Okay, just throwing stuff out there, but do you notice a lot of times, when writers attempt to disparage a good player, they refer to their numbers like they were compiled for the good of a contract and not the team, or some other bizarreness?
Then why has saves not become the most villified statistic of them all?
We all know in here that using your best reliever in 2-3 run games in the 9th for anything other than keeping him sharp is useless, as even a minimally competant MLB pitcher converts that save. For a lot of relievers, though, it's all about the almighty save, at least the top ones I should say. Often times, relievers are allowed to pursue and maximize their own save chances, often at the expense of the team (like I sort of already said, the save sometimes happens in the 7th, or the 8th, hey, maybe even the 6th).
I thought of this because I was reading the Jon Heyman dud of an article on his Hall of Fame vote, where he refered to Blyleven as a "compiler", all the while voting for K-Rod as AL MVP. And yes I know that there needs to be a little bit of an edge to a guy to pitch the 9th, and the current closer model works fine when you have a loaded pen a la the Red Sox / Yankees / etc, but is it really that necessary to use your best guy to shut the door on a 2-3 run lead?
There's so much to be said for this argument that it's hard to know where to begin. Yes, saves are entirely meaningless. Yes, they define "compiler" stat (at least with HR's/RsBI/R's for hitters or K's/W's for pitchers, the player is contributing to the team, whereas closers can be awful and still get saves - see Fuentes, Brian). Yes, the most important situations often happen in the earlier innings. Yes, it's wildly wrongheaded to use your best guy to come in and get three outs with a three run lead.
It's really quite bizarre how the writers have come to regard saves as the end-all, be-all of stats. They embrace it so easily, and yet they dismiss out of hand stats that make so much more sense - and to provide an example here is rather pointless, as the possibilities are nigh-endless.
But what's more bizarre to me is that teams are paying for saves. K-Rod gets a record-setting contract because his team handed him the most 9th-inning leads, and even though he wasn't particularly dominant, he converted 62 of them? He deserves a congratulations for breaking the record, sure, but not a $40 million contract. This is especially true because of the wildly varying nature of relief performances. With so few innings thrown per season, performances can fluctuate without rhyme or reason; it's the rare reliever who can be like Mariano Rivera or Joe Nathan and be consistently dominant, but guys can pull a Brad Lidge turn and go from great to pumpkin in one season.
I'm with Padman on this. The K-Rod example seems apparent given that, as of this posting, he's only had 34 saves with the Mets...the whole team can only win 67 games, so he's saved about half of those...but you can't get a save unless your team is in the ball game at all, and I can tell you from having watched a lot of Mets games this year, they haven't been.
Pursuant to this observation, it would be interesting to gauge the relevance of the save relative to the record of the team. Something like (Saves for Pitcher) / (Team Wins over .500) ? For the K-Rod and the Mets at this point in 2009, that would be 34 / -18, for a Team Save Value For Pitcher of -1.89. For K-Rod and the Angels last year, that would be 62 / 19, for a Team Save Value For Pitcher of +3.26.
If the team lost even though it had this player as a save artist, then the stat will be negative. If the team won, regardless of whether it had this player as a save artist, the stat will be positive. Because of the problem involved with dividing by zero, the stat would have to be simply be set to 0 if a team was precisely at .500.
If the player saved a number of games equal to or above the number of games the team was above/below .500, then the stat's absolute value will be above 1, indicating the player contributed a large number of saves. If the team is above .500, then the stat suggests efficacy of contribution. If the team is below .500, then the stat suggests futility of contribution.
So it would look like this:
+1 and above, the save artist contributes meaningfully to
team wins (the more above +1, the more the
contribution of the save artist)
over 0 but under +1, the save artist isn't contributing
meaningfully to team wins, but the team is at least
still winning
at 0, the team is exactly at .500, so any save provided
by the save artist is neither particularly helping nor
particularly hurting the team be a winning team
under 0 but over -1, the save artist isn't contributing
to team wins, and the team is losing
-1 and under, the save artist contributes to team
wins (the more below -1, the more the contribution of
the save artist) but because the team is losing, this
kind of contribution doesn't matter much.
A stat like this could make the small impact of save artists a little more clear, I'd think.
This doesn't work. Your equation means that a closer on an 82-80 team would be a whole lot more influential than that same closer on an 81-81 team. We all know that's not true. Dividing by a number approaching 0 increases the quotient, which further contradicts your scale.
I don't see how games over .500 will help the cause; it's just a transformation of using Saves/Wins. Win Probability Added is about as good of a way to show true value in relief innings as I've seen, but it would be interesting to see if saves can be "valued" in a fairly simple matter.
Your first comment is to the point, the second one isn't, I think.
If we wanted to be really mathematically accurate, I agree, we would be concerned that the number is really high when the team is at +1 or -1 with respect to .500 and flattens out when the team is more games above or below .500. There would have to be some kind of a scaling factor...and perhaps if such a factor actually were to be contemplated, it would be at that point that people might wonder whether this stat is more trouble than it's worth to create.
I do think that Games Over .500 is the point, though. Otherwise the Mets could honestly say that K-Rod is worth the gigantic contract they gave him, because he did save around half their games. The initial starting point from which I was proceeding, though, was that it radically isn't worth paying someone to pitch a few pitches in a few ball games when it really isn't helping your team be a winning ball club. Being some number of games over .500 is the definition of a winning ball club.
Anyway, I agree that the marginal effect does matter. Being a save artist that can help a team be +1 over .500 isn't several orders of magnitude over being a save artist that keeps the team at .500, and the impressiveness shouldn't steeply decline when the team is at +2 over .500. But I still think the stat gives a pretty decent idea of whether a save artist is actually helping the team or draining it of its finances.
Oh, and I agree that a more sophisticated stat than Saves (like WPA) would be a better measure of effectiveness, but that wasn't the theme of this thread...the theme was whether Saves is a good stat or not, and my comments were aimed at showing that it wasn't.
I think everyone who uses fangraphs knows saves is a dumb stat, this was more addressing how players like A-Rod can be disparaged for putting up numbers "selfishly", but so many sportswriters not only don't give saves the same treatment, but view them favorably.
That was what I was trying to discuss - K-Rod's saves with the Mets come off a bit more "selfish" because they're not really helping the team win, whereas K-Rod's saves with the Angels in 2008 come out a bit less "selfish" because they did seem to contribute to a team effort.
I'm not really sure a really good stat can be designed to show which saves are more "selfish saves", but what I was suggesting was along the lines of what I would like to see.
I've been thinking a bit more about LeeTro's comment earlier.
A stat does whatever it's modeled to do, and in my case, what I was trying to model is whether the Saves contributed by a relief pitcher help his team be a Winning ball club.
Given that this is what the stat was supposed to do, maybe it actually is the case that a relief pitcher contributing Saves to making a team 82-80 is strongly influential to the team's Winning season, and moreso than contributing Saves to make the team 81-81 (that is, non-Winning).
I'm more concerned about the way the numbers radically drop off as the denominator (Games over .500) increases. If we take K-Rod's 62 Saves in 2008 as an example...we really shouldn't see this happen as we marginally increase the Games over .500: 1 gm, +62; 2 gms, +31; 3 gms, +20.67; 4 gms, +15.5. That's just too steep of a drop off over just a couple of games.
All that being said, I still like the way this stat models what I was interested in. The numbers are too high and drop off too steeply, but they capture my sense of the Save artist's real contribution.
Would there be a way to level off this steep drop without making the stat too complex for a fan to use (and thus rendering it somewhat besides the point)?
Actually, "the futility of contribution" I'm trying to model on the "negative" side really should run in the other direction, now that I look at it. 62 saves for a team that loses a lot should register as "more futile" than 62 saves for a team that almost, but not quite, managed to get over the .500 hump. So that would be a fair criticism of what I was trying to do.
(I realize it's the tendency of a lot of people on this site to look at a stat that doesn't work, comment that it doesn't work, and then wash their hands of the person who was trying to design the stat...but I still think there's something with this stat that might work if I add the right thing to it, I'm just not sure what it is yet. The topic may or may not interest you, but if you have any ideas, certainly let me know.)
Perhaps instead of Games Over .500, I should use Games Over Worst in the League? Then the stat would be uniformly positive, and would only be zero for a player on a team which is in fact the worst in the league. And that would be proper, because your contribution to team Winning is properly considered 0 if you can't help the team be better than the worst in the league.
Granted, this makes the stat about "value over other team", which is problematic if, say, you have a competitive league where all the teams are more or less good. But otherwise this really avoids the problems the stat's design has encountered so far, I think.
Another possibility might be Games over .250 - that way you avoid problems associated with competitive leagues. (The reason for the choice is that this is what the '62 Mets got...but teams in baseball's pre-history did worse, so maybe the bar should be set lower. But in the modern era, that's the worst, so it's a pretty reasonable bar level.)
Let's look at Rivera vs. Nathan vs. Heath Bell in this argument. All 3 have had a similarly great statistical season. Rivera plays on the best team in baseball, Nathan is on an average team, and Bell is on a below average team. WPA for each pitcher: Nathan 3.05, Rivera 3.62, Bell 1.55. This shows Rivera only having 2 blown saves this year. I really don't know which pitcher is most valuable, but I do think teams just above .500 (aka Twins) do get the most marginal value for those wins a player would provide.
As for your scale, you could just invert the ratio, with 0 being most valuable.
Games over worst team is a better idea for the lower part of the spectrum. I don't know how you can find a way to increase value for playoff-bubble teams. If you do something like "Wins away from 88" or somewhere in that range, you can find the most value in players on teams that can make a difference in whether that team made the playoffs. You would not invert the ratio in this case.
If we try using Games over Worst Team (we'll say in all of MLB) as a basis for the stat...thus making it so the lowest possible value for the stat is 0 (save artist is pitching for the worst team in baseball and not helping it be anything but the worst team)...here's what happens...
K-Rod in 2008: 62 saves / (100 wins for Angels - 59 wins for Nationals as the worst team that year) --> 62 / 41 --> 1.51
K-Rod in 2009 (not including today's game): 35 saves / (69 wins for Mets - 58 wins for Nationals as the worst team this year) --> 35 / 11 --> 3.18
Just re-thought what I originally posted. Sorry if people read my earlier misinterpretation...I changed the preceding post.
The higher this stat is, the more a player's save would be "selfish" (a save that doesn't help the team win). So, yes, this does represent what I was trying to represent. The higher value for K-Rod on the Mets is correct. Getting 35 saves on a team only 11 games better than the worst in MLB is a good deal more "selfish".
Another possible representation might be to measure the "Selfish Saves" directly as Saves - Games Won over Worst Team. K-Rod in 2008 would be 62 - 41 = 21 selfish saves, whereas K-Rod in 2009 so far would be 35 - 11 = 24 selfish saves. I'm thinking that might be preferable because it avoids the problem with the gigantic marginal changes when the denominator is low...though that is not as much of an issue when Games over Worst Team is used in the denominator.
I'm not sure that this stat really measures all that much, except a combination of the ratio of close wins to blowout wins with the overall success of the team. The closer isn't the one making the call these days. Almost every manager uses the closer by the rulebook--for save situations in the ninth inning.
You can pretty much WPA and LI for an idea of whether the closer was needed. That gets an interesting closer evaluation stat into my head, though. Basically, you can use the LI of the closer's save opportunities to determine the probable save success rate of a replacement level closer (i.e. lots of 3-run opps means higher RepSv%, lots of 1-run opps means lower). RepSv% would be the expected Sv% of a replacement level relief pitcher. You could use it to calculate SORP, saves over replacement player. Sample size would probably doom the stat, but it would be a way of using saves as a more meaningful stat through contextualization.
Anyway, that's all I'm thinking. Doesn't add much to the point. The closer is getting paid for saves, but, though the closer wants to be "the closer," that's because every team (essentially, though maybe not TB) wants to have a "closer," which they'll use as such. Starters want to be the "ace" of the staff, but there's no special stat QSA (quality starts as the designated "ace") that they accumulate (thought they do accumulate stats).
