Fan vs. PECOTA Projections

[El Gordo]

Well, to tell you the truth, when all this started I didn't really care. All I knew was that LH can't hit and that he sure looked like Looie Aparicio (plus some) on my TV. So, my initial thought was that it's too bad he can't hit a little better and steal some bases. Then, when I looked at the MiL numbers that everybody was using to call him "the worst baseball player ever", and compared that to his small-sample of ML numbers, I decided "I hope he can remember how to take a friggin' walk." But now, with all this, I'm kinda hoping he does play, just so I can see what happens ;-)

My hunch is that there are some folks who have been reading all this, but who haven't said much, who are looking at what I posted and are sitting there thinking, "But-but-but... that *can't* be right!" I think it's right, at least as far as I know. I haven't invented a single thing, I'm just taking what other people say and seeing how the number work out, that's all. If we think stats have value, isn't that what we're supposed to do?

Here's what I think a lot of the disbelief is about: It's the constant myopia around here about OPS. I don't think it's mostly because anybody is *trying* to do that, I think it happens by accident. It is entirely normal around here for people to judge guys by looking at a single number, and in recent years BA has been replaced by OPS as that single number. People see OPS used as the main basis for judging guys all the time. So, as an accidental side-effect of that, some people accidentally decide that what matter is OPS and, by the way, a little D would be nice if you can get it. Maybe people don't quite say that out loud, but the net result is the same. I think most of that is due to the fact that we hear OPS all the time. To me, this is similar to what happens on the news about international finance. Almost nobody actually knows squat about international finance. But when you hear the news-heads talking about the Chinese yuan all the time, you figure it matters a lot. I bet any poll about which foreign currency matters most would show tons of votes for the Chinese yuan, just based on what we hear all the time. But if the news-heads went back to talking about the Japanese yen non-stop for a few months, then everybody would think that's what matters. I think the same thing happens here about OPS. Personally, I think OPS is a good stat. But I also think excellent-D counts more than a lot of folks seem to think it does.

But, anyway, all the talk about how playing LH will be tantamount to the FO being lazy and throwing the season is just nuts. I hope the FO worries about who they want long-term, and I don't want them wasting much time and resources worrying about getting a mediocre temp-solution. And, as for all the dime-a-dozen temp-solution who they can get for nothing, I'm still waiting to see that list. So far, it's mostly a bunch of older AAAA utility guys. I'd rather watch LH play SS. At least he's pretty to watch. While we're waiting for the team to get good, I want to have *somebody* who's pretty to watch.

I think the next interesting comparison would be LH vs Derek Jeter. How would a poor hitting, good fielding SS rate against the worst fielding SS in the game? I agree the value of defense is consistantly devalued because offensive numbers are easier to compute and are team independent for the most part. But the people who play the game are quick to give strong emphasis to defense. Makes you wonder.

[DrungoHazewood]

I don't see why not. I don't know what he will actually do if he plays, but everything I look at says that he can:

• To be league average, he needs to generate 63 runs.
• If hits according to Bill James' 2008 projection (.282/.296/.578), in 500 AB's he'll generate 41.7 RC.
• Adjusting the raw-RC by the factor Bill James uses in everybody's Win Shares, that 41.7 translates to a value of 40.1 RC.
• Which means he's 22.9 runs shy on offense. So, to be league average, his defense has to worth 22.9 runs.
• Adjusting DR by the factor Bill James uses in everybody's Win Shares, he has to be worth 22.0 raw-DR in the field to get the total up to 63 runs.
• Notice that the fact that I'm adjusting CR/DR like Bill James does in Win Shares actually hurts LH. Without doing that, his raw 41.7 RC means he only has to produce 21.3 DR to be average, but I'm saying he has to find 22.0 DR instead. I'm being harder on LH than are the folks who think we should just ignore the different relative value of CR and DR. So, it's not like I'm cheating in favor of LH. I'm just trying to do it right. (Ye of little faith, please raise your hand if you thought I was doing sneaky pro-LH cheating ;-)
• When 1970 did his exercise in the other thread using range data, his calculation for LH's defense came to 39 DR. Now, I understand that that is based just on ML data for all the guys he was looking at, and I agree that the sample is too small. But I think that gives us more reason to doubt the crappy DR it gave to Hu and Aybar than to have huge doubts about LH. If he actually showed the range to do that, why not give that *some* credence? I can see how Hu fans might complain, on the grounds that the sample didn't permit Hu to show off his range. But I don't see how it could completely fabricate range that LH doesn't have. However, I can see how the high value might seem suspect. In that same thread, 1970 said he thought that 30 DR was prolly more sensible than 39, which is discounting the stat-value by 23%. Either way, both of those are way higher than what LH needs to be league average. I don't see why it's a big stretch to think he would get 22.0 DR, which would be discounting the stat-calc of his range by more than 40%. Or, if you think I'm being too tough on LH, and want to use the 21.3 raw-DR instead, then you could discount the stat-calc of his range by fully 45% and he'd still be be a league-average SS. That doesn't seem highly unlikely to me.

If we use .782 runs for each extra play, that means that he's got to make 28.2 extra plays (or 27.2 extra plays if we don't do the Win Shares adjustment). What's so crazy about that?

ps: That's to be league average. IIRC, tangotiger was tolerating a net cost of half-a-win. If you wanna do that, then you can reduce the extra-plays needed by slightly more than 6.

I haven't had time to dissect this yet, but there have to be some serious flaws/optimistic assumptions in an argument that ends up saying a shortstop with the worst (or at least one of the very worst) offensive output in the majors ends up being league average.

With the bat Hernandez is about equal to the 2007 version of Adam Everett. If Everett had played 162 games he'd have been in the neighborhood of 2.5-3 wins over replacement. That's about +7.5 to +9 Win Shares - assuming that his defense is truly spectacular.

I think for your post to make sense you have to assume that Hernandez is either going to hit better in the majors than he did in AA, or he really is one of the 2-3 best defensive shortstops in the world. IMO that's a gigantic stretch.

It's one thing to say LH won't completely kill the team - I can construct that argument and believe it. I think it's much, much, much further out in left field to say he'd be a major league average shortstop.

[RShack]

I haven't had time to dissect this yet, but there have to be some serious flaws/optimistic assumptions in an argument that ends up saying a shortstop with the worst (or at least one of the very worst) offensive output in the majors ends up being league average.

With the bat Hernandez is about equal to the 2007 version of Adam Everett. If Everett had played 162 games he'd have been in the neighborhood of 2.5-3 wins over replacement. That's about +7.5 to +9 Win Shares - assuming that his defense is truly spectacular.

I think for your post to make sense you have to assume that Hernandez is either going to hit better in the majors than he did in AA, or he really is one of the 2-3 best defensive shortstops in the world. IMO that's a gigantic stretch.

Oh, it's easy to dissect if you just look at it for a sec. It all comes down to whether you believe that Win Shares make sense when they get down to the level of individual contribution. All along, I've said that I respect Win Shares as a top-down accounting method, but that it blurs all kinds of things together when it gets close to individual players. This becomes clear when you do what Bill James did in his article about how Everett and Jeter compared in 2005:

• I don't know how many different versions of Win Shares are floating around, but the one I found online for 2005 gives Everett 1.6 on offense, 8.9 on D, for a total of 10.5 Win Shares. Jeter got 19.5 on offense, 5 on D, for a total of 24.2 Win Shares. (Now, if Bill James Win Shares say something else, that's fine, I'm happy to use them, I just don't have them, so I'm using what I've got.) Using 3 Win Shares per win, that means Win Shares say Everett was worth 3.5 wins, and Jeter was worth 8.1 Wins. So, according to Win Shares, Jeter was worth 4.6 wins more than Everett was.
• But if you look in Bill James article comparing the 2 guys in detail, you see something quite different. There, he credits Everett with 61 RC and about 17 DR, which gives 78 total runs or about 7.8 wins. He credits Jeter with 105 RC and -19.5 DR's which gives 85.5 total runs or about 8.55 wins. So, using Win Shares Jeter is supposedly worth 4.6 wins more than Everett, but when Bill James looked at it closer, the answer shrinks to 0.75 Wins. So, who are you gonna believe: Bill James on Win Shares or Bill James when he looks closer? Notice that the version of Bill James that gave a bunch more weight to D looked at D 3 or 4 different ways to get the pro-D answer, not just one. In large part, this is because he fully admits that he doesn't have any single good way of assigning D-value. However, that doesn't seem to stop people from making decisions about D-value based on Win Shares.
• Of course the big fly in the Bill James ointment is that he figured plays which take away singles are worth about half-a-run or less. Meanwhile, both basic logic and tangotiger tell us that that's not true, simply because outs have some value too. When you look at what the second version of Bill James said, and adjust for that, then the conclusions change even more. If you use the .782 value for singles (it's for 2007, I dunno what the right value is for 2005), then the DR's change accordingly: Everett's goes to +26 and Jeter's to -30. Which gives them total runs of 87.6 and 74.5, respectively. In other words, it makes Everett have a net value of 8.8 wins vs. Jeter's 7.5 wins.

So, who are you gonna believe? Win Shares, or Bill James when he looks closely but uses only half-a-run for singles, or Bill James if you use .782 runs for singles instead of half-a-run? That's what it all comes down to.

This goes back to what I said some time ago, when I was goofing about Shack's Law: the earlier the method, the more it's gonna overrate whatever is easiest to measure. IMO, that's exactly what you see here:

• Looking just at offense, Win Shares gives way more credit to offense than just RC does. In the case of 2005 Jeter vs. Everett, WS-O credits Jeters bat as being worth 5.9 wins more than Everett's bat, but RC says Jeter's bat is worth "only" 4.4 more wins that Everett's. If you trust RC/10, that means Win Shares is inflating Jeter's offensive impact by 33%. That's a lot. Or, if you trust Win Shares, then you think RC/10 is off by 25%. That's also a lot. Which do you believe? No matter how you look it, something's wrong somewhere. Personally, I'm thinking it's Win Shares, but I could be wrong.
• Looking just at defense, the difference is way more dramatic. WS-D says that Everett's glove was worth 1.3 wins more than Jeter's glove. But when Bill James looked closer, it jumps up to 3.65 wins, which is 281% of what Win Shares says. Of course, that's using half-a-run for a single. If we use .782 instead, Everett's glove is 5.7 wins better than Jeter's, which is 439% of what Win Shares says.

No matter how you cut it, either the conclusions Bill James reached in his article on Everett vs. Jeter are just flat out bogus, or else Win Shares dramatically overvalues offense and even more dramatically undervalue defense. Which do you think it is? That's the real question here. The secondary question is how much a single is worth: is it about half-a-run like Bill James used in his informal reasoning? Or is it between .75 and .80 runs like tangotiger said. I really don't know, but I'd put my money on the latter.

The disagreement isn't so much about LH, as it is about having a reasonable way to value D. I think it's instructive that in the Everett vs. Jeter article, Bill James looked at D-contribution 3 or 4 different ways. I don't know what that tells you, but it tells me that Bill James is admitting that he doesn't have a single trustworthy way to do it, which is why he did it several different ways. And, since all the ways he looked at agreed with each other, that means that every one of them said that Win Shares is just wrong about this. To me, after looking at what Win Shares actually does, this is not in the least bit surprising: the closer it gets to the individual player, the more Win Shares blurs team phenomena up into one big stew and then ladles out portions to players in ways that holds them responsible for what their teammates did and didn't do. Again, I'm not dissing Win Shares as a top-down accounting system, I'm just saying that as it gets close to an individual player, it gets all blurry and therefore untrustworthy. It's a garden rake when what we need is a comb.

It's one thing to say LH won't completely kill the team - I can construct that argument and believe it. I think it's much, much, much further out in left field to say he'd be a major league average shortstop.

It's one thing to say that Win Shares have some utility. I think it's much further out in left field to say that Win Shares gives a useful view of the value of an individual player's defense. I think it's much more reasonable to conclude that Win Shares overstate the value of offense and dramatically understate the value of defense. This conclusion is entirely consistent with Bill James detailed look at Everett vs. Jeter. It is also consistent with the history of efforts to measure performance scientifically: whatever can be measured easiest gets measured first, and it always gets way overrated until people figure out how to rate the harder stuff. It happens every time, so it's not surprising that it's happening here. At the very least, we should just admit it.

[RShack]

Also, there's a couple things about RC that I'm a little dubious about. I'm hoping that somebody here who knows more about it than I do can fill me in...

In the previous post, I pointed out how Win Shares overstates offensive value compared to RC (by which I meant the AB*OBP*SLG formula). But it seems to me that even RC overstates it too, in a way that gives too much credit to good hitters and not enough credit to bad hitters. I'm not exactly sure about this, but here's how it seems to me. If RC is based on OBP and SLG, then it's implicitly assuming that each guy is hitting in a lineup of consists of 8 other guys just like him. I say this because OBP measures just getting on base, and the value of that has a lot to do with whether or not somebody else drives you in. So, how much Actual Value getting on base has is somewhat dependent on what somebody else in the batting order does after you get on base. For SLG, the value of an extra-base hit is somewhat determined by whether or not somebody else was already on base (so your double could drive them in) and whether or not somebody who bats after you gets a hit (so that you can score from 2B after you hit a double). If nobody does any of these things, then what you do with your bat doesn't matter nearly as much. So, the real value of a guy's OBP and SLG is influenced by what other guys in the lineup do.

But since RC considers nobody else, isn't it assuming that the player in question is also filling out the rest of the batting order? Isn't LH's RC based on the assumption that he's hitting in a lineup of 9 LH's? And isn't Jeter's RC assuming that he's hitting in a lineup of 9 Jeters? It seems to me that this virtually guarantees that RC will overstate the value of a good hitter (one who hits better than his team) and will understate the value of bad hitter who hits worse than the rest of his team. For example, once LH gets on base, whether or not he scores is not up to somebody who hits like him, it's up to somebody who hits better than him. So, while it appears that Win Shares overstates offensive value more than RC does, it also seems like RC would credit a good hitter with too many RC and a bad hitter with too few RC. So, if we compare RC for Jeter and LH, then we all agree that Jeter will generate more RC, but I would think that the formula gives Jeter credit for too many RC, while giving LH a RC number that's too low. If I'm wrong about this, can somebody explain why I'm wrong?

The other thing I wonder about is RC treating OBP and SLG like they have exactly the same weight. I don't know how much weight they should have, relative to one another. But it seems highly unlikely to me that they would both deserve exactly the same weight. That just seems like one hell of a coincidence to me. I would think that it would be much more likely that their weights should be different rather than exactly the same. If somebody had asked me which statement seems more likely to be true:

• That a run-scored has exactly the same value as a run-prevented, or
  
• That OBP and SLG each have exactly the same weight in creating runs,

then I would have bet that the first choice is way more likely than the second choice. Wouldn't you? Yet Bill James determined that runs prevented are 8.3% more valuable than runs scored. So, wouldn't it seem likely that OBP and SLG would turn out to deserve different weights? Doesn't it seem kinda odd that they're weighted exactly the same?

[RShack]

Reference RC, keep in mind that the formula you're referring to (AB * OBP * SLG) is intended to be used for quick and easy calculations. The actual formula is much more complicated (particularly the team based formula introduced by Bill James a couple of years ago).

As for the issue of treating OBP and SLG differently, why would you want to when the q&e calc method tends to give you results that match up so well with the results of the actual formula?

Keep in mind that I'm not saying they shouldn't be weighted equally for other purposes, only for purposes of the q&e calc method. I saw a study a few years ago that concluded that OBP should probably be about 1.8 times more (making it [1.8*OBP] + SLG).

Oh, it's not that I *want* OBP and SLG to be different, I'd just be surprised if their proper weights weren't different, that's all.

If I understand you correctly, you're saying that treating OBP and SLG as equal factors is appropriate simply because it does a good job of approximating the more-detailed formula... is that correct?

It always seemed to me that OBP should carry more weight, but I never really thought it through or looked into it. For that study that said it should be 1.8 OBP + SLG, do you recall what the context was? Were they proposing as different version of OPS as a single summary number? Or were they proposing a different way to calc RC? Or were they proposing something else? (Or do you have brain-rot like I do, in which case you prolly can't remember ;-)

As for the more-detailed formula for RC, I'd expect that it would also over-estimate the RC for a good hitter and underestimate it for a bad hitter, just like the short version. Do you have either news or an opinion about that?

[RShack]

Reference your last paragraph, I don't have an opinion. My guess is that it probably does, but maybe not as much.

I wasn't making any guess about much the RC-calc inflates RC for good hitters and deflates RC for bad hitters. It just seems inherent in the formula that it would, but I have no idea how by much it does it.

It seems pretty clear that Win Shares doesn't give enough credit to D, and therefore overstates offense, as shown by Bill James' analysis of Everett vs. Jeter. I just wondered how much the RC-calc over-credits good hitters and under-credits bad hitters. I think it does, and I'd guess the amount varies. I don't know if there's any rule-of-thumb about how much. I was just hoping somebody might know.

Drungo? tangotiger? anybody?

[El Gordo]

I wasn't making any guess about much the RC-calc inflates RC for good hitters and deflates RC for bad hitters. It just seems inherent in the formula that it would, but I have no idea how by much it does it.

It seems pretty clear that Win Shares doesn't give enough credit to D, and therefore overstates offense, as shown by Bill James' analysis of Everett vs. Jeter. I just wondered how much the RC-calc over-credits good hitters and under-credits bad hitters. I think it does, and I'd guess the amount varies. I don't know if there's any rule-of-thumb about how much. I was just hoping somebody might know.

Drungo? tangotiger? anybody?

Isn't it RC 27 that factors in a team of 9 of the same players. I don't know that RC does that, but then, most of this is Greek to me and I only speak a little Bulgarian.

[RShack]

Isn't it RC 27 that factors in a team of 9 of the same players. I don't know that RC does that, but then, most of this is Greek to me and I only speak a little Bulgarian.

I don't think RC (based on OBP and SLG, or the long version) does it on purpose. I think it does it implicitly.

[tangotiger]

Use Linear Weights, not RC.

And I was the one who showed the 1.8. It most closely aligns itself to Linear Weights. And, it also happens to correlate the best to team run scoring.

[RShack]

Use Linear Weights, not RC.

And I was the one who showed the 1.8. It most closely aligns itself to Linear Weights. And, it also happens to correlate the best to team run scoring.

OK. I have a couple questions:

1. Do you have a quick-and-dirty RC calc using that 1.8 factor that produces better results than the one that treats OBP and SLG equally?

2. As for the Linear Weights calc, that makes perfect sense to me. Is there a web site that provides the result of that calc for everybody in an ongoing way throughout the season, and projects it for the whole season?

3. Same queston as #2, but for DR using either of the range methods?

thanks...

[RShack]

I was hoping to see some responses to what I said in posts #91 and #96, above. What I said there is right as far as I can see. But maybe there's important stuff that I'm not seeing. So, it's not like I'm claiming it's right, I'm just saying I don't see how it's wrong.

I could care less about winning an argument about it, I just wanna know the best way to look at it. So if anybody has any suggestions about how to look at it better, and fix whatever I said that was wrong, I'm all ears.

[Baltimoron]

OK. I have a couple questions:

1. Do you have a quick-and-dirty RC calc using that 1.8 factor that produces better results than the one that treats OBP and SLG equally?

He did one here:

So, the real equation is this:

Runs: (1.73*OBP + SLG) * 0.27 * PA

In order to convert from runs to wins, you need to divide by something around 10 to 11 (that’s your runs to wins converter).

Wins: (1.73*OBP + SLG) * 0.025 * PA

In order to compare to league average, you would do:

lgOPS: 1.73*lgOBP + lgSLG, which, as luck would have it, is very close to 1.0 (it’s actually 1.014). For simplicity’s sake, if we make the equation:

1.69*OBP+SLG, this will give us a result of exactly 1 for the 2006 season.

So, to compare to league average, you can have this version of OPS:

Wins above average = (1.69*OBP+SLG - 1) * .025 * PA

A guy with a .430/.670 and 600 PA would give us:

Wins above average

= (1.69*.430+.670 - 1) * .025 * 600

= +6.0

here

[Baltimoron]

2. As for the Linear Weights calc, that makes perfect sense to me. Is there a web site that provides the result of that calc for everybody in an ongoing way throughout the season, and projects it for the whole season?

I think you could use BRAA on Fangraphs, which is LWTS by the 24 base/out states, for runs.

Or you could use WPA/LI, which is LWTS (in Wins) by the Game State (inning, score, base, outs), with the Leverage depressed.

[Baltimoron]

If we use .782 runs for each extra play, that means that he's got to make 28.2 extra plays (or 27.2 extra plays if we don't do the Win Shares adjustment). What's so crazy about that?

ps: That's to be league average. IIRC, tangotiger was tolerating a net cost of half-a-win. If you wanna do that, then you can reduce the extra-plays needed by slightly more than 6.

I am with you, but I think you fail to appreciate just how good a defender one needs to be to be 28 (or even 21) plays above average at SS.

<img src="http://www.billjamesonline.net/fieldingbible/charts/leaders1-07.gif">

<img src="http://www.billjamesonline.net/fieldingbible/charts/2006plus-minus.gif">

[DrungoHazewood]

I am with you, but I think you fail to appreciate just how good a defender one needs to be to be 28 (or even 21) plays above average at SS.

<img src="http://www.billjamesonline.net/fieldingbible/charts/leaders1-07.gif">

<img src="http://www.billjamesonline.net/fieldingbible/charts/2006plus-minus.gif">

That's the point I was trying to make many posts ago. For a shortstop with a .550-.600 OPS to be an average MLB player he has to be one of the 2-3 best defensive players in the world.

We have no reason to believe that LH is that. A lot of guys appear to be (and in fact are) really good, and then grade out at +5 or +10 plays on the +/- scale.