Buy The Book
Excellent Primer.
One point about the “age 29” thing: note that the player is 29 years old, and we are looking at his stats at age 26, 27, 28. So, when I say I don’t apply an aging factor at age 29, it’s because his talent level at age 29 is right in the middle of his talent level at age 26-28 (more or less).
Yeah, I don’t remember off hand. For some reason, I have “800” in my head. But, should be easy enough to reverse engineer, since I publish the r numbers for all the pitchers, right?
So, to make HAlladay come out to between .835 and .845, that would imply:
5790 * (1-.835)/.835 = 1144
5790 * (1-.845)/.845 = 758
Do that for a handful of pitchers (preferably those closer to r=.50, so that you don’t have such a big swing).
800 sounds right based on memory and it fits with above.
I didn’t realize (Remember anyway) you were triples alley. Great job in the article, you really laid it out well and explained it crisply.
I read the article and was happy to learn how Marcel works.
I was aware of the three parts (weighting, regression, aging) . . . I just thought there were other parts as well.
Thanks for the kind words!
When I plug in the numbers for various pitchers (pitchers whose r resides within .61 and .84), the constant appears to be somewhere around 1,065. When the estimate is off, it’s usually about a point.
Whoops, misread Tango #2.
Trying the constant as 1065 (1070 if you prefer) with pitchers around .5 yields an estimate usually within three points. Not perfect, but not bad.
May 25 12:40
Largest demonstration in Canadian history?
May 25 12:38
Do pitcher’s reach back for velocity when needed?
May 25 12:37
Chad Curtis
May 25 12:16
“Why Kickstarter works”
May 25 11:32
Howard Stern
May 25 11:26
Lack of hustle during a game
May 25 11:22
What sabermetrics is NOT
May 25 10:58
Rooting for laundry
May 25 02:38
NFLPA lawsuit against collusion
May 25 01:43
Neal Huntington’s best moves
Thanks for the link!
I’ve been meaning to ask- is there a different constant to determine pitcher reliability? I’m getting slightly lower figures when I tried to replicate the ‘10 Marcels. With Roy Halladay, for example, I get 5790/(5790+1200) = .83 (using the 3/2/1 weights, of course) rather than .84.