PrOPS: 2005 and Beyond
My latest article is posted at The Hardball Times: PrOPS: 2005 and Beyond. The complete 2005 PrOPS are now up for the AL and NL. Also, check out the 2006 PrOPS Projections.
If you like PrOPS, you can read more about it in The Hardball Times Baseball Annual 2006.
Addendum: A nugget for Braves fans. Here are the PrOPS numbers for guys who played most of their 2005 season on the Braves. Many of these players had very few PAs, so keep that in mind. I am most concerned about the rookies, and I’m not sure how much stock I would put in the numbers. For example, Francoeur’s projection is higher than what I think he will do. Since I happened to watch most of his at-bats, I think much of his good contact was the product of terrible scouting. I’ll never forget the 0-2 change-up that dropped into the zone Odalis Perez threw him, which Jeff deposited in the bleachers. I was thinking, “how dumb can you be?” That won’t happen again this year, but he will prove his worth as a pro by his adjustments. I can’t say how he’ll do, and neither can PrOPS. The same should be said for ALL of the rookies.
First Last 2005 2006 PrOPS Predicted Chipper Jones 0.999 0.932 Andruw Jones 1.004 0.930 Kelly Johnson 0.834 0.869 Jeff Francoeur 0.816 0.867 Brian McCann 0.806 0.859 Adam LaRoche 0.835 0.836 Ryan Langerhans 0.787 0.823 Horacio Ramirez 0.787 0.818 Julio Franco 0.835 0.817 Wilson Betemit 0.742 0.787 Rafael Furcal 0.773 0.780 Marcus Giles 0.750 0.776 Kyle Davies 0.697 0.768 Eddie Perez 0.789 0.754 Johnny Estrada 0.725 0.753 Mike Hampton 0.723 0.743 Peter Orr 0.640 0.703 Raul Mondesi 0.660 0.691 Andy Marte 0.551 0.654 Brian Jordan 0.596 0.641 Brayan Pena 0.514 0.620 John Thomson 0.538 0.601 Jorge Sosa 0.489 0.576 John Smoltz 0.477 0.542 Tim Hudson 0.420 0.515
7 Responses “PrOPS: 2005 and Beyond”
Trackbacks/Pingbacks
-
Giving the Phillies prOPS
Could the Phillies offense be in decline for next year?
JC, I was wondering what concocting some sort of PrBPS stat would entail? It seems like that would provide a slightly different facet of analysis (especially, to my mind, in the case of Francoeur). I’m just curious, and I don’t possess the statistical wherewithal to explore this beyond a theoretical degree, but perhaps it’s something you would be interested in? It’d probably entail looking at BABIP extensively, now that I think about it. Just a thought.
I posted this at BTF… I don’t think I get the meaning of the confidence intervals.
–
I’m having trouble understanding these confidence intervals.
Take Ichiro versus Francoeur. In the past two years, Ichiro has had 1501 PA (!) to Francoeur’s 274. I would think this would make Francoeur’s future performance harder to predict, leading to a wider confidence interval. But Ichiro’s confidence interval is 102 points and Francoeur’s is 81 points.
Am I missing something here? Do the numbers just not represent what I think they represent?
Couldn’t it mean that those two years are so far out of whack with each other that it actually lessens the confidence than if you were just considering one year? Like, if you have additional data so different from what you would expect, wouldn’t you have to widen the range of values? Since Ichiro had a much different 2004 than 2005, the addition of the 2004 data actually increases the range of performance you could reasonably expect next year.
Interesting idea. Still, check out Torii Hunter.
2004: .271/.330/.475 in 569 PA, 105 OPS+
2005: .269/.337/.452 in 416 PA, 107 OPS+
His confidence interval is 87 points, still wider than Francoeur’s.
In any case, I’m not so sure that a guy who hit .100 in 2004 and .500 in 2005 should have a wider confidence interval than a guy who hit .300 for the last half of 2005. I mean, intuitively he should, but if you’re aggregating the statistics (with appropriate weights) then I don’t think it would end up that way.
The confidence intervals for guys with only one year of data are estimated in a different regression, which has smaller standard errors than the standard model. This is why the CIs for these guys with fewer observations have smaller confidence intervals.
Okay, I see. But it still seems strange to me.
It’s as if you were saying that throwing away all of Torii’s 2004 and half of his 2005 would lead to a more accurate estimate of his future performance. I know that’s not what you’re saying, and you certainly can’t fudge the standard error just because you think it looks too small, but isn’t this cause for concern?
I suppose such issues will iron themselves out once the system has more data to work with.