Originally Posted by Soylent Green:
I'm with you bballman! Numbers are useful, but by no means foolproof. They can and are tweaked regularly for effect. One of the ways "statmasters" try to intimidate and puff themselves up is by accusing anyone who disagrees with them of "not understanding the math". I have an MBA hanging on the office wall (that and $4.50 will get you a cup of coffee) so I think I get the math. Definitely wise to question the methods and motivations... and often the math, BTW... of these arguments. Everyone has a viewpoint and human nature is to fit information into place that will validate a given viewpoint.
WAR is a great example... there are at least three seperate "official" ways to calculate this incessantly used gem.
Taking it a step further, none of these various WAR formulae take into account the "value" of the games in which the data is mined... so a meaningless Mariners v Indians late season tilt from 1984 (let alone a dozen or more of these a year x 20 or so years) impacts a guy's career totals the same as, say, a Tigers v Royals game with playoff implications that same August. It's the difference between formulas and the field.
Not that I disagree with your premise, but other than altering the initial data, like changing a hit to an error or ignoring some data, how do the numbers get “tweaked”?
Personally, I have absolutely no use at all for WAR because I only work with amateur ball and its impossible to calculate war there. But I am interested in how you or anyone else would assign “value” to any individual game, let alone an individual PA. I’ve tried many different ways to do that over the years, but its simply too complicated at my level due to lack of data.
Here’s a question I haven’t asked in a while, but this may be a good time to ask it again. Assume a team loses its first and last games of the season to the same team and misses the playoffs by 1 game. Did that last game have more “implications” than the loss they suffered on opening day, or does that last game only seem to have more riding on it? 
Stats - No offense intended at you friend! Was just replying specific to what you had posted and in kind. I don't think we're talking about stats such as height, weight, lineage, etc. Also not talking about basic stats used by scouts in the 50s & 60s. I was referring to advanced (sabre)metrics being used today, and the back and forth of this vs. the more traditional modes of evaluating talent.
None taken.
Well, I guess there’s the rub. To me a statistic is a statistic no matter what the form or who does it. You may be right for all I know because I’m not privy to ML front offices to know exactly how talent is evaluated. But I think there’s an entirely different evaluating system used in deciding who to draft than there is in deciding whether or not to sign a veteran player to a $20+ per year long term contract. And that brings to mind another question.
Assume stats equivalent to ML stats were available for every amateur player for every game they played from the time they started playing kid pitch. Would it then be reasonable to use those stats in the evaluation process?
The Choo example offered by Jacjac above is a good one; I don't think Nolan Ryan and his guys would have valued Choo nearly so highly and paid the freight. Jon Daniels and his guys are more sabremetrics minded, we'll see how that goes for them. My feeling is that it was a bad contract for the team... too much $ and 1-2 too many years. But as Jacjac points out, there is a lot of "sabremetric" justification being offered by the team as to just how great a deal this was for the Rangers. Again, time will tell... but I have my doubts about the deal and about the metrics being trotted out to justify the deal. It's a good example of what bballman was saying about "making numbers tell the story you want told".
Well, you may look askance at Choo, but I’m a Tribe fan and would gladly swap any outfielder we have to get him back. I watched him play a lot of games and can only say he was a great asset to the Tribe and to the Reds. I think you’ll be very happy with the deal at the end of the year.
…Doesn't change the overall point though; numbers can and are tweaked and moveable
I’ll ask again, how are numbers “tweaked”?
And relied upon to the exclusion of other equally vital on-field competitive realities. As the saying goes, Figures don't lie, but liars sure can figure".
Does that mean because every factor isn’t included, no stat is valid or of any use?
To me, as long as a metric is computed using the same algorithm, and the data from every game is included, its pretty difficult to “tweak” the numbers.