All Time All Mariner Roster: Center Field
Warning: This is both very long and very awesome.
Much like with the shortstop discussion, we’re not really discussing players, but rather years. The answer is Griffey. The only real question is whether we’re going to go with his 1993 or 1997 season. I’m going 1993.
’93 was his first year as a monster, and at 23, he established himself as one of the best young players we’d ever seen. The .309/.408/.617 mark made his offense worth about 61 runs better than a league average hitter that year, and he did it with an all around attack – 41 doubles+triples, 45 home runs, 96 walks, and 17 steals. He also played in 156 games and racked up 691 plate appearances, so it’s not like his rate stats were inflated by playing in a shortened season.
Among AL hitters that year, only Frank Thomas and John Olerud were better, and they both played first base. He finished 5th in the MVP voting, mostly because a lot of people in the BBWAA don’t really understand baseball, but once you factor in the positional difference, it’s not hard to argue that Junior was as good as anyone else in the league that year. 1993 was the year where Junior broke through, became a superstar, and put baseball in Seattle back on the map. He would have some more great years later in his career, but none ever quite matched that season – the best he’s ever had, and the best this organization has ever gotten from the position.
I also have it 93 or 97, and not at all shockingly, I went the other way.
In 1993, the league line was .267 .337 .408
In 1997, it was .271 .340 .428
That’s entirely normal, though the hitting blip is interesting.
Similarly, to look at the average at his position:
In 1993, the average CFer hit .275 .339 .410
In 1997, the average CFer hit .269 .343 .418
The 1997 is a little bit better. So when I compare the two years offensively, and 1997 comes up -.005 on average, -.026 in OBP, and +.029 in slugging, I wouldn’t make that swap. So I think 1993 might be the better season in raw offense, thought the difference is a couple runs if that.
The stolen base-thing, though: in 1993, he was 17 for 26 attempts, and 65% success is below even a generous break-even. 1997, he’s 15 for 19, a 79% success rate, which is valuable to a team. That’s a couple run swing in the other direction.
In terms of playing time, though… In 1993 he played 139 games in center and 19 at DH. In 1997, he played a full 153 in center. As we saw in the Davis-Olerud debate, I put a fair amount value on the player being in the lineup at the position we’re debating. 14 games is another … well, it’s not ten percent of the season, but it’s close.
If Griffey played another 14 games in center in 1993, you can have your choice of who you want out, either entirely or DHing:
– Brian Turang (“He’s sooo cute, Turang Turang Turang”) or
– Henry Cotto
Fourteen games of having some random guy running around in center, even if Griffey was DHing, makes a difference. Moving Griffey’s bat from CF to DH means you go from a stellar offensive contribution from a premium defensive position (and, presumably, a decent contribution from the DH) to getting that scrub’s offensive contribution from center and Griffey’s bat competing against DHs. That’s a fairly big deal to me.
In 1993, the average CFer hit .275 .339 .410
In 1993, the average DH hit .262 .332 .425
I hate it when the results get in the way of a perfectly good, logical argument like that. Griffey and Dykstra were great and really, it was just Paul Molitor at DH that year.
You get my point, though.
The only way I think I’d accept Griffey 93 over Griffey 97 is if Griffey 93 played such good defense compared to 97 that he made the same or better defensive contributions.
But how do we do we do that? Fielding percentage?
Griffey 93: .991 against a league mark of .986
Griffey 97: .985 against a league mark of .982
Range Factor? Do we even want to touch range factor? Griffey looks way worse in 93 compared to 97 there. Fielding metrics you don’t like have 97 significantly better than 93. Moreover, though, I generally try not to compare defensive stats season by season if I can avoid it. Looking at their three-year runs, the 92-94 block looks much worse than the 96-98 block.
I know that generally speaking defense declines first in a player’s career, but it looks like Griffey was getting much better with experience, broke his hand in 1995, came back, kept getting better, and really only started to tank from 1999 on (which, I think significantly, is the last time he won one). On the other hand, he won his first one in 1990. On the other other hand, Gold Gloves suck.
But I think my point here is even better than in the Davis/Olerud argument: if you think that their offensive performances are about the same (and with the stolen bases, I think so) then the decision focuses on defense and playing time. In 1997, you get Griffey for an extra 14 games out there in center, and by all available evidence he was as good or better defensively than 1993 Griffey.
And I agree that 1993 was a job on Griffey. If anyone ever needs a basis to argue for an east-coast media bias (which is totally true and I don’t know why people argue this isn’t true) I’d ask this: what’s the tally on legitimately-deserving-but-got-screwed versus undeserving-and-won-anyway awards in Seattle? Or the West Coast, for that matter?
These arguments are why Baseball Reference is so awesome. We can throw the hypotheticals right out the window and just compare the difference between the center fielders that played when Junior DH’d to the Mariners actual DH’s that year (and it’s hilarious how many they had). So, here goes.
Replacement CFs: Brian Turang (51 PA), Henry Cotto (30 PA), Mike Felder (23 PA), Dann Howitt (6 PA), Lee Tinsley (1 PA). As a group, they got 111 plate appearances while playing CF and racked up a pretty sorry .215/.260/.377 mark. So Derek’s on to something – the guys who had to play CF when Junior DH’d were terrible.
But don’t celebrate just yet – the M’s DH’s were pretty horrible too. Led by Pete O’Brien (who got the most PAs as DH), the Mariners got a .240/.330/.357 mark from their non-Griffey DH’s. Apparently, Edgar Martinez is the only guy that could save this organization from just wasting the designated hitter role, but that’s neither here nor there. Back to the Griffey argument.
If we’re going to deduct 1993 Junior for forfeiting ABs to useless players, we have to at least make sure we deduct the right amount. We’re really looking at the net difference between a .215/.260/.377 line vs a .240/.330/.357 mark over 62 plate appearances (the 111 that he ceded to reserves in 1993 minus the 49 he ceded to reserves in 1997). Yes, the DH’s were better, as OBP is more important than SLG, but a 50 point OPS difference over 62 plate appearances is about two runs.
It matters a little bit, but it doesn’t matter enough. Even including the SB rate, Junior’s offense by lienar weights was 3 runs better in 1993 than in 1997. Accounting for the playing time given to the backups, the gap closes to one run, meaning that the difference is basically meaningless, and for all intents and purposes, they were pretty much equally valuable offensively.
So, we get back to defense. And this is where I threaten to put you in the moderation queue – fielding percentage? Really? You just quoted fielding percentage. Punch yourself in the face.
I feel like we’re just going to have the Davis/Olerud argument again here, where you say that the best available metrics we have suggest that Junior was better in 1997 than he was in 1993, and I say that the best available metrics aren’t anywhere close to being good enough to give them any real credence. Fielding Percentage, Range Factor, and the BP stuff are equivalent to having a rowboat in a race around the world. Yea, it’s better than swimming, but the wise man just says “screw it, I can’t get around the world in a row boat” and goes home.
So, we’re left with a dilemma. 1993 and 1997 Griffey were basically equal offensively, so if there was any difference between them, it was with the glove, but we’re basically unable to measure the defense well enough to make a real call. So what do we do?
I say we look at his situational performance. Whether it’s a skill or not, getting a hit with the bases loaded is more valuable than a hit with the bases empty. Fangraphs charts WPA/LI, which is basically linear weights by game state, giving hitters more credit for hits where the run value is higher. This essentially encapsulates most of what we consider clutch hitting, just without the context specific element (whether the team is up by 10 runs or 1 run doesn’t matter in WPA/LI, and neither does the inning).
In 1993, Junior posted a 5.81 WPA/LI (it’s quantified in terms of wins added, so read that as 5.8 wins added offensively). In 1997, he posted a 6.06 WPA/LI, thanks to a better performance in high leverage situations (.966 OPS in highlev spots in 1997, .709 OPS in highlev spots in 1993). For whatever reason, Griffey was pretty horrible in game changing situations in ’93, enough that his inferior context neutral 1997 performance passes his 1993 mark when we include base-out state into the equation.
So, I guess I concede – 1997 Griffey it is, but really, by the tiniest of margins. We’re really splitting hairs here.
Also, we should note, just so people realize how ridiculous this is – the top nine offensive center field seasons in Mariner history all belong to Griffey. 2001 Mike Cameron climbs the list once we include his tremendous defense, but he still can’t get past these two amazing seasons Griffey had. They were just remarkable years from a remarkable talent.
That was a sweet DH/CF replacement argument. Still, I wonder how much we should consider how well the M’s did with the player or players available to them in a team sense in looking at a player-season in isolation. That’s a whole potentially crazy conversation on its own. And I’m still feeling that given the equal offensive contributions, if you had the choice between two Griffeys who could play x games, hit y, and play z-quality defense, that you’d rather take the one who could give you more games in center.
But certainly, in terms of what the M’s actually did with those jokers, you’re entirely correct, it didn’t make a difference. And to advance my own argument against myself, in a way I’m sort of arguing that a manager’s choice of where they play a guy can have an outsized effect on their value when it comes down to close situations like this. I’m somewhat uncomfortable with this, because if Griffey (or Davis, to return to that) isn’t injured and could have played center, does that really detract from their value?
That’s a tough answer. I’ve clearly opted to argue for the players that took more playing time at each position we’re considering.
Yes, I quoted fielding percentage. And I will not punch myself in the face.
Here’s the thing — fielding percentage is not a good stat, but I think your analogy does it a disservice. I’d say it’s more like navigation, where we start out orienting with map and compass, and now we’ve got much better maps based on high-quality surveys, we can use good clocks and other observation equipment, but we’re not anywhere near GPS yet. But you can still look at, say, an observation log by someone in the past and figure out where they were — with much less confidence, of course, but still.
I acknowledge entirely that it particularly penalizes players with great range and poor hands or throwing skills, because they’ll get to a ton of balls while making more total errors, and unfairly rewards Raul Ibanez — I mean, immobile players with steady gloves and weak but accurate arms. And I further concede that it’s a stat best suited for when errors were far, far more common.
But fielding percentage, and range factor, and zone rating, are bad stats that are better than nothing. Zone rating correlates extremely well with UZR, which pretty much everyone agrees is a fine system, and decently well with PMR, which is also one of the best (the correlation was .81 with UZR and .45 with PMR in this comparison).
Fielding percentage is not that bad. I ran the correlation between fielding percentage and UZR for shortstops with more than 100 expected outs last year, and it was .484 for the 29 shortstops who got over that bar.
If we consider for a second UZR the gold standard, that’s better than the Davenport Translations did in that correlation article I just linked to, and it’s not that far behind PMR. PMR! My other gold standard!
And it correlated at .281 with PMR, which is way behind the other candidates (which were around .45-.60) but it’s halfway between random numbers and the best systems we have right now. I don’t think it’s fair to dismiss it outright as meaningless or worthless, when it’s worth something.
Now having done that, I realize I should have done it for centerfield. Shoot. I figured I’d have more opportunity data with shortstop.
And yes, that’s one year’s worth of data, at one position, assembled by me and not someone more adept at this.
I do wish, though, that there was more research on this kind of thing: there’s a lot of debate to be found on the merits of favorite advanced defensive metrics, but there’s not a lot that says “Hey, WACKO stat measures defense amazingly well, so well it’s yay much better than UZR, and here’s its correlation to current metrics and readily-available metrics…”
I’m getting away from my point. If you’re in the wilderness and all you have is a map and a compass, the proper thing to do is go orienteering, rather than throw both away. Actually, I’d try and signal a plane with the reflective surface of the compass or something. I digress again.
We’re arguing defense in the 1980s, and all we have are rough measures. Using those rough measures, in the same way I was trying to argue we could do with Olerud, you can both acknowledge their shortcomings while coming to some cautious and reasonable conclusions.
And now for the twist ending I’ve set myself up so, so adeptly for… fielding percentage doesn’t at all support my assertion that Griffey’s defense in 1997 was better than 1993. At all. 1993’s FP is .991, and 1995’s is .985. Over those three-year spans I’m talking about, 92-94 is .991 and 96-98 is .988.
(random aside: I realized I misquoted league FP in 1997 as .982 instead of .988 — .982 was the FP for all outfielders. This doesn’t affect your response.)
I laughed when I realized this. And looking at only his fielding percentage numbers, he peaks at 22 and then there’s the long decline, which is what you’d expect given a standard aging. Soooo… uhhh…. his range factor scores! Yeah, let’s look at those!
I have no idea why the historical defensive system that will not be named rates 1997 as 10 runs better than the 1993 season. The difference in fielding percentages would — and please cross-apply all previous disclaimers and such — make me think 1993 was better by ~5 runs. I wish I had some more data to throw at this. If it’s five or ten runs difference in 1993’s favor, that’s our winner, I think. If it’s five or ten in the other direction — yeah. I’m in the woods, it’s night, it’s overcast so I can’t see the stars, and I’m pretty sure I’m near one of two towns, but I don’t know.
I believe this is the first one where at any point in the conversation we’ve ended up on different and still-opposing decisions. Please, be gracious and take 1993 Griffey so we can move on.
On the side note you raised — Griffey’s dominance is crazy. Even for an expansion team, the dominance at his position, the consistency of his contribution during his time here… it’s amazing to view in context. There isn’t a Griffey season that isn’t good. When I made my list I was curious about this and ended up with about seven Griffey seasons (including 94) and then Ichiro’s 07 about even with Cameron’s 01 and Griffey 92 and 99. What’s Griffey’s worst season here, 89 in terms of contribution while healthy, and 95 in terms of total contribution? If you dropped the other Griffey seasons from the franchise record, both of those would both be in the franchise top ten (assuming, of course, that whoever they got to play center those years would suck, which is perfectly valid because they did suck — 87-88 was Brantley, Moses, and Cotto. Hah.)
So, this little rant is only going to sort of be related to Griffey and Mariner center fielders, but it’s on topic, given your last awesome response. Here me out.
Correlation analysis is great – very useful, helps us understand what matters, and leads us towards better understanding. But, when used incorrectly, it can become, for lack of a better term, anti-knowledge. What do I mean by anti-knowledge? I mean statements that are factually correct but lead us away from the truth. We’ve all seen anti-knowledge at work in baseball – let’s use Willie as a pinch hitter because he’s 4 for 7 lifetime against the reliever they’re bringing in. Yes, it’s true, he’s 4 for 7, so that fact passes itself off as knowledge, but it means nothing. It tells us nothing, and helps us decide nothing. It is not knowledge – it just appears to be knowledge.
So it can be with correlations when one of our variables has no causal effect on the other. They could pass out neon wristbands at the all-star game for all the guys there to wear til the rest of the year as tokens of their greatness, and if we ran an end-of-year correlation between wristband wearing and second half performance, it would probably be pretty high. But that correlation would be anti-knowledge – a fact that distracts us from the truth.
So it is, in my opinion, with a correlation between fielding percentage and our perception of defensive ability. Some players will possess both real defensive abilities and low error rates, but I don’t buy the relationship between the two as real enough to act as any kind of proxy.
And this gets back to my larger point about why I disregard range factor, fielding percentage, and the BP defensive metrics entirely – they just don’t measure the things they’re supposed to measure. They might act as a proxy for some small part of defensive ability, but they miss so much of the real picture that we can’t allow their insights on 5% of the whole to lead our guesses about the rest of the 95%.
Or, to redo your map/compass analogy, trying to conclude anything based on those metrics is like trying to do astronomy from a geocentric perspective. Yes, if you decide that the earth is the center of the universe, you can come up with some plausible explanations for how things work and some of them might even be true, but in the end, you’re still totally wrong about the entire makeup of the solar system. Ptolemy was a pretty smart dude who produced some real astronomical observations, but that doesn’t change the fact that his views on how the solar system worked were almost entirely wrong.
There has to be a threshold of usefulness where we say “if you can’t measure the things we’re trying to measure, and there’s no cause-and-effect to allow for a real proxy, then I can’t put any emphasis on the data”. These metrics aren’t better than nothing – they’re worse than nothing, because while not leading us to the truth, they simultaneously give us false confidence that we’re on the right path. They are anti-knowledge.
Griffey is actually a perfect example of why. If I was going to put any stock in any defensive metric from pre-2001, it would be Sean Smith’s TotalZone system, which he’s built from Retrosheet data and a unique approach based on Tom Tango’s “With or Without You” model to use pitcher, hitter, and park as the proxies – you can read about the system here if you want. It’s far from perfect, but it’s better than the RF/Fld%/BP trifecta of doom, and it actually agrees with your initial assessment – Junior was bad early on, improved, then got bad again as his legs gave out. If you wanted a defensive system to back up your initial argument, TotalZone would be the way to go.
So, in the end, we have two Griffey seasons of near equal offensive value, with 1997 winning the clutch-hitting tiebreaker. The true difference between the two certainly lies in the defensive performance, but we have differing takes on whether we can actually measure that defensive performance enough to draw a real conclusion. While I’d like to claim victory and claim 1993 Griffey as the winner, I just can’t – after talking through all this, I don’t have any leg left to stand on to support a 1993 > 1997 theory. The real answer is we don’t know which season was better, but if we’re forced to pick one, there’s at least a bad argument for 1997, which trumps no argument for 1993.
Come back to your original side in the interest of harmony. 1997 Griffey it is.
That’s a very good argument about anti-knowledge. The thing I would mention though is that errors are not antiknowledge, and fielding percentage does measure something really useful, if not well.
Fielding percentage supposedly measures the number of times a player doesn’t get an out on a play they should have, compared to an average player.
In practice, it really measures something different, which is more like
number of times a player badly bobbled or threw away a ball / number of times a player got to a ball
That’s valuable information. Every advanced metric incorporates that information somehow, either in outs created on a better measure of opportunities, using throwing errors, or whatnot.
Now the old James argument is that having “number of times a player got to a ball” doesn’t adequately measure effectiveness. Derek Jeter, for instance, may not make any errors while waving his glove at balls rolling slowly past him all season, and that doesn’t make him great, though fielding percentage could be 1.000. And for extra criticism we can talk about the handedness of the pitching staff and other factors affecting the difficulty of the play.
At shortstop, since I crunched those already, the counted-by-scorer opportunities (as PO+A+Errors) varies from ~2.3 to ~3.5 per game. This makes range factor pretty useless, but that’s another story.
In practice, though, it works out okay: errors, in turning an out into a free runner, are like letting those singles past. If you compared two players with the same range, one of whom made a ton of errors, the second one would clearly be the worse fielder. Like last year — look at Yuni compared to another average-ranged shortstop. Yuni made 14 yippy-type errors, and you’d take a steady-handed, more accurate-armed version in an instant. Fielding percentage would differentiate between those two shortstops.
Then the question you have to ask is — how many more plays would that average-but-error-prone shortstop have to make in order to compensate for the high error rate? That’s not hard — you just work through the relative run values of errors against preventing singles, using your favorite calculation system, and voila — your middle-of-the-pack shorstop has to be about #8 or so in the league in getting to balls for it all to even out (and I think that’s low, but I’m doing napkin calculations here).
See, and now we’re in trouble, because most players aren’t #8. If you were going to make a blind assumption about the range of any player, you’d pick “average” for obvious reasons, and that’s where most players are clustered around. It’s hard to get to that many additional balls and make plays on them. Most players don’t.
Because here’s the flaw in James’ argument, as I see it: the player who gets to a ton of balls and makes a lot of errors is less common than the player of average range who is pretty steady. It’s not like fielding percentage is out there tagging 80% of players with horrible stats. In most cases, a player who can get to a ton of balls is able to be in position to make plays easily that result in errors for the other player, and they don’t see such a huge expansion of balls they can get to that the tradeoff is easy to make.
What you see in the weak correlation, I think, is not anti-information at all. It’s that errors, even measured by the official scorers, are an important indicator of a player’s defensive ability. It’s how often they totally blow a play they should make.
It’s like… it’s like batting average. It too has a bad denominator, though not as bad as FP, in which some things are arbitrarily removed and others count, but it correlates well with run scoring. This breaks down a little here, as batting average is way, way better correlated to offensive production than we think fielding percentage is. Wait, even better — it’s like if we were trying to evaluate historical offenses and all we had were walks, or stolen bases. Stolen bases might be the best proxy here. Teams can win without stealing, they can lose while stealing, but overall, teams racking up stolen bases have more runners and better baserunners, and they’re better offenses.
If you had more information to use, you would. But if that’s all you had: in 1902, Team A stole 150 bases in 140 games and Team B stole 90, well… for me, I’m willing to say that given the crappy data available, it’s pretty likely that Team A was the better offensive team, but there are many factors that could make that not true (and so on and so forth).
I feel like that’s where I am trying to look back on pre-play-by-play defenses — we’re armed with one dull knife to go kill some bison for the tribe. I’ve love to be better equipped, but I’m not, and we’ve all got to eat.
Have I used enough bad analogies yet? I think I have.
And fielding percentage correlates pretty well with overall team run prevention (this year in the AL, it’s a lot higher than walks allowed per game, and lower than strikeouts/game, for instance)(and again, I’d bet there’s a good chance someone who wants to do some serious data spelunking would be able to put way better numbers together).
None of which should be taken to read that fielding percentage is a particularly good stat, and certainly it shouldn’t be used where you have any kind of decent play-by-play information available. But as long as we’re willing to go into discussions about defense knowing its limitations and being cautious about its use, it can be helpful.
I totally dig the TotalZone stuff, though. That is some great and worthwhile work. And because it favors 1997 Griffey, I’m going with 1997 Griffey, and we can move forward.
Random aside: MGL did a “best at throw error saving” I came across while researching this, and Olerud came out at #5.