I think those spot starts are not only based on last year's performances for the three of them, but the last several years.  Furthermore, today in the Tribune, this was written:  Piniella said he's looking for a long reliever, and the loser of the competition for the fifth spot involving Mark Prior, Wade Miller, Marshall and perhaps Neal Cotts could be that man.

There are obviously more choices to be had than pitchers for that 5th spot and there will be several losers.  It's distinctly possible, Al, that both Prior and Miller fail to claim that 5th spot out of spring training and Guzman has a great spring.  Then, with his injury history, it's not at all unlikely for him to go down after 6 starts leaving the 5th spot open for Prior, Miller, or Marshall.  

Prior's stuff last season was just plain bad.  He was worse than Jason Marquis and Glendon Rusch.  Injuries or not, he was bad.  It takes a leap of faith to believe he's capable of making 20 starts and actually earning that 5th starter's spot at this point.  

I'm just glad to hear that Lou isn't going to give him the job if he doesn't earn it.  I've been afraid that no matter how poorly he pitched, if he was healthy it was his job.  According to that article, Lou has said he's going with the best pitcher.  

5th starters for teams generally make about 20 starts and teams (healthy teams) have 3 to 4 guys who make a handful of spot starts.  

I don't think those projected playing time numbers are that far off.  They may end up being way off, but do you feel confident that either Prior, Miller, or Guzman will be good enough to earn that spot?  I sure don't.  

I assume we could agree that past performance is the best indicator of future performance.  Obviously you can't predict the future and in only one season there is a wide amount of variance, but that doesn't mean we can't use the tools to come to conclusions.  These projections are basically giving us a realistic idea of what a player is capable of if the 2007 season was played thousands of times...what you may call the player's true ability.  He may out-perform these projections, under-perform them, and he may be dead on.  I don't believe anybody who lists these or does these believes they're going to be accurate for one individual player.  As you know, the variance is too great in just one season.  

I also assume you are basing many of your opinions about possible future production from players on their past performance.  For example, I think you'd agree Juan Pierre isn't going to hit 50 home runs.  We know this for a couple of reasons.  The first being the years of data we have to look at where he's hardly hit any per season and his size, which is a more flawed way to take a look at future production (small guys have power too and some big guys don't).  

When you say that you think Lee is capable of playing quite well (somewhat similar to his 2005 numbers), you are basing that opinion on past performance.  You may not know it when you do it, but when you're talking about how good or how bad this team may be, your conclusions are the result of instantaneous projections based on your opinions, which are certainly based on past production.

So I guess my question is this:  How are these projection systems so flawed, but your own predictions/projections aren't?  You're doing the same thing without the mathematical analysis.  You're even considering age (you've mentioned how Murton may develop power as he ages--people who do these projections consider a natural age progression in their formulas).  You surely believe that as a player ages into his mid-30s that his production will likely continue to decline.  

You're unconsciously doing the same thing that you are saying you dislike.  Is it the relatively complex math being brought forward that you dislike?  I doubt that's it.  Batting average and slugging percentage was once complex math for this game and I'm sure you know that.  

Obviously these are flawed.  I don't think anybody has said they aren't.  I believe I read a study recently on some Angels blog that compared the various projections for hitters who had over 500 plate appearances.  PECOTA turned out to be the most reliable having ONLY a correlation coefficient of .51 (slightly higher I believe).  Not a very strong correlation at all.  But there is one and I'd think you'd agree that any available information that can make the team better is worthwhile information.  

Bill James won't even bother projecting pitchers; says it can't be done.  And the study showed that the correlation was much lower, but there was still a correlation (zips having correlation coefficient the highest with pitching, PECOTA 2nd).  

It's information.  It's flawed, but no more so than looking at a player's batting average and RBI's in one single season.  

Somehow, we'll figure out a way to lose 90 of them.