChange11
5 years 11 months ago

Wow Joe Carter. I think everyone knows that situation.

At least you didn’t compare him to Dave Kingman, like BP.

What other players have accomplished a long string of 100 RBI seasons?

Like I said, the are BETTER metrics, but that doesn;t mean W and RBI are completly useless.

Not surprisingly, wOBA doesn’t like power hitters that do not walk a lot. Looking at Carter’s WAR … it’s all over the place, ranging from negative to 5+, yet his HR and RBI are consistent (whether he’s playing for the 86 Indians or the 92 Blue Jays).

It’s important not to take any stat in isolation.

Daern
Guest
Daern
5 years 11 months ago

Dylan–you were making a good point. Well, not really, but one that had elements of veracity. And then you ruined it.

“How did you just go from 98 to 100”

These sorts of arbitrary cutoffs are one of the main things wrong with counting stats, man! Don’t fall in that trap.

And while being marginally above-average isn’t to be ignored, Neil’s point–which I suspect you knew and were just being difficult about–is that RsBI and wins may tell you something after a while, but only in CONTEXT. Alone, they are next to useless.

Daern
Guest
Daern
5 years 11 months ago

Okay, wait. That was wrong. The first part is to Dylan, the overall theme is to CC. Mah bad.

Grovaltor
Guest
Grovaltor
5 years 11 months ago

What does wins and loses tell you that other, better metrics can’t?

Jason B
Guest
Jason B
5 years 11 months ago

EX. ACT. LY. what grov said.

Ryan Howard
Guest
Ryan Howard
5 years 11 months ago

“It is doubtful that an average player plays on such great teams consistently enough to rack up those numbers.”

If you meant this to be taken 100% literally, that is certainly true. There’s a sample bias there. The average player doesn’t get to be on great teams usually. He is signed by mediocre teams and to shorter contracts. This means he is much less likely to be surrounded by good hitters. Alternatively, a great player will often be signed by a great team- either due to good scouting or payroll advantages (i.e. your best bet for 100 RBI starts with playing on the Yankees).

But because of this, there’s a chicken-egg situation. Do they have so many RBI’s because a good team put them in the cleanup spot? Or are they in that position, on that team, because they hit well? Both are true, and with respect to RBIs I think it’s tough to say how those factors balance out. So sure, 100 RBI means you can hit very well. … Or it means you’re batting behind Chone Figgins and Kevin Youkilis when they’re in career seasons. Because that will probably lead to a 15-20 RBI swing versus having a couple Bengie Molinas in front of you.

I’d say if people wanted to use RBI’s, why not at least look at RBI conversion percentage? The # of RBIs one could have gotten versus one got? Or maybe an RBI opportunity WPA thing? The % of WPA a player added during RBI opportunities versus the maximum possible? (or alternatively, versus the league average) Those would both seem to tell you something that is primarily about the batter (and secondarily about how they’re being coached).

Mike
Guest
Mike
5 years 11 months ago

I know which one of those points is Matt Cain.

Kevin Yost
Guest
Kevin Yost
5 years 11 months ago

haha. I was looking for his point also. Poor Matt Cain : (
He has about the same ERA as King Felix in his career but nobody thinks he’s real.

Dealer A
Guest
Dealer A
5 years 11 months ago

Ha, I went right for Cain too.

Then I tried to plot Ben Sheets 2004 season, but I realized the graph only went to 160 ERA+.

awayish
Member
awayish
5 years 11 months ago

am i supposed to be acting all deceived and surprised? i must have missed the memo.

BX
Guest
BX
5 years 11 months ago

No, you’re supposed to be thinking that Jerry DiPoto is not that smart.

CircleChange11
Guest
CircleChange11
5 years 11 months ago

For kicks, I just looked up ALL of Saunders’ game logs to see whether he’s winning games he shouldn’t. What I found can be summed up generally, “When he’s on, he’s on … when he’s not, he’s not”.

I used the criteria for quality start to pull his “good starts”.

By year …

07: 9
08: 22
09: 13
10: 9
Total : 53

ERA in those QS …

07: 68 IP, 10 ER (1.32 ERA)
08: 155 IP, 35 ER (2.03 ERA)
09: 91.2 IP, 16 ER (1.57 ERA)
10: 67 IP, 10 ER (1.34 ERA)
Total: 381.2 IP, 71 ER (1.68 ERA)

Like I said, when he’s on, he’s on. Now subtracting these from his career totals to see what his stats are when he’s “not on”.

On: 381.1 IP, 71 ER (1.68 ERA)
Off: 310.1 IP, 259 ER (7.51 ERA)

He “earns” both his wins and losses.

I wish a stat like QS% was more common, as to indicate how many games a pitcher gives his team a good chance to win.

One also has to account for the “quality of opposition/offense” a pitcher goes up against. IMO, this is VERY important and should be reflected in a pitcher’s stats.

Saunder’s QS per opponent …

OAK: 10
SEA: 7
CWS: 6
TOR: 4
TEX: 4
COL: 3
KCR: 3
BOS: 3
BAL: 3
NYY: 2
MIN: 1
CLE: 1
TBR: 1
LAD: 1
ATL: 1
PHL: 1
DET: 1
SDP: 1

Daern
Guest
Daern
5 years 11 months ago

Ah, but how well do his QSs correlate to victories?

CircleChange11
Guest
CircleChange11
5 years 11 months ago

Oops, I counted 2010 twice (one as 2010, and once in place of 2007), here are the new totals, and the correlation you asked for.

2007: 12 QS (8-2-2), 77.1 IP, 22 ER (2.56 ERA)
2008: 22 QS (14-2-6), 155 IP, 35 ER (2.03 ERA)
2009: 13 QS (12-1-0), 91.2 IP, 16 ER (1.57 ERA)
2010: 9 QS (6-2-1), 67 IP, 10 ER (1.34 ERA)
TOTAL: 56 QS (40W-7L-10ND), 391 IP, 83 ER (1.91 ERA)

QS: 40-7 (10ND) — 391 IP, 83 ER (1.91 ERA)
Not: 14-25 (19 ND) — 301 IP, 247 ER (7.39 ERA)

Saunder’s QS per opponent …

OAK: 10
SEA: 8
CWS: 5
BAL: 4
TEX: 3
COL: 3
SDP: 3
BOS: 3
KCR: 2
TOR: 2
NYY: 2
MIN: 1
CLE: 1
TBR: 1
LAD: 1
ATL: 1
PHL: 1
DET: 1
STL: 1

Quality of opposition is certainly a factor … the REAL question is “how much compared to other pitchers?”

CircleChange11
Guest
CircleChange11
5 years 11 months ago

“Quality Start W-L-ND By Earned Runs Allowed”

0 ER = 12W – 0L – 0ND
1 ER = 10W – 2L -3ND
2 ER = 14 W – 2L – 1ND
3 ER = 3W – 3L – 4ND

Daern
Guest
Daern
5 years 11 months ago

Jack — Which of the pitchers on your graph least “earned” his victories?

BuzzingThalami
Guest
BuzzingThalami
5 years 11 months ago

I think back to a certain absolutely atrocious Astros starter who went 8-16 in 1987. I imagine DiPoto would have simply released Nolan Ryan, despite his leading the league in ERA with that record (okay, it was the Astrodome).

About the only way you can spin those comments well for DiPoto is to imagine that they had other reasons that the total package added up for them, but assumed the media would be too stupid to understand anything but winning percentage.

It takes quite a contortion act to manage, but I’m trying.

CircleChange11
Guest
CircleChange11
5 years 11 months ago

When you mentioned “led the league in ERA” you eliminated the factor of W-L record.

Does anyone really think if Haren was leading the league in ERA, he would have been traded? … or at least been traded for more prospect value.

The problem with Haren (in the eyes of AZ) is his ERA, not his W-L record.

Also, AZ is reffering to W-L record over 3.5 seasons (Saunders), not just a single season.

I don’t dispute the point you’re making, only the two situations really aren’t comparable at all.

BuzzingThalami
Guest
BuzzingThalami
5 years 11 months ago

My comment was more about DiPoto’s apparent reliance on W-L as a valuation method for Saunders. By his logic, Ryan was a pretty awful pitcher in 1987.

CircleChange11
Guest
CircleChange11
5 years 11 months ago

Dan Haren QS stats as a Diamondback:

2008: 23QS (70%), 162.2 IP, 36 ER (1.93 ERA)
2009: 24 QS (73%), 173 IP, 36 ER (1.87 ERA)
2010: 12 QS (55%), 86 IP, 25 ER (2.62 ERA)
TOTAL: 59 QS, 421.2 IP, 97 ER (2.07 ERA)

QS: 59 QS, 421.2 IP, 97 ER (2.07 ERA)
Not: 29 GS, 170 IP, 137 ER (7.25 ERA)

Wins and Losses
—————
QS: 35W – 9L – 15 ND
Not: 2W – 18L – 9 ND

W-L-ND By Earned Runs Allowed:

0 ER = 7-0-2
1 ER = 11-2-5
2 ER = 11-4-3
3 ER = 6-3-5

Dan Haren has been outstanding as a DBack (and just as good as an A), but I don’t find him W-L record to be all that “hard luck”, nor do I see a “pedigree” of winning with Saunders.

I’d be their is a correlation leaguewide with the number of ER allowed and Winning Pct (duh). I am doubtful that any pitcher is consistently “good luck” or “bad luck” over 5 or so consecutive seasons, although it certainly can (and has) happened in a single season.

Even the lowest scoring team in baseball will average 4 runs per game over the year. Giving up 2 ER or less over 6+ innings, is going to get you a win a lot of the time (75+%)?

Steve Marino
Guest
5 years 11 months ago

If you flipped a fair coin 86 (54+32) times and assuming a normal distribution then there is insufficient evidence at the 1% level to say that the coin is anything but fair. Under that basis then a coin flipped 86 times but comes up heads 54 times (a .627 percentage) could very well be a fair coin that in the future will give a 50/50 return.

Really, some GMs say the funniest things.

Oscar
Guest
Oscar
5 years 11 months ago

I’m sorry, but this is a terrible article. That pitching wins correlate reasonably with pitching skill over a large enough sample is obvious. It is still completely useless as a predictive statistic, especially in the context of having readily available statistics that are a level of abstraction closer to true talent (e.g. ERA).

You know what else correlates with ERA+? Innings pitched, because bad pitchers don’t have as long careers. Does that mean that IP has ANY value as a statistic for predicting a pitcher’s true talent? No.

studes
Guest
studes
5 years 11 months ago

Jack, I’ve got a graphing tip for you. On scatter plots, be sure to put the independent variable (in this case, ERA+) on the “x” axis. Makes interpretation much easier, in that pitchers who posted more wins than expected are above the line (more=above).

Unless you’re trying to say that win-loss percentage is the independent variable. If that is the case, I’ll leave an entirely different type of comment!

neuter_your_dogma
Guest
neuter_your_dogma
5 years 11 months ago

“We see Saunders is well below the ERA+ that we would expect given how many of his games he has won. That’s because Saunders has spent his career playing for good teams, for the most part, and also with decent defenses.” I agree. But what explains the disparity between pitchers on the same team?

wpDiscuz