Tango on Baseball Archives

Double-counting Replacement Level (August 25, 2003)

I sent the following to Baseball Prospectus. I'll let you know if I get a reply. If I understand your process, Wins above replacement is equal to the offensive wins above offensive replacement plus the defensive wins above the defensive replacement.

In essence, offensive replacement is 2 wins below average, and defensive replacement is 2 wins below average, so that an average player is 4 wins above replacement.

Did I get this right?

So, a team of average hitters would be 36 wins above replacement. Assuming you've got a team of average pitchers at 18 wins above replacement, that gives you a team of average players as 54 wins above replacement (or 81-54=27 or .167).

In order to do "above replacement", you can't combine it as I'm describing in the first paragraph. You first compare to average for everything, and then, in the very last step, do your above replacement calculation. You will see that the team of average players will be about 18 wins above replacement, so that an average team of hitters and pitchers will be 36 wins above replacement (which sets the team replacement level at 81-36=45 wins (or .280), something that makes more sense.

***
This part was not sent to BP.

You can see how Mike Schmidt looks that shows that the replacement level is being double-counted.

If you want to think about it logically, the backup 2B or SS on your team is not both a replacement level fielder AND a replacement level hitter. Such a player would be playing in Double-A ball at best.

The implication with the double-counting error at BP is
1 - players with long careers are overvalued (by 2 wins per year)
2 - hitters are overvalued compared to pitchers

--posted by TangoTiger at 02:42 PM EDT

Posted 3:00 p.m., August 25, 2003 (#1) - Patriot
The interesting part about the BP's problem with replacement level is that the Woolner study should have shown them that this was the case. The Woolner study set the replacement level at 80%, except for catcher(85%) and first base(75%). A simple and IMO logical explanation for this is that since first baseman are chosen more on offensive ability, the replacement first baseman tend to be worse hitters but better fielders then the replacements at other positions(relative to their position of course), and the opposite for catchers.

Posted 3:32 p.m., August 25, 2003 (#2) - tangotiger
Great perspective Patriot!

FWIW, using on 2001 superLWTS, and setting 300 PA as the line between regulars and backups, I get, on a /680 PA:
regulars: +6 overall, + 6 batting, 0 fielding
backups: -22 overall, -21 battting, -1 fielding

The *players* that are replacement level (backups, or shades below backups) are *average* fielders.

There's no such thing as a replacement level fielder or replacement level hitter... there are replacement level *players*. A replacement level player turns out to be an average fielder.

Posted 3:52 p.m., August 25, 2003 (#3) - Patriot
That has interesting applications beyond just the old baseline debates it seems. For instance, it would seem as if the proper baseline for a fielding method would be .500, since the replacement level player will do the same. Obviously we know there are better fielders out there, but if they never actually get to play, we shouldn't really care in a value method.

Furthermore, that suggests that the proper way to measure a player is to compare their hitting to a replacement at their position(I would still advocate the chained replacement or some different approach altogether), and then just tack on the runs saved above average on defense. It helps Edgar Martinez' HOF case, IMO. If we know that the guy who "replaces" him in the field will most likely be average, he shouldn't be penalized for being a DH. I guess you could say that his being a DH limits the team's flexibility, so maybe a small penalty, but not the crippling one that he would get if everybody else has their fielding measured vs. a replacement level fielder.

Posted 3:55 p.m., August 25, 2003 (#4) - David Smyth
That's an interesting comparison. I'm not sure what conclusion to draw. It almost seems like, on an overall basis, managers don't worry about fielding; they just select the best hitters at a position. If they happen to get a good fielder also, that's great, but they don't seem to actively seek it out. Maybe that's because they simply have very little ability to evaluate fielding. IOW, they think that some of these regulars are better defenders than they really are.

Posted 4:22 p.m., August 25, 2003 (#5) - David Smyth
I think that BP would (will?) say that they are aware that their setup does not add up to an actual ML repl player, and that they did not intend for it to be taken as such. They want to compare a plyer to the worst players in MLB in every phase of the game. There is probably some logic in doing this, but the problem is that people will regularly misinterpret the results.

Posted 4:29 p.m., August 25, 2003 (#6) - Patriot
That may well be the case, and there may be some very defendable logic behind it. But as Tango pointed out, it kind of torches the whole idea of comparing pitchers to hitters. Also, the entire idea of a replacement level in the first place rather than some other baseline is to attempt to conform to the economic/structural realities of baseball teams. If you're not doing that, what's the point?

Posted 4:30 p.m., August 25, 2003 (#7) - tangotiger
In terms of "value" for Edgar, I always look at it this way: "How would a baseline player do, if he was in Edgar's shoes?" From that standpoint, your baseline player will have no fielding contributions, just like your baseline AL pitcher has no hitting contributions.

I agree that Edgar theoretically limits the way the team is set up, in that if they had a truly horrible fielder, they couldn't hide him at DH..... except, have you looked at the way teams use the DH? It's not reserved to just the bad fielders. You'll get decent fielders in there. Edgar being in the DH doesn't really affect the way teams do their business.

Posted 4:50 p.m., August 25, 2003 (#8) - tangotiger
I agree with Patriot. BP should never have added it up as they do.

I mean, why stop there? Why not set the replacement level for batting, for stealing, for taking the extra base, for range, for throwing, for every facet of play? And then add it up.

The idea of replacement level is exactly what Patriot is saying: that a replacement level player = 0 wins = 0 (or 300K) dollars in salary. It's the minimum level of play in which you will get paid MLB dollars.

As I mentioned in my scales article a few months ago, *first* compare everything to average, and then, as a *final* step, compare to replacement. Do *not* have your intermediary steps do replacement as well.

Posted 6:17 p.m., August 25, 2003 (#9) - Robert Dudek
Tango,

I wonder how regulars/backups break down by position. For example, are backup 1Bs/SS/2B etc better or worse than average fielders ?

Posted 6:41 p.m., August 25, 2003 (#10) - David Smyth
"I agree with Patriot..."

So do I, if you were referring to my post. I was trying to explain it, not defend it...

Posted 9:48 p.m., August 25, 2003 (#11) - Patriot
I knew that. I said "That may well be the case, and there may be some very defendable logic behind it". Don't take my refutation that follows as trying to explain why you were wrong; just fleshing out my position a little more.

I sent an email to BP a while back about how MLV still is wrong for using RC as its basis, and never got a reply. I also sent Shandler an email about his lousy RAR method and never got a reply. However, I assume they get a lot of email and don't have time to answer all of them. But this is a very logical criticism and I would hope that they at least consider it.

Posted 10:52 a.m., August 26, 2003 (#12) - tangotiger
I looked at 1999-2002 UZR. I selected, by year, all players with at least 81 "UZR games" (treat that as "full" games), including if they had 60 games at 2b and 30 games at SS. Those are my regulars.

Then, by position, I figured the regular's UZR/162. I did the same for the backups. Here are the results.

pos Regular Backups diff
3 0.5 -1.8 2.3
4 0.8 -1.9 2.7
5 2.0 -5.0 7.0
6 1.7 -6.0 7.7
7 0.9 -1.5 2.4
8 1.2 -3.9 5.1
9 0.6 -0.9 1.5

So, we see that the regulars are slightly above average fielding-wise, at about +1 relative to all players at their position. The backups are -3 relative to all players at their position. That makes the difference, 4 runs, how much an average regular is better than an average backup, fielding-wise.

If someone wants to repeat this exercise for hitters (I'd HIGHLY suggest using LWTS) by position, that would be nice.

Posted 6:23 p.m., August 26, 2003 (#13) - Patriot
I have some numbers based on ERP/out from 1990-1993 that I did a while ago. I know it's not exactly what you're looking for but I have all the data already. I took the guy with the most Games Played at the position as the starter. Here are the ERP/outs as % of league average at each position:
pos reg others
2 98.1 70.7
3 124.9 99.3
4 99.3 75.0
5 103.3 78.9
6 83.7 74.3
of 111.4 84.9
dh 1.171 89.9

the differences/450 outs(assuming .175 r/o for the league)
pos diff
2 21.6
3 20.1
4 19.1
5 19.2
6 7.4
of 20.9
dh 21.4

About 20 runs difference except for SS which is a major outlier here.

Posted 8:49 p.m., August 29, 2003 (#14) - Michael
Are you sure that all BP does rep level that way? Someone asked GH about replacement level at the last Pizza Feed and he claimed a replacement level player as he used them at BP was a replacement level hitter who was a league average fielder at his position.

Posted 9:02 p.m., August 29, 2003 (#15) - Tangotiger
I don't know what "all BP" does. I'm just telling you what is on their site, and you can see the result by looking at Mike Schmidt.

Look at they do, and not as they say.

Posted 10:33 a.m., August 31, 2003 (#16) - David Smyth
Tango found -22 runs/162 for the backups (all players with <300 PAs). Just for the record, I prefer a different concept of repl level. It doesn't matter to me what an avg backup hits, or the effect of chaining. It doesn't matter to me how good an actual expected repl would be. What matters to me is the performance of the the worst players who are still good enough to play in the league, because these players are making the baseline contribution to wins and losses.

So I also looked at the Slwts (the 2000-2002 instead of just the 2001). I found the avg of the worst 20 players who had at least 900 PAs over that period. It was about -35 runs/162. That works out to a .350 player. A team of such players would win about 20% of their games. Is that hard to accept, considering that the Tigers, who are bad but certainly not composed of the 25 worst players in baseball, are winning about 25% of theirs?

Posted 10:44 a.m., August 31, 2003 (#17) - David Smyth
Let me add that most of those 20 players are still playing in 2003. There are reasons for this. In short, the managers think that these guys are better than they really are, for a few different reasons. If the baseball selection process were perfect, then the baseline player would be somewhat better, say .400 instead of .350. But it really doesn't matter, because wins and losses are simply a product of who is actually playing and how they are performing, not who *would* be playing in a perfectly selected league.

Posted 12:56 p.m., August 31, 2003 (#18) - Tangotiger
You have to regress their observed performance to establish their true rates. At 900 PAs, you probably regress about 25%, so that -35 runs would come in at -26 runs.

Posted 12:58 p.m., August 31, 2003 (#19) - Tangotiger
To put it another way, you selectively sampled your players by looking at their performance after the fact, and selecting on that. That's a no-no.

However, if you take ALL those players based on 2000-2002, AND THEN, tell me what their average performance was in 2003, then, that's the correct figure to use as your replacement level.

And, it'll probably come in at around -26 runs or so.

Posted 2:06 p.m., August 31, 2003 (#20) - David Smyth
I don't see why regression should be part of this. I don't care "who" the players are each year who perform at -35 (IOW, whether they regress towards the mean the next year and another group of -35 players appears). As long as the bottom level is pretty consistent each year at -35, then that is what should be used (according to my concept of replacement).

Posted 4:45 p.m., August 31, 2003 (#21) - Tangotiger
You can't selectively sample your group after-the-fact on the metric that you are studying. To combat this selection issue, you regress. Otherwise, your sample is tainted.

Why did you choose a PA cutoff then? Why not select ALL players, and take the worst runs / PA of the bunch? If you have a guy who's rate is -60 runs / 600 PA, but he did this after only 25 PAs, then so be it. I agree, ridiculous.

Posted 4:55 p.m., August 31, 2003 (#22) - David Smyth
Tango, you're right (as usual). You should regress, even in this application.

So I have an observation (more of a question, I guess). Let's say that a team will replace a player when there is more than a 50% chance that he is of replacement ability. So they could use the standard 3 weighted seasons approach. But in the real world teams have to make these the decisions in real time, and there is an "urgency" consideration, to minimize the "damage" by such a player.

So let's say that a player is 35 yrs old, and was a .475 player the last 2 full seasons. He starts out the season 0 for 100. Adding this to his prior seasons will suggest that he is still an above-repl player (I assume). But what if you apply a suitable, larger regression to those 100 PAs. Is it possible that on such a test the player will show as having a >50% chance of being below repl level? IOW, how well have these regression factors been determined? And how do we know that we should apply the same regression to every 100 PA sample, regardless of the performance quality of the sample. I mean, if some test shows that 100 PA samples should be regressed 70% (just making up a number), why should we assume that it applies to players who go 0 for 100, or 26 for 100, or 100 for 100?

Posted 9:40 p.m., August 31, 2003 (#23) - Tangotiger
The reliability of a metric is always dependent on the number of trials, and not the number of successes.

If you flip a weighted coin, and you get 73 heads in 100 flips, is that 73 successes or is it 27 successes? 73 hits may be a success to a hitter, but 27 would be the success to a pitcher.

Posted 7:38 a.m., September 1, 2003 (#24) - David Smyth
So if you are supposed to regress, say, 95% for a 100 PA sample, and if a Roy Hobbs comes out of nowhere and goes 98 for his 1st 100 ABs, you would still regress that 95% of the way towards .260, for an estimate that Hobbs is actually a .296 hitter?

Posted 7:46 a.m., September 1, 2003 (#25) - David Smyth
The difference between a coin flip and a hitter is that the bias of the coin cannot change, while the ability of a hitter can change (see Bonds, Barry). I suspect that some statistician might say that the normal regression would not apply in my Hobbs example above.

Posted 7:57 a.m., September 1, 2003 (#26) - Tangotiger
Yes, it would apply. That is simply your best guess.

In the case of 100 PA, you'd regress probably 70% towards the mean. So, if goes 100 for 100, and the league mean is .300, your best guess as to the true talent level that would produce such an observed rate is .510.

You can instead of using a weighted coin, you can use an unbalanced die, where the weighting of the die changes for every roll, but skewed towards say landing on 4,5,6. This would be like a human where his "true rate" changes PA-by-PA, but centered around something.

Posted 1:58 p.m., September 2, 2003 (#27) - J Cross
Wouldn't you just figure that the 100 for 100 batter isn't playing by the same rules and isn't part of the same distribution as the other players? I don't like the chances of a .510 hitter going 100 for 100. It's probably more likely that the batter is life form another planet who has come to earth with the sole purpose of messing with Tango's stats.

Posted 4:30 p.m., September 3, 2003 (#28) - Kevin Sharp
Following up on #27, a simple binomial distribution would tell you that a Hobbs who goes 98 for 100 is probably at least a .930 hitter. Applying the same regression to a .290 hitter as a .980 hitter seems a little foolish. At the extremes (even in real baseball), there can be signature significance even within (relatively) small samples.

Posted 4:51 p.m., September 3, 2003 (#29) - tangotiger
I was thinking a bit about this. The problem is that we use regression towards the mean on rate stats, when I'm not sure that's entirely accurate, especially when you have distributions such as this.

So, I propose the following, with an example. Say the league mean is .300 and your 100 PA player is a .950 player. The regression towards the mean is set at .700 for a player with 100 PAs.

Let's break out our ratios.
.300 = .300/.700 = .429
.950 = 19.000

.400 = .667

With a .400 player with 100 PA, we would normally do a regression as 70% towards .300, or .330. With our new-fangled ratio method, that would become
new ratio = .667 - (.667 - .429) * .700 = .500
new rate = .5 / (1+.5) = .333 (as opposed to our previous .330)

With your .950 player and 100 PA, that becomes
new ratio = 19 - (19-.429) * .700 = 6.00
new rate = 6 / (6+1) = .857

I don't know if that even makes mathematical sense, but I find that my trusty ratios always come through in the pinch.

(That 70% regression for a rate might translate to 73% for a ratio, or something.)

Posted 4:55 p.m., September 3, 2003 (#30) - tangotiger
Btw, your .980 player becomes a .936 player using this process. I think I may be onto something here. Maybe I should break out my stats books from 15 years ago.

Posted 6:45 p.m., September 3, 2003 (#31) - David Smyth
"I think I may be onto something here."

Well, I hope so. I've been making occasional posts about "signature significance" (as Bill James called it) on fanhome for a couple years and never got any "positive" response from the more statistically qualified people. Maybe my latest attempt at it here will satisfy my urge to know for sure (whatever the result may end up being).

Posted 12:14 p.m., September 8, 2003 (#32) - Joe Dimino(e-mail)
Back to the original post made by Tango to start this off . . .

You said that players are probably being overrated by 2 wins per year, but this wouldn't apply to pitchers would it? I've been calculating pennants added over on the Hall of Merit thread, and I'm going to adjust everyone down by 2 wins per full season. Basically I'll take their WARP1 or WARP3 (adjusted to 162 team games), and subtract PA / (lgPA / 9 * #tm) * 2 from everyone. Does that make sense? And should I do this for pitchers also? I'd think not, since you are saying the error is in their weighing fielding incorrectly.

Thanks for the help!

Posted 1:53 p.m., September 8, 2003 (#33) - tangotiger
Joe, that looks about right, though I can't comment on what the replacement level for fielding that is used for those golden years. Assuming it's set the same way, then yes.

As for pitchers, there's no double-counting going on, though they should have their own runs-per-win converter. I think you already do this.

Posted 3:49 p.m., September 9, 2003 (#34) - tangotiger (homepage)
Michael: what you are reporting about what Gary told you is inconsistent with what Clay is reporting at the above link. BP is, to the best that I can tell, double-counting the replacement level. This, according to the WARP-3, makes Loiaza a very viable MVP candidate.

My note to BP from last month was left unanswered, and therefore, I will report a "no comment" from them.

Posted 3:52 p.m., September 9, 2003 (#35) - tangotiger
To clarify, if WARP-3 did not double-count the replacement level for non-pitchers, it sees Loaiza as a viable MVP candidate.

Posted 4:25 p.m., September 9, 2003 (#36) - ColinM
I was late in finding this thread, but I want to say great job tangotiger on noticing this and letting BP know. I'm not surprised about Loiza. WARP-3 and even worse, Win Shares, really downgrade the contributions of top pitchers versus top position players. Most historical research I've done suggests that the top pitchers are roughly as valuable as the top position players. I would guess on average maybe 5 out of the top 15 players in a given year would be pitchers. Not that you guys have any reason to just take my word for this :)

Posted 2:55 p.m., September 11, 2003 (#37) - ColinM
I posted this in clutch hits but it seems to have some relevance over here also. In the thread we noticed that Randy Johnson strangely has more career WARP3 than Greg Maddux. I then ran across this in the BP glossary:

XIP
Adjusted Innings Pitched; used for the PRAA and PRAR statistics. There are two separate adjustments:

1) Decisions. Innings are redistributed among the members of the team to favor those who took part in more decisions (wins, losses, and saves) than their innings alone would lead you to expect. The main incentive was to do a better job recognizing the value of closers than a simple runs above average approach would permit. XIPA for the team, after this adjustment, will equal team innings. First, adjust the wins and saves; let X = (team wins) / (team wins + saves). Multiply that by individual (wins + saves) to get an adjusted win total. Add losses. Multiply by team innings divided by team wins and losses.

2) Pitcher/fielder share. When I do the pitch/field breakdown for individuals, one of the stats that gets separated is innings. If an individual pitcher has more pitcher-specific innings than an average pitcher with the same total innings would have, than the difference is added to his XIPA. If a pitcher has fewer than average, the difference is subtracted. This creates a deliberate bias in favor of pitchers who are more independent of their fielders (the strikeout pitchers, basically), and against those who are highly dependent on their defenses (the Tommy John types).

Ignoring #1 for the moment and just concentrating on #2, this is...well...terrible. I mean, if you want to use DIPS theory and try to figure out the quality of defense behind a pitcher, then sure, thats fine. But to give a pitcher credit over another just because he has more defense independant plays? This tells you nothing about the quality of the defense behind him.

Look at it this way. Say RJ and Maddux both save 60 runs above replacement. Do you give Maddux less credit because he accomplished it by not giving up homers and walks? In this case the fact that he has less BB and HR, less defense independant plays, is a GOOD thing. You don't give more to RJ for having more Ks.

If I'm wrong about how this works, then I apologize to BP. But it sure seems to fit the results and the description. I also found these numbers:

IP, ERA+, WARP3
Nolan Ryan 5386, 112, 130.2
Gaylord Perry 5350, 117, 113.3

Nolan Ryan has over 380 XIP in 1973 and 74. Looks to me like WARP3 for pitchers is just as unreliable as WARP3 for batters.

Posted 10:10 a.m., October 31, 2003 (#38) - tangotiger (homepage)
Clay says: ...an assumption that the ultimate replacement level team was the Cleveland Spiders of 1899, a combination of craptastic hitting, pitching, and fielding. That puts my "replacement level" player at a .130 win pct., significantly below the "freely available" threshold (which typically involes a .300-.350 win pct), but still above the "no contribution whatsoever" of Win Shares.

Is this reasonable? Is it reasonable that you can have crappy hitting AND crappy fielding from the same players at the MLB level, in today's day and age? Is this who you are trying to be better against?

The most reasonable baseline is that a MLB scrub non-pitcher is a bit over 2 wins / 162 GP worse than average (hitting and fielding). For pitchers, it's probably a bit under 3 wins / 27 full games. So a team of non-pitchers would be -2 x 9 = -18. A team of pitchers would be -3 x 6 = -18 wins as well. That's -36 wins from an average team of 81 wins, or 45 wins, or .278.

You can present data, based on your varying assumptions, that'll put the baseline at somewhere between .250 and .350 for a team. If you want to set the replacement level to .130, this would mean you have a team scoring 2.90 runs and allowing 7.75. I find that completely unreasonable.

***********

When the Tigers and the Mets and the Spiders get brought up as examples, I have to remind the readers about (a) the difference between a sample performance and a true talent level, as well as (b) the non-random distribution of talent among teams.

Taking that last one (b): there's some players on the 62 Mets and 03 Tigers that would not have seen the light of day on any other team.

As for (a): the stats of players are samples... SAMPLES.... OBSERVATIONS... of their true talent level. You simply can't take a player's stats and assume that that's representative of their true talent level, and therefore base a theoretical team from those stats.

If you've got a theoretical team of .300 talent, they will NOT play .300. They will play between .200 and .400. So, if you knew (which is impossible) that you had a team that is expected to win .300 over 1 million games, it's quite possible that they will play .200 over 162 games.

So, if you've got the 62 Mets and 03 Tigers or the 99 Spiders, you must, absolutely must, regress their performance to some degree to establish the true talent level of that team of players.

Posted 10:26 a.m., October 31, 2003 (#39) - tangotiger
Not to pick on Clay, since he's doing great work with translations, but his statement on Matsui:

...it sent me back to the drawing board with respect to Japanese translations, with Extra Special Attention paid to power. The result is a revised system specifically meant to deal with Japan, and not treating it like every other league in the States. If I had had these revisions in the spring, my forecast would have been more like .290/.375/.479 (22 HR) instead of the .290/.421/.567 (32 HR) we actually forecast - since his actual line was .287/.353/.435 (16 HR), that cuts more than half the error away.

PECOTA does the same thing, as just about every regression equation out there. You can't include your samples to establish an equation, and then use those samples to test against. All you are doing is best-fitting your samples, which is not necessarily predictive of data outside your samples.

If Clay did not include Matsui in his samples, then that's another story, and I'd have no problem with it... as long as the equations being developed had no knowledge of Matsui. If on the other hand, Matsui was included, then you can't talk about "cutting errors in half", since Matsui was part of the sample group you established the equations on.

As an example, John Jarvis shows, using regression, that the value of a double is .67 runs. This is laughable. And then, he shows the RMSE of all teams, and shows that a LWTS equation, with the .67 figure, comes out the best! Well, it was best-fitted to do so. The true test would be to best-fit say the 1974-1990 time period, and then test against the 1961-1973 time period and 1991-2003 time period.

For PECOTA, MLEs, and other translation systems, you should only use a certain percentage of the data, and then test it against the rest of the data.

(Forewarning: if you're going to comment that my comment is too "harsh" or I'm picking on anyone, then send me an email to that effect, and we can discuss it privately. I'm not going to debate this type of issue in an open forum any more.)

Posted 4:40 p.m., October 31, 2003 (#40) - David Smyth
---"As an example, John Jarvis shows, using regression, that the value of a double is .67 runs. This is laughable."

Come on, Tango. You know better than I do that is not what the .67 is intended to mean. People will read that and think that regression is bad, when all it really means is that regression must be interpreted properly. Regression is a useful tool in analysis.

Posted 5:03 p.m., October 31, 2003 (#41) - tangotiger
Well, John Jarvis then goes out and starts using that figure for other purposes.

As well, the regression value itself has a margin of error, which is ignored as well.

Posted 6:23 p.m., October 31, 2003 (#42) - Joe Dimino(e-mail)
I agree Tango. Clay says that the 0 FRAR player should be the worst regular in the league and on the verge of a position switch, but presumably, that guy can hit some AND you have a replacement on the bench that is a significantly better fielder.

I agree with everything you've said. Thanks for pointing this out.

Posted 6:23 p.m., October 31, 2003 (#43) - David Smyth
Looking at the actual Davenport stuff, Tango is right on. Clay essentially is using the "copout" of saying that you can define replacement anyway you want, and as long as you handle the math properly, none is inherently better than another. That's just another of the "everything is relative", politically correct type of nonsense we are subjected to all the time in America nowadays. Radical Islamist homicide bombers are just "freedom fighters", no worse than the Americans of the Revolutionary War. Rap music is every bit as good as classical music; after all, music is subjective.

And on and on. Come on, Clay---admit you blew it with the super-low Spiders repl level. It is not the equal of the .375 concept, simply because it does not answer the relevant questions about modern baseball nearly as well.

Posted 4:07 p.m., November 4, 2003 (#44) - Silver King
What's the name of that fallacy where you take logic that applies in a certain situation and then slosh it all over society? Is David (or ?) an impartial or accurate arbiter of that which is--inherently obviously, sounds like--better? "Everything is relative" is a more dismissable version of "everything is context dependent" - to whom, for what purposes, under what conditions?

Anyway, thanks for the commentaries on Clay's chat comment; I was wonderin' about that.

I'm really curious about how far the new/revised/expanded/? player cards will go. They've promised for a few years now that they'll put up translations for every player season in history. That would be really fun. Also, Clay has mentioned before that he's revised his defensive measures. I expect that's forthcoming; I hope there'll be interesting article(s) about his improvements.

Oh, and I'm _sure_ looking forward to whatever MGL and Tango are writing. It'd be delightful to have a lot of this stuff in clearly-worded (an MGL specialty, I think) all-in-one-place form. I'd happily subscribe to you guys' premium website. ;)

Posted 5:17 p.m., November 4, 2003 (#45) - tangotiger
We're slowing trudging along. The outline is all written, and the data is all pretty-well parsed for easy querying. The hard part is really managing our family (and for me work) lives with this project. And, I know that writing and presenting the report will take up 80% of the time. As for a "pay-for" website, we've discussed it, but still not sure yet if/when to implement that.

Posted 7:09 p.m., November 4, 2003 (#46) - David Smyth
Silver King, no, I don't try to set myself up as an especially important judge of what is "better". I am probably wrong my share of the time. But that doesn't mean that everything is created equal. The way to judge which repl level is better is to determine which one is best capable of answering the most relevant and important questions about modern baseball. And I believe that such a determination can be done with enough objectiveness that we CAN say that the .375 level or thereabouts is "better" than the .130 level or whatever it was.

The reluctance to use our brains to make logical assessments about what is "better" is one of the problems of our society. Eating dogshit is not the equal of eating filet mignon, no matter how many people say that it's simply a matter of taste.

Posted 10:28 p.m., November 4, 2003 (#47) - RossCW
Eating dogshit is not the equal of eating filet mignon

'Tis true, there are many people in the world with a clear understand of the difference between right and wrong who would tell you that eating dogshit is, morally, vastly superior.

I am probably wrong my share of the time. But that doesn't mean that everything is created equal.

Certainly not all shares are equal - you must be a lion.

Posted 11:44 p.m., November 4, 2003 (#48) - Silver King
I agree with your point, David, about replacement level.

Our logical assessments of things beyond baseball are in far more peril of not taking nearly enough into account, while we tend to leap to assume we're making a clear-cut correct value judgment. As RossCW neatly caught, even the chosen-for-its-utter-obviousness steak/poop dichotomy misses a serious consideration.

We often don't get enough information (little access, or it's buried under 'mainstream' and trivia) about other perspectives to know if we're making a viable assessment of them. We should always be working on this, and re-forming our necessarily provisional assessments.

Tango, as long as you're making a book, can you put stuff in the appendices like SLWTS player reports for the past decade, etc? =)

Posted 8:41 a.m., November 5, 2003 (#49) - David Smyth
---"'Tis true, there are many people in the world with a clear understand of the difference between right and wrong who would tell you that eating dogshit is, morally, vastly superior."

Well, you know how I am sure that eating filet is superior? Because if you eat nothing but pit bull turds, you will die. So eating steak is "better", no matter what some confused people think is "morally" superior.

And yes, Silver, good information is critical, but not really necessary to form one's core principles.