Production
Posted: September 8th, 2009 | Filed under: Baseball | 80 Comments »
Well, we have brilliant readers … and we have BRILLIANT READERS. And so Rob V gets our all-caps Brilliant Reader award for the week for the idea of calling a baseball stat “Production.” That’s it. That’s the whole name of the stat. Production. As in:
“Wow, look at Albert Pujols’ production.”
or
“You cannot argue with the production you get from Joe Mauer.”
And so on.
I know people have been arguing and arguing about the Hitting Average stat, but I must admit that I have not read it as closely as usual because I’m in Milwaukee for Sports Illustrated*, where I have been working and getting yelled at by old guys in the press box. Well, actually, it was just one old guy. He was seated two or three seats over, and I was talking a little ball with Keith Law, and I guess maybe we were talking a bit loud. Maybe. I doubt it. But maybe. My voice can get loud at times …
Anyway, the guy glares at us for two innings (according to Keith … I didn’t notice) and finally shouts, “Hey, we’re not talking too loud over here are we?” I instinctively said something like, “Hey, sorry, we’ll keep it down,” because I’m an apologizer. People will ask sometimes “Where do you live?” And I say Kansas City, but I should say that I live in the Mode of Contrition. But of course after apologizing I soon realized that this guy — no idea who he was, maybe some old player from the 50s or something — actually never shut up himself. He talked for the next five innings straight. Of course, I’m just not the type to snap out: “Uh, actually, yeah, you are now talking to loud over there” — we residents of MOC don’t do that. But I did notice that the stuff he was saying was so inane (“This guy Pujols is a good hitter!”) that I realized he probably more annoyed by WHAT we were saying rather than HOW LOUD we were saying it. Fair to say the Milwaukee King would not appreciate a new baseball stat.
*What could the guy be doing in Milwaukee? A story on the great German restaurant Karl Ratzch’s? A retrospective on kicker Tom Dempsey? Ah, the tension builds.
ANYWAY, I love this idea of a stat we would call “Production.” And I think it should be a stat that simply totals up a players’ production — not a percentage, not an average, not a fraction at all. Just total up everything.
I’m open to how we can do it, but I’ll suggest that maybe the best way to do it is just take the positives column from the “Hitting Average” stat and use that. (We could also use runs created … like I say, I’m open).
In case you forgot, it gives you a point for a point for a single, 1.6 points for a double, 2.2 points for a triple, 3 points for a homer and some various fractions for stolen bases, walks, intentional walks and so on. The system was devised by the brilliant Tom Tango (inspired, no doubt, by the linear weights work of Pete Palmer among others), and Tom almost certainly has spent a whole lot more time than you have thinking about it. He came up with these proportions because he figured out that they correspond very closely to runs scored. I ran a little spreadsheet that figured out the “Production” of every single team this decade and then compared that to the number of runs scored by each of those teams (using the “CORRELATE” feature) and they correlation was .95 — better than OPS, or OBP, or SLG and MUCH BETTER than batting average.
You ask: Who cares? Why do we need another statistic? Well, of course, we don’t need another statistic. We don’t need the statistics we have now. But I like statistics because I like trying to get a little bit better insight into baseball. Get a little bit closer to what really drives the game. I think one of the reasons some people don’t like advanced statistics is because they tend to pop the pretty myths we have in our minds. I like myths too, I really do. But I also like to know that they ARE myths.
For instance, one constant theme that people never stop arguing about is this idea of clutch hitting. I understand. We want to believe in clutch hitters, those people who have something a little bit more sturdy in their souls, those people who can raise their game to the moment.
I’m not saying clutch hitting doesn’t exist — of course it exists. Ninth inning, tie score, man on second, and the batter rifles a single up the middle — that’s a clutch hit.
But we often look for clutch hitting in all the wrong places. Did the guy at the plate knock in that run because he has some invisible talent for getting a hit in the big situation? Did it come down to his moral fiber and his inner strength? Or did he get the hit because he’s a good baseball player?
For instance, one BR was comparing the stats of Mark Teixeira and Adam Dunn. They are somewhat similar (Dunn with a career 133 OPS+, Tex with a 135 OPS+). But, the BR was suggesting, there is something resolute about Tex and something flimsy about Dunn. Look, for his career, Tex hits .317/.438/.597 with runner in scoring position while Dunn hits 231/.421/.486. See? Who would you rather have come to the plate with runners in scoring position?
Well, for one, I think Tex is a better hitter than Dunn. Sure, I am a big proponent of walks, but I have never suggested that a walk is as good as a hit … it often isn’t. Dunn’s offensive game is pretty much homers and walks and he strikes out a ton while Tex has much more variety to his game.
Still, if you want to talk Tex and Dunn, that’s fine. Look at these two lines of numbers:
Player 1: .281/.461/.587
Player 2: .273/.414/.473
OK … who is the better hitter? You’re no doubt way ahead of me: Those are the batting lines for two players with runners in scoring position THIS YEAR. Player 1 is Adam Dunn. Player 2 is Mark Teixeira.
Well, these are the games you can play with small sample sizes.
Anyway, I would just like to find a quick baseball number that we could look at — a number more meaningful and more connected to a players value than RBIs or runs or batting average or even OPS — and legitimately say: “Wow, that guy’s having a great, great season.” No, there is no shortage of such numbers — WPA and Eqa and runs created and Win Shares and VORP and a bunch of others. But for whatever reason, I wish there was something else.
So … production. I like it. Come up with the production stat you like best. And we’ll vote. For now, I’m going with the Tango Production Model …
Here, then, are the all-time Production Seasons
- Babe Ruth, 1921, 468
- Barry Bonds, 2001, 457
- Babe Ruth, 1923, 441
- Lou Gehrig, 1927, 441*
- Jimmie Foxx, 1932, 436
- Babe Ruth, 1927, 433
- Lou Gehrig, 1936, 427
- Sammy Sosa, 2001, 427
- Hack Wilson, 1930, 424
- Mark McGwire, 1998, 423
*Did you know that in 1927, Gehrig struck out a career high 84 times? And apparently this bothered him so much he basically cut way down on strikeouts for the next 10 years — the most amazing year was 1934 when he won the triple crown, hit 49 homers, drove in 165 RBIs, punched up a 208 OPS+ in 690 plate appearances … and struck out just 31 times.
And here are the 2009 Production leaders
- Albert Pujols, 348
- Prince Fielder, 311
- Mark Teixeira, 302
- Ryan Braun, 292
- Chase Utley, 291
- Adam Dunn, 291
- Ryan Howard, 291
- Adrian Gonzalez, 290
- Hanley Ramirez, 286
- Mark Reynolds, 284
- Miggy Cabrera, 284
Jeter is at 283. Mauer, because he missed some time, is at 266 as is Justin Morneau. Not saying you want to build your whole baseball experience around it … but it just seems like a fun thing to know.
Loading ...
Jeez, I am first? No way you are circling me…
Should we divide these by ABs or PAs or games to get rate stats, to account for missed time due to injury? Or do you just want to leave it as a counting stat and make the argument that it measures consistency as well (being able to remain in the lineup, etc)
Either way I love this way of looking at it. I’m going to see if the ESPN program can accept decimals for fantasy scoring purposes next year and abandon the traditional H, R, RBI nonsense.
Joe,
Forgive me, I haven’t had time to read all the comments about your last post. So I am not sure if I am repeating something here but you do realize their is a great stat called Weighted On Base Average? wOBA is probably the best offensive rate stat in the game, I believe it was created by Tango but I am not certain on that.
I know it is fun to come up with new stats, but I don’t see any improvement on wOBA here . In addition, I think your Production stat would pale in comparison to WAR over at fangraphs.com. It accounts for offensive and defensive runs above average, along with replacement adjustments.
Could this be inverted and divided by innings pitch to allow one to evaluate pitchers?
“In case you forgot, it gives you a point for a point for a single, 1.6 points for a double, 2.2 points for a triple, 3 points for a homer and some various fractions for stolen bases, walks, intentional walks and so on.”
Perhaps it was first printed on that sign on Jimmie Dimmick’s lawn?
Any chance we can see the entire formula?
It seems like so much baseball stattery is aimed at divorcing performance from the situation.
Meanwhile, I watch a huge amount of tennis, where ‘playing the score’ and ‘mental toughness’ are integral to success; these terms are essentially equivalent to ‘clutch hitting’.
While the rhythms of baseball are far different, I would love to see a stats-person with skill (and access to the data) consider a stat based entirely on situational performance.
It would be a closer look at players’ performances based on who’s on base, the score and inning, with carefully considered subjective weight applied to each situation.
Like — a solo HR in the 9th while down by 11 runs will have little weight, while a game-winning, walk-off single would get some weight (though less than a HR).
I don’t know if it would show situational stats as more ore less meaningful, but I’d love to have the tools to look.
I like the Production+ number (HA+) from a couple of posts ago better. Take away from production for getting caught stealing and all that.
It’s not really fair to compare Barry Bonds to Babe Ruth in this stat since you get less credit for an intentional walk. We don’t really know how many of Ruth’s walks were unintentional so you count them all as .7. Perhaps if we’d have real stats on intentional walks from the 20s and 30s we’d have less people from that era in the top 10.
If I knew you were in Milwaukee (with Keith Law, no less) I would have come to the game and stared at you for the whole thing
i suppose it’s useful when thinking of an mvp/injuries as well. you see a .300 batting and have no idea of knowing if it’s over a week or a season, but the 266 shows the injury.
id still vote mauer.
and i believe in clutch.
I think you need to take a break from trying to come up with new stats and start figuring out a better means for the ones that we already have!!
I’m sick and tired of watching SC and they come up with these obscure stats just for the sake of stats.
And I love the choices for AL Cy Young.
Is this normalized for the era’s run-scoring environment? Is it? If it isn’t then OMG you’re just as bad as the rest of them!!
PS. What’s it like to “watch” a “game”? I never have. =(
I like that, by simply adding up the weighted values for each event, you get a number that depends heavily on the number of PAs a player gets during the course of a season. Place in the lineup doesn’t seem to be a major issue, since leadoff hitters don’t usually rack up a lot of XBHs. What you end up with, then, is a measure of production that takes durability (games played) into account. Of course, in doing so, you privilege players who play on good offensive teams, since those players will naturally have more PAs. It’s an interesting sta(r)t, but the simplicity of it creates some problems.
Joe, i just can’t believe you published a stat that currently has Mark Teixeira as the top-rated player in the AL. Now those aboard the Tex bandwagon can say “look even Mauer’s biggest supporter rates Teixeira more highly!”
@mwu
“I would love to see a stats-person with skill (and access to the data) consider a stat based entirely on situational performance.”
You are looking for Win Probability Added (WPA). See the link in my name. It basically looks at each at-bat and calculates a team’s chances of winning the game before and after the at-bat. The change in the win probability is assigned to the batter, and then you just total up all his at bats. So, like you said, a HR in the 9th of a blowout is basically worthless, while a game-winning RBI single in the bottom of the 9th is huge.
@Wooden U. Lykteneau
“Any chance we can see the entire formula?”
It was in a post from earlier this week. I’m not going to bother looking it up, since you can do that as easily as me.
In response to Matt at #3…I think it’s weird that you’re saying we don’t need another stat because we already have wOBA, when the reality is that we don’t need wOBA because we already have EqA. EqA’s been around since the early 90s – I remember reading Clay Davenport’s translations on the rec.sport.baseball newsgroup back then. EqA also has the advantage of being scaled to batting average, which makes it a lot more accessible to the casual fan than wOBA, which is scaled to OBA. (Everyone would quickly learn that a .300 EqA is good; not everyone would quickly learn that a .300 wOBA is bad.) Like you, I’m reluctant to embrace a new stat when there’s an existing stat that works just as well. Which is why I see no reason for this sudden new embrace of wOBA when we’ve been using EqA for over 15 years.
If wOBA correlates with run scoring better than EqA does, that would be a point in its favor. I have not seen any evidence that it does.
Had a similar thought to Matt’s at #14. After this blog spent weeks saying that Mauer is the clear-cut MVP – backed up by many BRs who wrote that anyone who failed to see this was an idiot or worse – the Official Blog Stat now says that Teixeira is the most productive player in the AL to date. Holy cannoli.
why, why, why, no AL MVP candidate from the west? we exist out here. we really do. and even though it pains me to say it, as I am a loyal Mariner fan (sorry about the whole Yuni thing, but no, we don’t want him back) , the Angels are kicking everyone’s a** and the bat in their lineup that can’t be replaced, no matter how unforeseen this was, is Kendry Morales. Take Texeira out of the lineup and the yankees are still the yankees. Take Jeter out and the yankees are still the yankees. (Maybe not winning at the current ridiculous rate, but still winning). but take out morales… bigger dent.
his OBP is .357 to Tex’s .380
his SLG is .588 to Tex’s .551
his Avg is .313 to Tex’s .282
he’s got 30 HR to Tex’s 35 HR
95 RBI to Tex’s 106 RBI
2 triples to Tex’s 0 triples
39 doubles to Tex’s 35 doubles
151 hits to tex’s 149 hits.
So, yeah, I’m with you on the Mauer thing, and Jeter is having his best season yet, but comparing 1st baseman to 1st baseman, on a playoff-bound team, Morales means more, does more for his team than Tex. Fight the east coast mentality — pump Mauer over the next obvious candidate, Kendry Morales. I still love Mauer, but Kendry should at least make the poll. All I’m saying’.
Also, loved the AL Cy Young poll. Love me some Felix, but Zack is the man.
DavidH (15), thanks for providing that link to WPA. I’m linking to a fangraphs ranking of this season’s WPA leaders for anyone who wants to check it out.
In your definition you make a few references to at-bats but you mean plate appearances, right?
Wouldn’t “PozPro” be at once folksier and also more au courant?
Teixeira. (Always forget that first “i”, even though I grew up with a whole pack of Teixeiras and should know better).
and just to be clear — the AL MVP is obviously Mauer.
I don’t like the counting nature of the stat. I have no idea how to compare Albert Pujols’ current season to Babe Ruth’s season until the end and both are complete. This is another reason that rate stats are popular.
I voted for Zack Grienke in the poll. Somehow, he’s only third! Who are the people voting for other “Zack Grienke”s? What are they looking at?!??!1!
I forgot what the stat was called but I remember seeing one that decided what the odds were of a team winning the game at each point in the game. If you take the odds of winning after a play subtract it from the odds before the play and assigning this value to the hitter and pitcher (or baserunner or fielder) then you have a leveraged value that tells you the value of every contribution based on the game situation. This values an out in a one run game by the closer more than an out in a mop up game. I would like to see these type of stats. You could even break it down by a hit into this zone at this trajectory and velocity turns into a single this often and break it’s value down for the fielder, hitter, pitcher, and even the baserunner who will advance two bases this fraction of the time. Anyone know what this stat is/was called?
How far down the “production” list must we go to find a 2009 Pirate (and still a Pirate)?
Perhaps that could be considered the new Mendoza line? We should probably give it another name (although Mendoza was originally a Pirate) but I don’t think that there is a current Pirate worthy of said line so maybe let’s get the Royals involved as well and go with the “Yuni line”.
“This “Yuni line” is not to be confused with the bus, subway or trolley line that connects a population base to a univerity campus.”
Joe,
You will probably tell me this stat already exists, but here goes.
Why doesn’t every play a batter makes have a point value. Basically a single is worth .XXX runs. That is because when a player gets a single the probability of a run scoring goes up by that amount. HBP and BB are probably the same, but less than a single. A bunt might actually have a negative score in some (or most) cases. An error would have positive score, and I know that an error is technically beyond a hitters control. However is it possible that certain hitters hit the ball harder more consistently therefore causing them to be on the high end of the error scale? What about the speedsters that make the fielders rush the play or the throw, might they get on by error more often? If a hitter makes an out that is worth -.XXX points but it is not as bad as a K because runs can score and runners can advance on groundouts so a K would have a higher negative point total, the only thing worse being a DP. It just seems to me that every play a player is involved in affects the probability of a score positively (getting on base) or negatively (making an out). It may be that the way your stat is already calculated already reflects what I am talking about.
I am no statistician, but it seems to me that this stat should have a super high correlation to runs scored. I know this might not be a simple stat, but the values could be posted somewhere since they would be a calculated constant, and people could access them to add up the totals. Every player based on his performance could have a Run Score for every game like a pitcher has now with Bill James’ Game Score.
Nice work as always Joe – might want to throw in a link to Tom’s blog for those who don’t go looking for such things. Some of the best reading out there.
Okay, while I was writing my question someone mentioned WPA, that was basically what I was talking about. I think this is a great stat for measuring actual production/value. Which is probably what MVP should be based on rather than some more predictive number that will tell you you had a lot of line drives but fewer hits than expected or you had an unusually low BABIP and should expect to return to the norm next year. Good predictive information but says little about value in the here and now.
Tango responds here:
http://www.insidethebook.com/ee/index.php/site/article/pozs_production/
As far as the correlation – the problem is that it doesn’t really matter how you account for outs at the team level, since all teams have essentially the same number of outs (27 per game, more or less). This is NOT the case when you break it down to the player level. So the stat may correlate well for teams – but it should not be used to evaluate individual players. (That’s where the Mauer/Tex problem comes in.)
Tango’s proposed method may resolve that issue.
Joe in Jersey – you’re basically describing linear weights.
@25 Marty,
I like your idea for that stat. Fangraphs already tracks the probability of winning throughout a game. I don’t know if your stat already exists, but it very clearly adds and subtracts based on the batter’s outcome. Also you are right that in close games (what one might call clutch situations) these results would be bigger positives or negatives. Very interesting.
Wow. The shiny new production stat actually implies that Tex really IS the AL MVP. Poz seems like a pretty fair guy, the kind who admits when he’s wrong; why didn’t he mention this?
@31 Colin,
Thanks. Where’s the best place to see this? Why isn’t this used more frequently? is the name?
I’ve always wondered who would have won the 1927 AL MVP award if the voters had been allowed to vote for the Babe. (Gehrig won the award that year; writers were not allowed to vote for a player who had already won the award).
Knowing baseball voters, Ruth’s record-breaking 60 homers would have pushed him over the top, but as the Production stat shows, Lou probably had a slightly better season (although the difference is razor-thin).
Joe – Fangraphs has wRAA and wRC, which are both linear weights (baselined to average and zero, respectively), as well as wOBA, which is essentially a rate version of linear weights. Baseball Reference has Batting Runs, or linear weights relative to average. BPro has Equivelent Average, which might as well be linear weights. They’re really pretty commonly used.
I would recommend the Fangraphs values – wOBA typically grades the highest of all publicly available linear weights implementations when I do accuracy tests.
@31 Colin,
I just looked at linear weights on Tango’s site and he doesn’t account for every play possibility. Also a CS is the point opposite of SB, when it has been shown that an out made cost more in run probability than the extra base adds (That’s why many agree that you need to be successful more than 2/3 of the time to add value). So linear weights good, but should be more accurate.
Wow, the 1980’s sure was a decade of mediocre players. Any BRs able to name all of these?
http://www.sporcle.com/games/hrleaders_1980s.php
@36 Colin
Thanks again, will look into those.
Did anybody notice how much the National League dominated the production numbers for 2009 above? Is there some other factor at work here? It seems very strange that only one of the top 10 and 2 of the top 11 are in the AL.
Wouldn’t this stat be better if you divided the positive production by the negative production and then multiplied by the number of plate appearances? It would still act as a counting stat, but the negative outcomes a player amasses would be then be incorporated, not just the positive ones.
Joe – There are many different linear weights implementations, depending on the use case. (Tango’s site probably has a half dozen.) From the sounds of it, you’re probably looking at Linear Weights Ratio, which was what JoePos called Hitting Average. The value of the CS versus the SB was explained in the comments there.
David – Yeah, there’s something going on there. NL players tend to play in hitters parks, relative to the AL. (That’s not all of it, of course.) If you do a proper league adjustment, the AL players will rise in the rankings.
@ Steve G,
Steve, EqA, while a great stat, doesn’t use linear weights. It equally weighs all Total Bases arrived at from hits (much like Slugging %). So two single would be worth the same as a double, which is not the case. At least in my understanding.
EqA is nice that is is scaled to Batting Average, making it more easily understood than wOBA in it’s final form… but that doesn’t make up for it’s flaws.
I agree w/ Matt @ 14. While it is accurate, and more substantial, it is essentially just a counting stat, so you will ineviatably leave out the brilliance of players who don’t get the same number of tries.
I’m sure sombody else has already thought of it (and probably done it eariler in the comments, or in the comments of the last post), but can’t you just take the number and divide it by PA’s??
That seems much more accurate, plus it seems to make the line from good to great right @ .500, and that’s a nice, round, easy to understand number.
2009 Production Average Leaders:
1.) Albert Pujols .584
2.) Joe Mauer .538
3.) Hanley Ramirez .514
4.) Prince Fielder .513
5.) Adam Dunn .506
6.) Miguel Cabrera .503
7.) Mark Reynolds .502
8.) Chase Utley .501
9.) Adrian Gonzalez .500
10.) Ryan Howard .494
11.) Ryan Braun .492
12.) Mark Texerishexxera .490
??.) Captain Yankee-Pants .454
matt/43: I think you’re being unfair to EqR here. I certainly think wOBA is better, but it’s a matter of degrees. The basic formula for EqR is:
(H + TB + 1.5*(BB + HBP) + SB) / (AB + BB + HBP + SH + SF + CS + SB)
(I’m leaving out a lot of the surplus math to go from that rather OPS-like unit into absolute runs and then to go from there into the familiar EqA.)
So the relative weights of everything looks something like:
1.5 = Walk
2 = Single
3 = Double
4 = Triple
5 = HR
Or, relative to the single as 1 (like LWR/”Hitting Average”):
.75 = Walk
1 = Single
1.5 = Double
2 = Triple
2.5 = HR
Which is a lot better than you’re making it out to be. Again – you get a bit extra precision out of doing it “right”, but not a lot.
I think if you’re going to have a stat like this you should also take away “points” for making outs. Something like: Regular out:-1 Sacrifice hit: -0.5, hitting into a double play of any kind -1.5.
That would solve the problem of guys with the most plate appearances having a huge advantage.
Also, if it really is a value stat aimed at the MVP some defensive positional bonus or penalty should also probably be included (catchers and shortstops should be given a bonus compared to DHes and first basemen).
The thing that appeals to me about this hypothetical stat is the ability to value careers, especially as pertains to “peak” vs. “longevity” guys for the HOF.
One would hope that this stat would establish an in/not-in line for hall of famers.
Joe,
While in Milwaukee, you have to try Real Chilli especially the Marquette store at about 16th & Wells. It is a little better when you have been drinking but its still pretty good when sober.
Love the stat.
“Wow. The shiny new production stat actually implies that Tex really IS the AL MVP.”
Until you factor for Defense and production on a Value vs Positional Average basis.
I am fairly confident about going on the record as saying a “302 Production is much easier to obtain then a 266 Production behind the plate.”
In fact, I might even go as far as saying 302 at 1B is half as valuable as 266 at Catcher – but that would be a wild stab in the dark, so I will not commit to that rate
@ 47: Marco
“The thing that appeals to me about this hypothetical stat is the ability to value careers, especially as pertains to “peak” vs. “longevity” guys for the HOF.
One would hope that this stat would establish an in/not-in line for hall of famers.”
Again, it doesnt include defense or defense vs positional average, so comparing say Ozzie Smith to Reggie Jackson solely on “production” will not give you a very good comparison.
“He came up with these proportions because he figured out that they correspond very closely to runs scored.”
…any mathematician should be afraid of this stat because of that description. See, it may look like it works right now, but if there’s no real math reason for assigning those weights, then it’s just random numbers and that just skews the whole formula’s relation to reality. Of course, it could happen to be perfect, but… I’d still need to understand why those weights are reasonable before putting any stock in this stat.
OK – i haven’t read all the comments, but in case it hasn’t been brought up.
It sure seems like most of those players are NL players, and we have been conditioned to believe that the AL is the better league, so shouldn’t the majority of players on this list be AL players?
Does Production need to be put through some kind of league/stadium context equation or something?
Just saying . . .
“It sure seems like most of those players are NL players, and we have been conditioned to believe that the AL is the better league, so shouldn’t the majority of players on this list be AL players?”
Just because the best is in the NL does not mean the NL is better – it means the NL happens to have the best players and nothing more. When you do this for every single player, your results will indicate that the average NL score is lower then the average AL score if the AL is really the better of the two leagues.
Or think of it this way. Say I am picking apples and placing them into two barrels. These are my barrels:
a) In this barrel I put all the good to great looking apples. – it has 25 good to great apples in the end.
b) In this one I put all the worm-eaten and rotten apples, as well as three pristine apples at the top. – it too has 25 apples at the end, 22 horrendous ones and 3 which are pristine.
which barrel do you want? Which would you consider the “better” barrel? Easy choice, isnt it?
As it relates to the MVP debate, I like that Production is a counting stat. MVP should be based mostly on counting stats, as they reflect what a player actually contributed to his team’s 162 games.
Counting stats don’t necessarily have to be caveman stats, as in the case of WPA, which also affirms Tex as the league’s most valuable offensive player.
Mauer certainly deserves an upgrade for playing catcher, but don’t you also have to adjust for Tex playing roughly 50 percent more innings of defense than Mauer? I think you do but how to make that adjustment is up to people smarter than I.
Devon: read the wOBA article, 2nd to last
http://www.insidethebook.com/articles.shtml
This stat should not be read to say that NL hitters are better than AL hitters. It only says that they produce better stats. These stats are highly context dependent. I say most importantly the NL batters accumulate the stats against inferior pitchers and fielders. There is, of course also the context of the home stadium. It is possible that the best hitters are in the NL and that they are just distributed more evenly in the AL and the average player is better, but I doubt it.
@52 Briggs: Adding to JoeyO’s point (53) – Weaker talent pools tend to have wider variances in performance. Think about it this way: take any of the best five hitters in baseball. For the sake of argument let’s say that is Pujols, Utley, Mauer, Hanley, Tex. Would you expect each of them to have better numbers in the stronger or weaker league? All other things equal, a single player will have better numbers in the weaker league. The NL may just be more top-heavy, or some the best NL players may have even better numbers than their AL counterparts due to the weaker competition.
I understand that this appears that Poz just found a way to say that Tex is the MVP, but everyone is overlooking that Poz also mentions, “Mauer, because he missed some time, is at 266…” One could assume that Poz is hinting that if Mauer had played a full season, his Production would be higher and possibly even better than Tex’s, who has played a full season.
i think bellwether was onto something in post #44. dividing by PA allows you to take away the advantage of playing for a higher socring team and gaining more PA for your counting stats. if you took that list (from #44) and then multipled each player’s score by the number of games he played, you would return the (deserved) advantage of players who’ve been able to play in more games. the stat still wouldn’t take park effects into account, and still would have nothing to do with defense, but i think it would give a closer sense of what joe (i think) intended for it to display.
i also agree with those who have argued this stat does NOT claim that teixeira should beat mauer for MVP, because it makes no attempt to bring defense into the mix. if you were to take “production” and eliminate the advantage of playing for a higher scoring team while keeping the advantage of playing in more games, teixeira comes out a bit under 9% of mauer. i am sure most of us would agree that playing catcher makes up far more than 9% over playing firstbase.
I’m not saying Yuni is a good player……….he is clearly NOT one.
But he’s done O.K. for the past five or six games……
And I hate him..
So I’m not sure why I’m posting this.
Here.
Bye now.
Joe,
For my money (which, of course, is nothing — thanks for the free site!) I vote for RC/27, or Runs Created per 27 outs. I “love” the name Production — let’s just call RC/27 “Production.”
The only flaw with Tango’s numbers is that one cannot immediately look at them and recognize how great a year the hitter is having. You have written before that the problem is, partly, we can just LOOK at a .300 average and know that is good.
Well, RC/27 more than passes the “look-at-it-and-know” test. It spits out, essentially, the runs your team would score a game if your lineup consisted entirely that player.
For example, Pujols sits at 11 right now. Dunn? 8.9. Willie Taveras? 2.8 Betancourt? 2.3.
You look at the numbers, and they make sense. Also, you can INSTANTLY understand how good Pujols is and how bad Taveras is. . .
@ #3
Depends what you want. EqA is adjusted to park and era while wOBA is just adjusted to average OBP of the era.
IMHO I think EqA is still better for purely comparing Player A vs. Player B.
Runs Created is always fun to check, though. For anyone who’s a fan of needlessly dissecting and FJMing Hall of Fame candidates, check out the linear weighted RC’s of Tim Raines v. Andre Dawson. From what I recall, Raines’ lwRC/27 was about .75 more than Dawson.
Why is a HR in the 9th more valuable than the one in the first that gave your team the lead? *
Why is hitting a single to drive in a run in the 9th more valuable than the guy that got on base, stole 2nd, and then ran really fast to score on the routine single?
I don’t get the concept that somehow a run in the 9th is more difficult or valuable than a run in the first, third, or fifth innings. Isn’t it harder to get a hit off of most starters than most relievers?
My issue with percentage stats is that they discount counting too much. There is value in playing more games than less games. Of course, counting stats over value guys that play every game (whether they should or not), because they rarely have anything in their formula that penalizes outs or caught stealing….
*it’s like the MVP race. Somehow the stats in September are more important, even if some guy was hot for four months and got his team into first place existed. somehow, getting your team into first early in the season is less important than doing well late in the season. It makes no sense. How is it better to suck for 3 months, and then get hot and help your team come from behind, than it is to be good for 5 months while getting your team in first and keeping them there?
The problem with just adding up the “good stuff” in a stat like this is that it ignores the fact that guys who play more made a bunch more outs. Mauer missing a month is not contributing anything, one way or another, to his team during that time and Teixeira will end up with more hits and walks because he played more. But Teixera has also made more outs than Mauer and that is a negative.
“I don’t get the concept that somehow a run in the 9th is more difficult or valuable than a run in the first, third, or fifth innings. Isn’t it harder to get a hit off of most starters than most relievers?”
A one run lead in the 9th inning is a lot more likely to hold up than a one run lead in the first inning. It follows then that a home run in the 9th to take the lead changes the probability of winning a lot more than the same home run hit in the 1st inning.
Put more generally, as the clock runs down, the probable deviation of the current score and the final score gets smaller.
WPA itself is fundamentally sound, and it’s perfectly reasonable to say that it is giving you a measure of value. Less clear is whether or not that measure should be considered the true expression of value, or rather a current approximation of valye with a degree of uncertainty that changes as the game progresses.
Decisions in the game need to be made without information about the future, so WPA is a better expression of how an event affected the outcome of the game than a backwards looking view. Likewise, WPA ( and “leverage”, which is roughly how much WPA is possible right now) represent fairly well the drama of the moment. But there are some odd things that you need to be reconciled to (example: how “valuable” is a complete game in a 1-0 victory? Well, that depends on when the 1 run scores. huh?)
The alternative – that the value of an event is determined by the eventual conclusion, rather than the current circumstance, also has problems. Is the first run of a 10-9 win as valuable as the last run? Are those the same values you have in a 10-0 blowout, when the last 5 runs were scored against the weak part of the bullpen, etc.
As per usual, you look at the tradeoffs, and decide if they work for you; then learn to recognize when the person you are arguing with prefers a different set of tradeoffs.
I kind of like it being a counting stat too. I get the argument against it, but I think people should be rewarded for not missing a month of the season. Not to knock Mauer, who has been fantastic, but my one and only problem with him as MVP (and I’m not even saying he shouldn’t win it), is that he missed a whole month. A month he made zero contributions. And I’m sorry, that’s a big deal, especially if the team misses the playoffs by a few games. Everyone wants to close their eyes and stick their fingers in their ears and pretend that didn’t happen while they announce anyone who doesn’t think Mauer should win is an idiot who knows nothing, but missing a month does matter.
You forgot to add ass-terisks next to the names “Sosa” and “McGwire” in your table of all-time production leaders. Not Pozterisks, not asterisks, ass-terisks.
“Mauer missing a month is not contributing anything, one way or another, to his team during that time”
Yes, exactly, and that’s a BAD thing. If this award was “Who has the best OPS+” or who hit the best, then that’d be one thing. It’s not. It’s Most Valuable Player, and not contributing anything to your team for a month makes you less valuable. Now Mauer’s having an amazing season and you can absolutely make the case that he’s STILL the most valuable despite missing that month. I may agree with you, I’m waiting til the end of the season to really decide. But this whole attitude of “You’re an idiot for thinking Mauer isn’t the clear MVP” when he missed a month is ridiculous, in my opinion, ya know? It does matter that he missed a month in terms of his value to the Twins, which is supposed to be what the whole award is about, and his “OPS+” or slash stats doesn’t show that.
If they had an award for hitters that was the equivalent of Cy Young, Mauer would be my pick and it’s not even that close, just like Grienke to me is the clear choice for Cy Young, despite other guys having more wins (not his fault). But they don’t.
@69: Yes it should count against Mauer that he missed a month. It is a bad thing. it would have increased his value if he had been in the lineup all that time.
The point though, and the problem with Joe’s proposed production stat, is that it ignores making outs.
Making outs is also bad. it also makes a player less valuable. If the system took into account the extra outs that Teixera has made compared to Mauer it would be more accurate.
I don’t think it’s dumb to say that Mauer isn’t the MVP because of the time he’s missed. (I disagree but can see why an intelligent person might argue that.)
But to say that he isn’t even in the top 11 “producers” in baseball, like Joe’s proposed stat does, is kind of dumb.
OK, I’ll bite. If the proportions (1 point for a single, 1.6 for a double, 2.2 for a triple, etc) correspond closely with runs scored, why not take it a step further. Why not consider situation. You know, something like Dan Leavitt uses here:
http://baseballanalysts.com/archives/2006/07/empirical_analy_1.php
that shows a leadoff hit or walk is worth .877 runs, a leadoff double 1.147, a single that advances a runner to third with one out is worth 1.187 runs, etc. I’m sure I am overcomplicating this, but wouldn’t that be more “exact”?
@66 – good point, on the fact that there is less time for the outcome to change, but I’m not sure that means that run is more or less valuable in producing the outcome.
The issue with Joe’s stat, or any attempt to make a new “easy” stat, is that it won’t be as predictive as stat heads want. The issue with stat head stats is that they are complicated to understand.
As a fan, something like Joe’s stat (if we also subtract outs) is fine.
As an exec, trying to build a team, I’d need more.
@ 67: KyleLitke
“A month he made zero contributions. And I’m sorry, that’s a big deal, especially if the team misses the playoffs by a few games.”
Ok, valid. But then you have this issue. If two players are fairly close on value provided, but one provided that value over 100 games and the other over 80 games, which was more valuable? Well, the number of games played doesn’t matter in a counting stat – only the end results. If the second person did it in 80 games, he actually played much better giving his team a chance to win a much higher percent of those specific 80. The player with the same value spread over 100 games gave his team much less of an opportunity to win, he just spread his out over a longer amount of time giving him the same value to his team then the other guy.
Now, of course, in our specific situation we are talking two players without an equal value but still have the problem I outlined above where the Value Per Position scale needs to be considered. If you put it on this scale (since we have to have both positions represented)
Tex 302 + Average Catcher
J.M. 266 + Average 1B
Which of the two of those do you think will be higher?
So where it might seem like a big deal that Tex played more games, if his counting value over the games played based off his positional average or expectation is lower then that of Mauer, then Mauer was more valuable to his team on the season even though he missed that time.
72: Mike in Mn
“The issue with Joe’s stat, or any attempt to make a new “easy” stat, is that it won’t be as predictive as stat heads want. The issue with stat head stats is that they are complicated to understand.”
This is one of the all-time worse statements I have ever seen.
Stat heads would love for stats to be easy – they just generally are not.
And how do we know this? Two of the most important stats to a “stat head” with regard to pitchers are BB/9 and K/9. These stats are as easy to figure out as ERA, if not even easier since the rate used is specifically given in the name of the stat – where ERA implies ER/IP and does not give the “Per 9″ rate used.
74 Joey:
Yup, that was too short of a statement. Happens when I’m at work sometimes.
I was not stating (or not trying to state) that stat heads want things to be difficult, just that for ONE stat to be predictive, it will likely need to be more complicated than “fans” want.
I was also talking about a hitting stat, that is “comprehensive” like Joe is trying to create. Again, I should take more time with some of m posts, but I am “trying to work” here.
that said, you offer two excellent examples of simple stats that I think are predictive.
“Yup, that was too short of a statement. Happens when I’m at work sometimes.”
No sweet dude, happens to us all. And I wanted to actually offer my apologies for being snippy with you in my response. Have an idiot trying to argue % isnt x/100 in the other thread and think it coupled with a couple buzz phrases (“stat head”, “not easy”, “predictive”) must have created a red-flag effect.
Did want to point a couple things out though with regards to hitting stats – BB%, K%, BB/K, ISO, HR/FB (if you know FB), LD%, etc… They are all very simple stats. And the complicated stats you allude to are often complicated mainly because of a serious of individual easy parts being factored together to provide a singular statistic.
Hello Joe, I enjoy reading your ultimate stat articles and I wanted to mention one that you might not have heard of. It is the Baseball Point System by Steve Mann. He published several Fantasy baseball guides in the 1990’s. The formula is (2xhits) + (1.5xwalks) + (Total Basesx1) +(Stolen Bases x 1) – (2 x CS) – (.50 x At-Bats) equals POINTS. You then divide points by At-Bats plus Walks. To get Runs Produced you divide by 4.4. For example take Johnny Bench’s 1970 season. He had 177 hits , 54 walks, 355 total bases, 5 stolen bases , 2 caught stealing, and 605 At-Bats. He had 488.5 points and 659 appearances ( AB + Walks) so his rate is 488.5 divided by 659 is .741 points per appearance. He had 111 runs produced 488.5 divided by 4.4 is 111. If you take this and you apply it to the league so far it is 99.5 % close to actual runs scored. Try it yourself for any season since 1951 when caught stealing was available in both the American and National Leagues. Please let me know what you think of Steve Mann’s Formula? I would be glad to hear from you. Remember that RBI’s and Runs scored are not very good estimators of batting ability, because take Pete Rose for example. He bats leadoff and there are not many times where someone is on base for him to get an RBI. He will score a lot of Runs but he will be short in the RBI department. Those two stats are batting order dependent and team dependent. It matters how good your team is. If nobody gets on base for you then you can’t get an RBI unless you get a home run. Thanks Joe. I hope to hear from you. Gary.
Of course clutch hitting exists; the thing is that no one adequately defines what it really is. Sabremetricians (generally) say it doesn’t exist because they define it as someone lifting their game at key points in a game or that they actually get better in those times.
Of course that cannot be found because if a player did improve over the norm by very much they would soon be trying to find a way to do it all the time.
Let me lay out what I think we can all agree are a few facts:
1) Situations in sports create adrenaline
2) Adrenaline can seriously impact how you handle those situations.
I.E. if a rising heartbeat and dry mouth etc get in the way of your performing, then you will be said to “choke” while if you succeed you will be said to be “clutch.”
Simply put a clutch hitter is someone who gets past the interference of adrenaline and related aspects of tough situations and who can use the extra “rush” to
To finish my thought (embarrassing pause)
get past negatives associated with the “rush” and DO AS WELL IN THAT SITUATION AS THEY DO IN NORMAL SITUATIONS.
So why not define what those adrenaline inducing situations are and then see who maintains their performance? And after the 7th inning isn’t good enough – sometimes you have a 5-10 run lead and there is no adrenaline.
Heck sometimes, in the first inning of the first game of the season (blasphemy I know) can be more adrenaline inducing than the 8th inning of a game Greinke or Carpenter is dominating and has a 3 run lead.
I recall all too well the effect unnecessary adrenaline had on Jack Clark in 1985 when he dropped a foul popup shortly after arguing fruitlessly with Don Denkinger, for example
[...] it all? A statistic that everyone could understand and agree on? My new favorite sports writer, Joe Posnanski, feels the same way. I love this idea of a stat we would call “Production.” And I think it [...]