New York Met's Official 2005 Thread

[GForce]

Originally Posted by Anonymous Ghost

2005 PECOTA 5-year Forecast (weighted mean):
2005 .271
2006 .282
2007 .287
2008 .287
2009 .292

That's a little above-average for a 1B or corner outfielder, but its hardly the makings of a future perennial All Star. If he were still at 2B, he could be a future Jeff Kent. But further down the defensive spectrum, the marginal value of that bat is rather minimal.

BTW, with a win today, the Mets are on pace for 46 wins.

projections 5 years down the line for a guy who hasn't played any significant time in the majors don't really matter too much, IMO. Of course, nobody said he was a future perennial All-Star anyway, but that's neither here nor there.

The Yanks are currently on pace for 69 wins, Kaz Matsui in on pace for 138 RBI, and Marlon Anderson should hit a nice .667 over about 140 ABs for the year. So, SO informative this "pace" stuff is.

GH

[Anonymous Ghost]

Actually a career projection for a 23 year-old with as much of a minor league track record as Diaz (1715 ABs) is pretty reliable when using a system as sophisticated as PECOTA. Not so much when the projections are WAGs. Had he just had just half of season in Rookie ball plus a college career then your point might have some merit.

[GForce]

Originally Posted by Anonymous Ghost

Actually a career projection for a 23 year-old with as much of a minor league track record as Diaz (1715 ABs) is pretty reliable when using a system as sophisticated as PECOTA. Not so much when the projections are WAGs. Had he just had just half of season in Rookie ball plus a college career then your point might have some merit.

Actually, my point has significant merit. PECOTA's reliability is OK relative to other projection tools, but by no means "reliable." Being the best tool available now doesn't make it reliable, it just makes it more reliable than alternatives. A system that assumes all players age the same and has no real system to judge playing time, is far from "reliable." But then again, you knew all that I'm sure.

GH

[Anonymous Ghost]

Originally Posted by GForce

Actually, my point has significant merit. PECOTA's reliability is OK relative to other projection tools, but by no means "reliable." Being the best tool available now doesn't make it reliable, it just makes it more reliable than alternatives. A system that assumes all players age the same and has no real system to judge playing time, is far from "reliable." But then again, you knew all that I'm sure.

From the highligthed part, its clear that you don't understand how PECOTA works. First, it most certainly does not assume that all players age the same. Rather, its premise is that similar players (determined by past performance, position, physical attributes, etc) age similarly. The math behind it might be a bit over your head, but you should be able to grasp the fundamental concepts behind it.

Second, PECOTA endeavors to project future true talent level, not accurately predict exact counting statistics. Predicting playing time is wholly irrelevent for this purpose, except to the degree that fewer PAs mean wider error bars on the EQAs.

[GForce]

Originally Posted by Anonymous Ghost

From the highligthed part, its clear that you don't understand how PECOTA works. First, it most certainly does not assume that all players age the same. Rather, its premise is that similar players (determined by past performance, position, physical attributes, etc) age similarly. The math behind it might be a bit over your head, but you should be able to grasp the fundamental concepts behind it.

Second, PECOTA endeavors to project future true talent level, not accurately predict exact counting statistics. Predicting playing time is wholly irrelevent for this purpose, except to the degree that fewer PAs mean wider error bars on the EQAs.

Trust me, nothing is over my head (particularly nothing you could even come close to grasping). The premise that a current player will age the same as an entirely different player is fundamentally flawed. This is not a new criticism of PECOTA by any stretch, but again, I'm sure you knew that.

Of course, you've fabricated an argument here anyway, as nobody said Diaz would be a perennial all-star as you brought up. I merely said he could certainly hit, and the info you provided did nothing to disprove that. But I'm sure you knew that, too.

The fact is, even the guys who have done PECOTA have said for all intents and purposes, "Don't look at the result, look at the method." Which kind of throws results, while not out the window by any means, certainly a good distance from "reliable."

Sorry if I went over your head!

GH

[Anonymous Ghost]

Again, you misstate one of the basic premises of PECOTA: its not projection based on all players, but rather projects based upon the historical performance of a subset that share similar characteristics. In other words, its a scientific approach, not WAGs by some egomaniacal 30 year-old loser plugging numbers into some arbitrarily defined, ad hoc Excel spreadsheets at 4am in his boxers.

Silver certainly isn't the first person to do this type of research, but he's done the most extensive, publicly available (for a nominal subscription) projection set. He focuses on popularizing his method more than his results because he understands a basic statistical concept that you seem to have difficulty with: variance around a sample mean, some of which are the result of exogenuous factors beyond a player's control (suboptimal usage/playing time, freak injuries, etc). And again, the point of PECOTA is not to project exact "counting" stats, its to sift through the noise and properly try to gauge a player's true talent, present and future.

As for Diaz in particular, you said "Hitting isn't the question with this guy. He can hit, but his defense is...well....he can hit" which I interpreted as meaning you believe him to be a star hitter with subpar defense. The evidence--PECOTA or otherwise--pretty clearly indicate that he's not. That's an honest misunderstanding for which your imprecise language is as much at fault as my assumption. Let's move on and focus on your imprecise understanding of the application multivariable regression. If you drop the smug attitude going forward, I'll be happy to answer any questions that you may have.

[GForce]

Originally Posted by Anonymous Ghost

Again, you misstate one of the basic premises of PECOTA: its not projection based on all players

Which, of course, was never said. But you are doing very well as misrepresentation, so I'll let you carry on with it. You're likely good at very few things, but for misrepresentation and distortion, you're A-1 Aces, baby.

Silver certainly isn't the first person to do this type of research, but he's done the most extensive, publicly available (for a nominal subscription) projection set. He focuses on popularizing his method more than his results because he understands a basic statistical concept that you seem to have difficulty with: variance around a sample mean, some of which are the result of exogenuous factors beyond a player's control (suboptimal usage/playing time, freak injuries, etc). And again, the point of PECOTA is not to project exact "counting" stats, its to sift through the noise and properly try to gauge a player's true talent, present and future.

You can continue to blather on nonsensically about my not understanding...I understand PECOTA fine, likely far better than you do. It's fascinating you spit out a 5-year forecast, say "a career projection...is pretty reliable when using a system as sophisticated as PECOTA"...then you say projections aren't the point when I confront you on the questionable accuracy of those projections. In essence, you ran your entire argument right into the ground. Well done!

Anything that projects ANYTHING is subject to being measured by the accuracy of its results. THAT is a basic concept that you apparently fail to recognize. Again, PECOTA is the best tool right now. I don't argue that point in the slightest, and Nat's work is commendable. And it will improve because he's dedicated to it.

As for Diaz in particular, you said "Hitting isn't the question with this guy. He can hit, but his defense is...well....he can hit" which I interpreted as meaning you believe him to be a star hitter with subpar defense. The evidence--PECOTA or otherwise--pretty clearly indicate that he's not. That's an honest misunderstanding for which your imprecise language is as much at fault as my assumption. Let's move on and focus on your imprecise understanding of the application multivariable regression. If you drop the smug attitude going forward, I'll be happy to answer any questions that you may have.

Your asinine assumption is your own fault. I said he could hit, you noted the projection (which you later wanted to disregard) put him above average at 1B or corner outfield. Above average is certainly positive, though not glowing...kind of like saying "he can hit" is positive but not calling him a "perennial all-star." You made the leap to try to create an argument that has fallen on its face. Not my fault.

Funny...after your "by no means a perennial all-star" comment, you're the one who drew the comparisons to Jeff Kent...a 4-time All Star.

We've now deviated sufficiently from the thread. If you care to, feel free to PM me and I'll continue your education free of charge.

GH

[Joshv02]

Originally Posted by Anonymous Ghost

not WAGs by some [egomaniacal] 30 year-old [loser] plugging numbers into some arbitrarily defined, ad hoc Excel spreadsheets at 4am in his boxers.

Hey! That is me (umm, with my edits?)!

Originally Posted by Anonymous Ghost

If you drop the smug attitude going forward, I'll be happy to answer any questions that you may have.

. . .

Originally Posted by GForce

Your asinine assumption is your own fault.

Doh! Guess that won't happen, AG.

[ctorg]

[marioh]

I watched the replay last night since I couldn't watch the game live, and did Diaz look bad on that one fly ball. But he made up for it the next inning. The RF's on both teams looked terrible at times, defensively, yesterday.

[GForce]

Originally Posted by Joshv02

Doh! Guess that won't happen, AG.

And why should it? He's the one who started personal insults with me, not the other way around. But I'm sure you realized that.

Anyway, time to get the thread back on topic...

GH

[Anonymous Ghost]

Originally Posted by GForce

Which, of course, was never said. But you are doing very well as misrepresentation, so I'll let you carry on with it. You're likely good at very few things, but for misrepresentation and distortion, you're A-1 Aces, baby.

LOL, who's misrepresenting points here? Let's take a look:

then you say projections aren't the point when I confront you on the questionable accuracy of those projections. In essence, you ran your entire argument right into the ground. Well done!

No, I said that PECOTA doesn't endeavor to seek to predict specific statistical performance, but estimate present and predict future project true talent. PECOTA isn't reliable to tell you how many of whatever counting stats a player will hit in 2006 *because* his playing time is subject to so many exogenuous factors. But it can give you a pretty good idea of what his true talent will be, which Silver summarizes with EQA, from which it is possible to make a prediction as to whether a player figures to be an elite hitter, a league-average hitter for his position (more on this point later), a replacement-level hitter, or worse.

Anything that projects ANYTHING is subject to being measured by the accuracy of its results. THAT is a basic concept that you apparently fail to recognize.

I'll refer you to the above paragraph and what I actually wrote earlier (try re-reading, because you didn't get it the first time). PECOTA accurately forecasts a player a little better than 65% of the time in single-year forecasts (however, that doesn't take into account the many improvements that Silver has implemented since the introduction). We don't have data yet to fully evaluate his 5 year projections (it hasn't been around that long), but Silver has compared historical players' forecasts to what they actually did and found there to be a statistically significant relationship between what PECOTA would have projected versus actual result.

You're making a weak attempt at turning this conversation from the actual topic to one of semantics: "what qualifies as 'reliable?'" If you define "reliable" as 95% accurate, then PECOTA (along with basically every other forecasting model devised by man) fails. If you relax your criterion to a more realistic expectation (~50%, for example), then PECOTA performs quite well.

Again, PECOTA is the best tool right now. I don't argue that point in the slightest, and Nat's work is commendable. And it will improve because he's dedicated to it.

This is the first thing in your post that actually has any basis in fact. Good job.

Your asinine assumption is your own fault. I said he could hit, you noted the projection (which you later wanted to disregard) put him above average at 1B or corner outfield. Above average is certainly positive, though not glowing...kind of like saying "he can hit" is positive but not calling him a "perennial all-star." You made the leap to try to create an argument that has fallen on its face. Not my fault.

I said that he looks to be a slightly-above average *hitting* 1B/OF. I haven't deviated from that position; however, as someone who actually has some idea of how to properly evaluate the marginal value of players, I understand that a slightly-above average hitter with very poor defense is not a special player. If he's worth 25 VORP on offense (approximately what a .280 EQA RF is worth) but gives 10 runs back on defense plus a few on the basepaths, he's a pretty fungible commodity. He's above replacement-level, but he's hardly someone to get excited about.

Funny...after your "by no means a perennial all-star" comment, you're the one who drew the comparisons to Jeff Kent...a 4-time All Star.

Sigh, what I actually wrote is that had he stayed at 2B, he could be comparable to Jeff Kent. He didn't. By moving down the defensive spectrum (let me know if you need an explanation of this term), his marginal value diminishes greatly. At 1B/RF, he gets lost in the crowd when just looking at his hitting, and when taking into account his defense and baserunning, he's below average on the whole.

We haven't deviated from the thread. Pivotal to any discussion about the 2005 Mets is how their performance compares to the expectations of fans. After reading your posts, its become clear to me that the casual Met fan has a very provincial understanding of how to apply basic statistics and sabermetrics to player evaluation. I'm happy to have the chance to distill them into digestable amounts for your consumption.


So, to review, here are the core ideas that you should try to digest:
1) There's a difference between a player's actual performance and true talent. This is due to something called "sampling error." Sampling error can be reduced through an increase in the size of the sample (ie, more PAs) and a proper understanding that random variation doesn't in itself invalidate the true talent mean. Why? Because the true talent mean is based on a player's entire history not just a single season, hence a greater sample size.

2) With any statistically related, one's definition of "reliable" isn't 100% accuracy; rather, its demonstrating a statistically significant relationship. I don't want to overwhelm you, so we'll save hypothesis testing for another day. Suffice to say, "reliable" means being right more often than not. Evaluations of PECOTA (Silver's as well as others) show that's the case.

3) A player's marginal value is the sum of his value in all areas: hitting, defense, and baserunning. Also, a run earned on defense is worth the same as one earned on offense (actually, a run on defense is worth a little bit more because it represents a more significant change in the scoring environment--but we can save that minor distinction for another day). If Diaz is an above average hitter but a poor defender and baserunner, then in theory overall he could be a below-average player. In application, our best estimates of his true talent in those three areas show that to be the case.

4) Finally, about 20-25 years ago Bill James introduced the concept of the defensive spectrum. Essentially, his hypothesis is that because some positions are harder to play, the average hitting ability from position to position can vary because there are more hitters who can play the easier defensive positions. Thus, a .280 EQA 2B is worth a lot more than a .280 EQA 1B, assuming that their respective defense and baserunning values are worth 0 runs (ie, league-average).

[Melo]

If not for the bomb that broke the Cubbies back and Cameron's injury Diaz wouldn't have even made the club. It was after that homer that he began to be referred to as "Baby Manny". I think he'll hit in the bigs but not enough for a corner outfielder. I think he'll improve as an outfielder, he takes bad angles and misses the cutoff but that will improve with experience. It sounds like I may be down on him but I like him and hope he can develop. I don't see why he couldn't be a .280/15-20 hr/75-90 rbi guy. The question is can that offense be supplemented from a traditionally lesser offensive position.

[Melo]

Well, the Mets pitching looked mighty good last night, allowing zero runs in 11 innings against the 'stros. Ishii only walked 3 and looked very sharp. He was confident in his curve and got some k's with it, even on a full count with Bagwell at the plate in an RBI situation. Carlos Hernandez continued to look sharp. Looper got out of his last inning but something looks wrong. Keith Hernandez mentioned it in the telecast, noting that his fastball lacks zip. He struggled and got the 3rd out with the bases packed. His fastball was clocked at 86? This bears closer scrutiny.

[ctorg]

Three wins in a row now, eh? Pretty streaky.

I didn't watch the game, but Looper just inspires zero confidence in me as a closer. I like closers who strike out a lot of guys, because to me, a closer needs to be able to come in with guys on second and third and no outs and not allow anyone to score. Sure, a groundball is usually good there, but it leaves room for error.

[marioh]

Nice piece of hitting by Reyes last night in that game winning AB. He actually showed, *gasp*, patience at the plate and waited for a fat pitch to hammer back through the box.
A very nice outing for Ishii as well, matching the Rocket's great pitching performance.
Looper was shaky, again. It's also Roberto Hernandez Melo, unless the Mets have more than 1 Hernandez pitching for them. Roberto has looked sharp this year, not bad for a guy who's around 40. I think I'd have more faith with the 40 year old former closer than Looper at this point in time.

[lgkeeper]

Mets record before my son was born: 0-5.

Mets record since my son was born: 3-0.

My son is named after Gary Carter.

Coincidence? I think not.

[marioh]

Originally Posted by lgkeeper

Mets record before my son was born: 0-5.

Mets record since my son was born: 3-0.

My son is named after Gary Carter.

Coincidence? I think not.

You need to start having more kids.

[GForce]

Would've been tough to lose a game like last night where Ishii matches Clemens. Glad they scraped across the run.

Reyes won't hit .360 for the year, though...the kid is electric, but he needs to learn some patience.

As for Looper, he's certainly looking shaky as ctorg said, and his velocity is well down. Hopefully not a sign of anything too alarming, because as well as Roberto is throwing right now I really have no desire to see him as my closer over the duration.

GH

[Melo]

Originally Posted by marioh

It's also Roberto Hernandez Melo, unless the Mets have more than 1 Hernandez pitching for them.

oops, thanks for that

Grats on your son, lgkeeper. I guess this means we'll win out thru the series?