Second, there is no replacement level for defense, they are using league average for defense and baserunning. They use that standard because the theory is if you hit at replacement level, you have to perform at league average to be worth anything. A replacement level hitter wouldn't get playing time if he was also a replacement (which doesn't exist in war) level defender.

Replacement level isn't just about offense or defense; it's a combination. An average hitter who's unplayable on defense is roughly replacement level.

Replacement level isn't just about offense or defense; it's a combination. An average hitter who's unplayable on defense is roughly replacement level.

Again, I know why they chose to do that, but I feel it unfairly penalizes below average defenders who play everyday compared to above average defenders who only play 110 games. If Trout had 150+ games this wouldn't have been an issue, but it is an issue because it makes the assumption that a below average defender doesn't provide value by just being in the lineup everyday.

As far as I'm concerned, a guy who posts exactly replacement level offense, and slightly below average level defense, but who(for whatever reason) is in the lineup 150 games of the year, is a plus to the team. By war he wouldn't be, by war he would be considered inferior to the guy who went 1-5, but made several good defensive plays in his one and only game of the year. There is value to being able to play in the field everyday that war doesn't account for. (and I'm not sure how it would account for it to be honest.)

Again. Trout should have been the MVP, but I feel that War overstates it because of (several reasons actually) how it accounts for defense. I think Win Shares nailed the discrepency in their value pretty accurately.

Second, there is no replacement level for defense, they are using league average for defense and baserunning. They use that standard because the theory is if you hit at replacement level, you have to perform at league average to be worth anything.

Everyone is basically right here. Eric and JJ are correct that in WAR, technically, replacement is a combined offensive/defensive metric. There are no separate replacement levels for offense and defense. But it is also true in practice that the average replacement level player is far below average on offense and about average defensively. A player who provides average defense with very weak hitting is not delivering any value, as WAR correctly tabulates.

a guy who posts exactly replacement level offense, and slightly below average level defense, but who(for whatever reason) is in the lineup 150 games of the year, is a plus to the team.

No, this player has negative value. If replacement level is set correctly, better players are available at the league minimum salary. (In the short term, this might not be true for a specific team, but that's likely a managment failure.)

No, this player has negative value. If replacement level is set correctly, better players are available at the league minimum salary. (In the short term, this might not be true for a specific team, but that's likely a managment failure.)

And I don't agree with that. Of course my point is to illustrate the problem I have with using average defense, not really this particular player. There is value to the team in knowing who is going to play everyday at a position. There is value in a below average defender who plays farther left on the defensive spectrum than he should in order to help the team get a better player in the lineup. Any system that punishes a player who is a better defender than a butcher, but is worse than average, is not a system that I can fully support.

As I said, Trout should have been the MVP, but war in my opinion, overstates the difference.

Any system that punishes a player who is a better defender than a butcher, but is worse than average, is not a system that I can fully support.

Why?

A below average defender who can't hit at all is worthless. There are plenty of average to above average defenders who can't hit.

There is value to the team in knowing who is going to play everyday at a position.

There is no value in knowing you'll have a below replacement level player in there every day. There is no value in being able to pencil in Yuniesky Betancourt for 150 games of -1 WAR.

You're better off throwing a bunch of random minor leaguers and waiver wire pick ups at the position.

To claim it is wins, is ridiculous

Just curious (I have no interest in arguing over whether WAR is any good; I find all this stuff equally interesting):

Are the various WARs occasionally reconciled with team wins? I can see why WAR might not track team wins with absolute precision (after all, Pythag doesn't), but does it come close enough to make sense as roughly valid?

Last year (per B-Ref) the Rangers had 18.6 batters' WAR and 23.4 pitchers' WAR for a total of 42; they won 93 games, which would put replacement level at 51 (I know that's not how it's calculated; just working it backwards to take an initial rough look).

Oakland had 18.6 and 23 for, heck, round that up to 42; they won 94, (implicit) replacement would be 52.

The Angels had 37.9 and 2.6 (cripes!) for 41, rounded up; they won 89, replacement level 48.

Seattle: 14.2, 13.6, total 28, 75 wins and 47.

The minor discrepancies, as I said, are neither here nor there; looks to me like a replacement-level ballclub would win ~50 games. Sound right? is that an assumption in the system? The Astros had a total of 7.5 WAR as a team last year and won 55, which is in the range as well.

One implication of a pretty stable team baseline is that yes, WAR is apportioning actual wins, at least in practical terms. It may be apportioning them incorrectly, but players have to have some kind of value, and some have to contribute more than others, and WAR is a serious and reasonable attempt to do so.

But that's just a quibble. Replacement level for defense is just a parameter. One that's particularly hard to define objectively.

And I don't agree with that. Of course my point is to illustrate the problem I have with using average defense, not really this particular player. There is value to the team in knowing who is going to play everyday at a position. There is value in a below average defender who plays farther left on the defensive spectrum than he should in order to help the team get a better player in the lineup. Any system that punishes a player who is a better defender than a butcher, but is worse than average, is not a system that I can fully support.

This all means that you think replacement level is too high (or at least replacement level at premium defensive positions). It has nothing to do with the offense/defense distribution.

This is what people mean by arrogance. War is a structure that attempts to qualify the totality of a players performance in a wins like system. To claim it is wins, is ridiculous, the accuracy is not even close to being there, and outside of Win Shares, none of the systems even start with wins as the baseline.

Stop trying to lawyer the language and change the argument when you can't rebut it..

WAR is wins.Theoretical wins sure, but wins are always theoretical in how we, sportswriters, GMs and fans use them during analysis of players past or potential contributions. Even if the wins are estimates, and the accuracy isn't perfect, WAR is just the a framework built around the most eternal concept and question in baseball...

As I wrote ad nauseum in that post (and which you ignored so you could cherry pick) GMs and fans from the beginnings of baseball have tried to quantify the contributions of players as Wins. Virtually every baseball roster decision in history is based on the WAR framework, how many wins player A would add (theoretically obviously) over player B, or an easily available replacement.. Fans have argued forever over how many wins players contributed or will contribute, whether they do so in circumspect ways (he was the key to our championship, he's the missing peace this team needs) or directly.

As it stands, Miggy put up the better offense over the course of the 162 game season,

Wrong. Miggy 47 bRuns on offense, Trout 65 bRuns on offense. Offense includes base-running and double plays, unless you play some odd variant of baseball where what happens after the bat hits the ball is just "theoretical", LOL.

Offense is the component in in WAR that has the most accurate measure of value, and Miggy isn't particularly close.

And you don't believe in park adjustments? You think the best hitters of the 90s were all Colorado Rockies, who just had the misfortune to lose their batting eye when they were traded?

Even if you think park adjustments were wrong for Anaheim and Detroit, you have a long way to go to show Miggy as the better offensive player. The park adjustments would have to be nearly as large in the opposite direction.

Detroit's park factor for hitters in 2012 was 104, for the last three years? 104.

Anaheim's park factor for hitters was 91 in 2012. For last 3 years? 92.

LOL, good luck with that argument. And then you can quantify what WAR doesn't, the fact that Trout played in the west against tougher teams on average than Miggy did in the central.

Trout just gains ground by better defense, park adjustments, positional adjustments and the silliness of a system that thinks there is no benefit to a guy who puts on a uniform 20+ times over the other guy.

Again this is ludicrous. You found one minor adjustment of little consequence that you didn't like, and make ridiculous assertions like this. If Miggy played 20 less games, his WAR declines by a similar amount. If Trout isn't sent down for 20 games out of spring training, his WAR is most likely even higher (he hit .400 in AAA while awaiting callup).

Sure WAR is imprecise. Sure the fact it uses single season defensive measurements that don't have the sample sizes to get the accuracy (from a predictive standpoint) single season hitting samples give. So Mike Trout may not have been an all time great defensive center fielder. And Miggy may not have been as bad as he looks.

But Mike deserves the WAR that was measured, just like Maris deserved the 61 home runs he hit. Within the framework of the measurements we used, both excelled, whether either can ever duplicate their seasons again.

Stop trying to lawyer the language and change the argument when you can't rebut it..

Not trying. The point is that the arrogance to say something as moronic as "War is wins" is exactly what this article is talking about the arrogance put forth by the war disciples. They can't see anything other than their pet stat.

I like war, even with all it's flaws, but I don't adhere to it religiously, there are clear discrepencies in my opinion, the difference between Trout and Cabrera is a textbook example of the flaws with war and the strength of war.

Sure WAR is imprecise. Sure the fact it uses single season defensive measurements that don't have the sample sizes to get the accuracy (from a predictive standpoint) single season hitting samples give. So Mike Trout may not have been an all time great defensive center fielder. And Miggy may not have been as bad as he looks.

And I don't quibble with their rating of either players defense. My quibble is that it punishes everyday players when compared to a guy who played fewer games.

If Miggy played 20 less games, his WAR declines by a similar amount. If Trout isn't sent down for 20 games out of spring training, his WAR is most likely even higher (he hit .400 in AAA while awaiting callup).

Again, win shares does a better job of showing what is the real quality difference between the two seasons. I think War overstates it, because it doesn't give credit for just being in the lineup. It's not Trout's fault, but MVP isn't a blame game, it's based upon what happened in the past, regardless of the reason it happened.

And I have no problem with Mike getting the war that was measured(for the most part) my point is in comparing the two, Miggy is losing ground for being out on the field 20 more games, by the mere fact he was on the field. He is losing war defensively, every single extra game he plays, for being less than an average defender, not a butcher, but a less than average fielder.

I think War overstates it, because it doesn't give credit for just being in the lineup.

Yes it does; that's what Rrep is. It just doesn't give as much credit as you seem to want it to.

He is losing war defensively, every single extra game he plays, for being less than an average defender, not a butcher, but a less than average fielder

But isn't that how every manager has evaluated less-than-average defenders since the dawn of managers? You either tolerate the guy if his bat allows, move him if his bat allows, or, well, replace him.

Not trying. The point is that the arrogance to say something as moronic as "War is wins" is exactly what this article is talking about the arrogance put forth by the war disciples. They can't see anything other than their pet stat.

The reason KT is reacting so violently here is because the critique is so clearly and obviously true about him.

The reason KT is reacting so violently here is because the critique is so clearly and obviously true about him.

Well, clearly you are an expert when it comes to arrogance...

He is losing war defensively, every single extra game he plays, for being less than an average defender, not a butcher, but a less than average fielder

No, he is not losing "war defensively". He is losing defensive runs relative to average and gaining replacement runs. If he were an average hitter, he'd be gaining WAR.

I think this is what CFB is saying: If Cabrera had played 20 more games at DH instead of in the field, his WAR would be higher, despite the fact that he was doing less. He suffers for playing games in the field.

CFB, all you are doing is arguing with the notion of a replacement level player which means you're just saying b-r sets it too high. But the zero point isn't going to change the gap. Fangraphs has a lower replacement level than b-r and they put the gap at 3 wins (they put Trout's defense about 1 win lower than b-r).

But, even so, let's buy your argument. Trout started 138 games, Cabrera started 154. It's only a 16 game gap. Just how many WINS do you think 16 extra games of playing time can be worth? That's 10% of a season. Even if you put replacement level down at 0 wins that doesn't add up to 1 win for an average player. There's simply no way you can significantly reduce the gap between Trout and Cabrera with greater rewards for playing time. So if you think the gap was substantially smaller than 3-4 wins, you've got much bigger problems with WAR.

And look at the nature of your arguments -- "I feel this", "I think that", "I believe Win Shares has it about right" -- based on nothing to back it up. At least lay out the components of win shares and how they differ from bWAR and fWAR.

As to park factors, have you looked at the raw stats for Anaheim the last 3 years? In 2012, the Angels scored 71 more runs on the road and hit 23 more HR. Their opponents score 81 more runs and hit 28 more HR. That's a one run per game difference. 2010 and 11 are not as extreme but the gap is on the order of 50-60 runs and 15 HR. Those are big park effects, why should they be ignored?

And I always find this one kinda entertaining:

MT road: 951 OPS
MC road: 913 OPS

MT home: 976 OPS
MC home: 1094 OPS

One of these guys numbers benefited hugely from his home park and it wasn't Trout.

I think this is what CFB is saying: If Cabrera had played 20 more games at DH instead of in the field, his WAR would be higher, despite the fact that he was doing less. He suffers for playing games in the field.

If he is, then it's not right. Cabrera was something like -10 defensively on the season which is less than the positional adjustment between 3B and DH. I thought he was talking about a theoretical sub-replacement level player, though.

He is losing war defensively, every single extra game he plays, for being less than an average defender, not a butcher, but a less than average fielder.

Right. You seem to think that providing average defense, absent a decent bat, has value. This is demonstrably false. Think about it: every single starting SS in AAA is a much better fielder than probably 80% of all MLB infielders. Every starting CF in AAA is a better fielder than probably 2/3 of MLB OF. It just isn't a scarce comodity.

Or imagine baseball were like the NFL, with separate hitting and fielding squads. Hitting quality would be somewhat better. But fielding would be vastly better, an enormous upgrade. Probably 70% of all current starters would lose their jobs.

Right. You seem to think that providing average defense, absent a decent bat, has value. This is demonstrably false. Think about it: every single starting SS in AAA is a much better fielder than probably 80% of all MLB infielders. Every starting CF in AAA is a better fielder than probably 2/3 of MLB OF. It just isn't a scarce comodity.

Or imagine baseball were like the NFL, with separate hitting and fielding squads. Hitting quality would be somewhat better. But fielding would be vastly better, an enormous upgrade. Probably 70% of all current starters would lose their jobs.

Intriguing, but how do you demonstrate this empirically? And what would the impact of such an arrangement be on overall run scoring levels?

Those are big park effects, why should they be ignored?

Because they are systematically exaggerated in the way they have been computed since they were first invented? Because if we take a look at the variations in measurement of these values based on differing assumptions, they tend to strongly regress to the mean? And when we do that, we find that the difference between these two players (who we'll have to be hearing about until the end of recorded time, apparently...) is a good bit less than 3-4 wins?

Note we aren't proposing to "ignore" park factors, but when we look more closely at them, we find that they aren't quite the "mirror of truth" that many seem to believe they are. This is another instance of a good idea/theory running itself off the rails. There are pretty straightforward ways to adjust/correct for this situation; one day, perhaps, they will take hold.

He is losing war defensively, every single extra game he plays, for being less than an average defender, not a butcher, but a less than average fielder

Cardsfanboy, I think you are misunderstanding how the replacement baseline operates. Replacement in WAR is set as the overall level of offense plus defense possessed by freely available talent at a position. A replacement level 3B may be a -22 hitter and an average fielder, he could be a -18 hitter and a -4 fielder, or he could be a league average hitter and a -22 fielder -- the configuration of the value is irrelevant as long as the sum is the correct value of freely availabletalent. AROM's system defaults to displaying a certain configuration of talent -- average defense and bad hitting -- but he could just as easily shift the offensive baseline a little higher and the defensive baseline a little lower, and it wouldn't change a single result in the system. It's just a convention to use the average defense / bad hitter paradigm as a way of looking at the data, because that configuration of replacement level talent is common. You could if you wanted use a different convention of a -4 fielder who is a 4 run better hitter than the current offensive replacement level, and Cabrera and Trout would wind up with the exact same number of WAR. The distinction for how you attribute replacement level value between offense and defense is entirely irrelevant.

A player who is below average defensively but above replacement overall loses fielding runs with each game played but gains replacement runs and batting runs at a faster rate than he loses fielding runs. He adds total value with every game played. You could shift down the defensive replacement level and up the offensive replacement level and get the same player accumulating a positive defensive value but an offsetting lower amount of offensive value than he did before; he would accumulate the same total value as he did before.

I've also wondered about what the game would look like if it had separate offensive and defensive teams like the NFL. I suspect there would be no turnover among offensive stars, as essentially every above average hitter -- regardless of defense -- is in MLB now (which also speaks to DH replacement level). Maybe the lower 30% or so of current offensive players, the typical or below average middle infielders and catchers, would be replaced my minor league corner players. Over time there would probably be a shift toward developing a few more offense only Dunn types, but overall the upper half of the league on the offensive side wouldn't change much.

On the defensive side, initially I think GuyM is right that maybe 20%-35% of current players would still play defense. All the corner players would be replaced, but some up the middle players could keep their jobs. Over time, however, the sport would start to generate true defense-only players, who would over time take away 100% of the defensive slots. All outfielders might look like less muscular cornerbacks whose main skill is sprinting. First basemen might look like NBA guards, tall and quick. I suspect middle infielders and catchers, those already somewhat selected for defense, would change the least in appearance.

The reason people hate RBI is because "old school baseball men" love it so much, not because it's not useful if deployed rationally.

No. The reason I can't stand rbi is because they are almost totally a function of power (most simply represented by SLG, though breaking it down into BA and ISO actually works a bit better when explaining rbi) and opportunity (most simply represented by AB with runners on base)

If you've got SLG, rbi add nothing of value.

There is no value in knowing you'll have a below replacement level player in there every day. There is no value in being able to pencil in Yuniesky Betancourt for 150 games of -1 WAR.

You're better off throwing a bunch of random minor leaguers and waiver wire pick ups at the position

This is the polar opposite of right. There's no guarantee that a replacement level player will perform at replacement level. This is not finance, where a risk-free interest rate is obtainable. Risk-free replacement level performance is not available in the marketplace -- a fundamental flaw in the model.

#17 I've checked -- though my study was more than a decade ago and limited by the data at hand.

As always, there are a handful of players who meet a definition of clutch/choke, provided you set the line at 95% confidence. And that is completely predictable even if there is no such thing as clutch hitting ability.

I don't have the study handy, but I do have the top hitters from the 1990s in late/close.

The best of the 90s in late close are (by hand so I may have missed somebody): Pretty much a who's who of hitters from the 90s, though perhaps not in the precise order you'd expect.



(No it's not by OPS. It's by OBP*1.7+SLG)

A related study by David Grabiner suggests that good left-handed hitters with large platoon splits will tend to show up as slight choke hitters. They're more likely to see good LHP in a high leverage situation and be left in there to face the pitcher.

I'm agnostic on the subject. As I'm sure you know my standard response to anybody who brings up clutch is to ask them for a definition so I can actually run a study on the matter.

When I've looked at RISP, late close and various other definitions what I've consistently found is that a small number of players perform better or worse in the situation under study. That overall stats and situational stats aren't perfectly correlated (but are very highly correlated) and that the number of runs/wins at stake are vanishingly small.

No surprise really. No offensive metric is more precise than 5 runs (for a full-time player) and the primary source of error is clutch/timing (word choice is dependent on religious view of the situation)

All that to say is that OBP and SLG explain almost spot on 89% of the variance in team runs scored. Trying to build a winning team by focusing on the timing aspect (even supposing you could correctly identify the players with clutch ability) is not going to work.

If you've got SLG, rbi add nothing of value.

As descriptors of what happened in actual baseball games, RBI add a lot of value. The decision about whether to care about what happened in modeled baseball games, as opposed to what happened in actual baseball games is a philosophical choice, not a mandate.

near calculation is better than the TB*OBP approach.

For individuals yes. The difference is in the noise at the team level, but the basic problem with the multiplicative approach is that (using example from back in the day) a HR hit by Frank Thomas shows up as considerably more valuable than one hit by Joe Carter.

(From an old usenet post)

Doubt this? Take any of Joe Carter's seasons. Calculate the runs created. Then add a 1-1 with a HR and recalculate the RC. Repeat with any of Frank Thomas'. To pick one year at not random, in 1995 that extra HR is worth around 1.43 for Carter and 2.02 runs for Thomas.

For extreme players the error that this created gets very large.

I find it easy to make this point to strat players. Mike Admas' 2011 card has zip against left. Chapman's card has a whole pile of walks (but no hits). If (as one guy in my lague does) you use OBP*SLG they come out as equal.

the basic problem with the multiplicative approach is that (using example from back in the day) a HR hit by Frank Thomas shows up as considerably more valuable than one hit by Joe Carter.

The standard solution to this is to combine the player with eight average teammates, isn't it? Probably doesn't completely eliminate the effect, but reduces it a lot.

I don't think this is true. People who think Nolan Ryan is close to the best pitcher of all time are focusing on two statistics: strikeouts and no-hitters. Looking at even his other basic stats shows him as a very good-but-not-great pitcher. His ERA (3.19) is very good, but not outstanding. His ERA+ (112) is actually pretty mediocre for a Hall of Famer. His winning percentage (.526) is about the same. Ryan never won a Cy Young (and his Cy Young shares total is 30th all-time, which isn't exactly legendary). Of course, he walked the most batters of any pitcher in history, by far.

His wins total is his most impressive number, aside from the Ks and no-hitters, but it's also exactly the same as Don Sutton's, and no one ever accused Don Sutton of being the greatest pitcher of all time.

Those are all very basic numbers, except for maybe ERA+. Anyone who takes five minutes to study Nolan Ryan's career statistics would quickly understand that he doesn't belong in the discussion for greatest pitcher of all time. You don't need WAR or Win Shares to lead you to that conclusion.

I agree with all of this.

With Ryan I think what should be highlighted is that he pitched forever, which has value. Yes, his 112 ERA+ is no great shakes for a HOFer, but he maintained that ERA+ over 5400 innings. Which leads me back to your conclusion: very good pitcher, a deserving HOFer, but not in the discussion for all-time great. I agree one doesn't need WAR to get there.

I'm not a huge WAR supporter as people know, but I do find it useful in setting a baseline or ballpark figure of a player's value. In this case I want to know how much bulk value Ryan had over his career by posting 5400 innings of a 112 ERA+ with the defenses behind him. WAR tells me 77. That's useful to me.

This is the polar opposite of right. There's no guarantee that a replacement level player will perform at replacement level. This is not finance, where a risk-free interest rate is obtainable. Risk-free replacement level performance is not available in the marketplace -- a fundamental flaw in the model.

No, you fundamentally don't understand. Certainty is a bad thing at or below replacement level.

At or below replacement level, you don't want risk-free, you want risky. You want the highest variance you can get.

If I have a choice of 2 players who project as equivalent, one who has 4000 MLB PAs, and one that just got off the boat from Cuba, who I have fragmentary stats and a cursory scouting report on, I'll pick the unknown guy every time.

The unknown guy may turn out to be good. If not, discard him and replace.

Variance is your friend when the expected outcome is bad.

There is negative value in locking in sucky performance.

War for pitchers is useless

Repeating this doesn't make it true. I happen to prefer (the old versions) of Support Neutral W/L to WAR, but the two are highly correlated. WAR's handling of defensive support is potentially problematic, but that's something that can be argued on a case by case basis.

Now if you're talking Fangraph's WAR, fine. This is a fine example of smart people coming up with something silly.

Which necessarily argues for the inaccuracy of any or all of them.

Well it's important to understand the standard error of any given metric. I know that you just can't evaluate a player's offensive contribution more precisely than +/- 5 runs and their defensive contribution is more contentious -- though best I can tell the standard error of the estimates (taking the consensus of the best metrics that are generally available) is on the order of 7 runs.

I've argued for a long time that presenting the results to a single decimal place (with no indication of the standard error) is asserting more precision than the methodology supports (and that's true whether you're talking Fangraphs WAR, BBRef's WAR, Win shares ...), but that's not the same thing as inaccurate.

No, you fundamentally don't understand. Certainty is a bad thing at or below replacement level.

At or below replacement level, you don't want risk-free, you want risky. You want the highest variance you can get.

I do understand that.

The WAR model presumes that replacement level talent that will perform at the replacement level is available in the marketplace. It isn't.

That doesn't really matter if we're simply using the tool to evaluate performance ex post, as we can just call "replacement level" the level we're using as our baseline. If we're using the tool for any ex ante discussion, it absolutely does matter that there's no such thing as risk-free replacement level production available in the marketplace.

The point you raise is valid, but needs the important assumption that the expected value of production is the same. Sure, if one guy projects at an expected value of -1 WAR, with zero variance and another guy projects at an EV of -1 WAR with a lot of variance, it may be that you'll trade the shot at a 5 WAR year for the risk of a -5 WAR year. I don't think the answer is as clear cut as you suggest, though -- a team may be at a point or have the type of roster that makes it just take the sure mediocrity and not take the risk of disaster.

Under no circumstances should you take a guy with a big upside/big downside, but an expected value of -2 WAR, over a risk-free -1 WAR.

Intriguing, but how do you demonstrate this empirically?

The idea of an offense/defense platoon baseball game is actually interesting, in theory (though I'll stick with baseball, in the NL, with no DH's in reality, thanks much.) Why would you need to have 9 hitters in your offensive platoon? Why not go with 9 defenders but just five really fantastic hitters? And if you do that, and you're literally only sending the top 150 or so hitters to the plate in a season, why not give the defense a bit of a break and drop a rover/4th OF into the mix as well?

The idea of an offense/defense platoon baseball game is actually interesting, in theory

It's a thought experiment, to illustrate how much more scarce offensive talent is than fielding talent. The idea wasn't to design a new game. Though of course if someone wants to imagine a new sport using bats and balls, it's a free country....

As for what scoring levels would look like, with slightly better hitters and much better fielders, it's an interesting question. If you thought that BABIP might drop by 20 points (about the amount it rose in 1993-94 with the juiced ball), that would reduce scoring by about .5 runs/game for each team -- a big drop. But I dont' know if a 20-point drop is realistic. My guess is scoring would decline a bit overall, maybe .2 runs/game.

#154 It's also worth noting that a players may have a true talent of "replacement level" (whatever definition you use. And simply have a bad year.

The year to variation in offensive output for a regular player is on the order of 14 runs. IOW it's completely predictable that some teams will get -14 value over replacement (or worse) from a position.

Are the various WARs occasionally reconciled with team wins?

Win shares is. But the way that James gets there is (to put it mildly) extremely questionable.

BB-Ref's reconciles at a teams pythag. Actually that's not correct either. It reconciles at the pythag given the predicted runs scored from offensive events. Fangraphs WAR reconciles at pythag for "runs created" and projected runs allowed given the team defense and the parts of pitching (ie Ks/BBs and HR) that have no defensive influence.

I suspect that two-platoon baseball would be a fact by now if it really made a lot of sense. The existence of the DH tends to show that. But there are many big-leaguers who would play both ways at every other position but pitcher, and nothing really to be gained by "resting" them for the half-inning their platoon is off duty.

Put another way (and more germane to the discussion, those scarce good-offense players that GuyM is talking about are very often damn good defensive players too. Of the top 20 players in B-Ref's rField last year, 12 had OPS+ above 100 (and a couple of others, like Michael Bourn and Brett Lawrie, were basically league-average hitters). There's just no advantage to be gained from two platoons: even at SS or C, enough of the guys can hit to make it impractical. Perhaps one, at the outside two flexible "extra DHs" for great fielders might make some sense, but it's not like the guys who play DH right now are a phenomenal offensive contingent: guys hanging around in the minors with great hitting potential and no position are very thin on the ground.

BB-Ref's reconciles at a teams pythag. Actually that's not correct either. It reconciles at the pythag given the predicted runs scored from offensive events. Fangraphs WAR reconciles at pythag for "runs created" and projected runs allowed given the team defense and the parts of pitching (ie Ks/BBs and HR) that have no defensive influence

Thanks – that's very interesting. I knew that Win Shares starts with actual wins (I remember Bill James's explanation of why that sometimes produced weird results, and how he chalked it up to hell, they won those games anyway, live with it :)

It's interesting to think of a team composed truly of replacement players. I mean, the 2012 Astros come damn close. If there aren't guys hanging around the tryout camps who can hit better than Brian Bogusevic did last year, then the art of scouting is clearly in trouble …

#167 That would be true of below aver 1B/corner OF being compared to DHs. But it's not true of Cabrera -- who played third.

It's a tricky method issue. One that is actually handled better in a Win Shares structure (all defensive contributions are positive, just an issue of how much)

But there are many big-leaguers who would play both ways at every other position but pitcher,

I suppose it depends on what you mean by "many." But virtually no corner OFs or 1B would have jobs on defense, and a decent number of 3B and 2B would lose their fielding jobs too. On offense the impact would be smaller, but a fair number of MI and C wouldn't hit.

And FYI, you can't use rField to measure defensive value -- you have to add the position adjustment.

If there aren't guys hanging around the tryout camps who can hit better than Brian Bogusevic did last year, then the art of scouting is clearly in trouble …

No actual team will ever field a squad of replacement level players - by which I mean a team of freely available, near break-even talent level AAA journeymen. Because that makes no damned sense for a baseball franchise to do. You're either going to field above-replacement level talent, and pay the premium, or you're going to cut the bottom out of the rebuild and play whatever projectable talent you might have in your system, in order to get them reps and experience toward building "the next good MLB squad" for your franchise.

(all defensive contributions are positive, just an issue of how much)


As they should be. WAR adds together two concepts that aren't really even in the same units (though they claim to be) -- all batting events count, but only certain fielding events count. Fielding is expressed in something akin to "marginal runs," batting is expressed in "runs."

At its most fundamental level, Miguel Cabrera stood where a "third baseman" usually stands, with minor adjustments by his manager, and turned balls that would have been hits had he not been standing there, into outs. He, therefore, prevented runs.(*) How could his fielding value, on the same scale as his batting value, therefore be negative? That fundamentally doesn't work.

(*) He may not have prevented runs that other fielders wouldn't have prevented, but that's a different concept than batting runs, where we don't measure things at the margin, we simply measure things.

And FYI, you can't use rField to measure defensive value -- you have to add the position adjustment

My bad, I thought that was the one that did take that into account (that must be dWAR or something).

And I dunno about "no corner outfielders." I guess there's an almost endless supply of young centerfielders who would effectively fill all OF spots, though there are some pretty good major-league RFs around.

No actual team will ever field a squad of replacement level players

Not in an era with a draft and cost control of younger players, I agree. I think it pretty much was done sometimes back in the dark ages (by Connie Mack after a fire sale, e.g.) But yes, even the Marlins have Giancarlo Stanton, and as long as they're paying him damn all, why bother to replace him.

No actual team will ever field a squad of replacement level players - by which I mean a team of freely available, near break-even talent level AAA journeymen. Because that makes no damned sense for a baseball franchise to do. You're either going to field above-replacement level talent, and pay the premium, or you're going to cut the bottom out of the rebuild and play whatever projectable talent you might have in your system, in order to get them reps and experience toward building "the next good MLB squad" for your franchise.

It makes sense if all the players will have higher future expected WAR by putting up 0 WAR in the major leagues this year, as opposed to playing in the minor leagues.

Or someone could create a new statistical acronym that spells out LOVE.

LOVE stinks.

As they should be. WAR adds together two concepts that aren't really even in the same units (though they claim to be) -- all batting events count, but only certain fielding events count. Fielding is expressed in something akin to "marginal runs," batting is expressed in "runs."

This is wrong. Cabrera's Rbat last year was 52 runs. I assure you that is not Cabrera's total runs created. It is his runs above average.

So much passion, so little information. What's that about?

This is wrong. Cabrera's Rbat last year was 52 runs. I assure you that is not Cabrera's total runs created. It is his runs above average.

Expressing Rbat in terms of average doesn't change the fact that the inputs on offense are what the player did, while the inputs on defense (Defensive Runs Saved) are what the player did versus what other players did or were expected to do. If you hit a double, you are credited for a double even if 99% of MLBers would have hit the same pitch for a double. If you field a routine ground ball that 99% of MLBers would have fielded you get (essentially) no credit. One's a direct measurement, the other is a measurement of marginal impact.

That's the asymmetry I was referring to.

#179 Dave Tate came up with a solution (one that Bill James adopted. Probably without knowing that Tate had got there a decade earlier). Calculated team runs created. Calculate team runs created with the player's stats removed. Credit the player with the difference between the two totals. It's called marginal lineup value.

It's an awful lot of work, and doesn't address other minor issues with the runs created metric (for instance including DPs with no adjustment for opportunity. This is actually worse than not including them at all. Plus provably incorrect weights for sac bunts.)