differences in ops+ on baseball reference and the baseball encyclopedia

[fenrir]

anyone know why they are different?? in some cases pretty large. i was looking at the 1961 season for mantle and seen that his ops+ on baseball reference is 206 but in the new baseball encyclopedia it's 210. why the difference? is there one that is more accurate or does it not matter?

[Patriot]

In all likelihood, the difference stems from the different park adjustment methods. Forman uses the Bill James' "Willie Davis" method to translate the player's whole stat line, whereas Palmer in the Encyclopedia uses the run factor to directly adjust OPS+.

I prefer the Palmer method, but OPS+ should be used as an approximate method anyway, so I wouldn't lose any sleep over the discrepancies.

[AstrosFan]

Why do you prefer the Palmer method? The Baseball Reference method is designed that way because park factors are meant to apply to runs. BB-Ref is extrapolating the park adjusted on-base and slugging league figures from a runs created formula, while Palmer simply applies the park factor directly to the relative on-base and slugging figures. Wouldn't runs correlate better with runs created than with on-base or slugging, or even OPS?

[Patriot]

Why do you prefer the Palmer method? The Baseball Reference method is designed that way because park factors are meant to apply to runs. BB-Ref is extrapolating the park adjusted on-base and slugging league figures from a runs created formula, while Palmer simply applies the park factor directly to the relative on-base and slugging figures. Wouldn't runs correlate better with runs created than with on-base or slugging, or even OPS?

OPS+ is itself an approximation of runs created per out relative to the league average. It is not a particularly good one, but neither is runs created. Both conversions are based on the run park factor.

So, as far as I see it:

1) RC is not clearly superior to OPS+ as an estimator of the player's value in runs, so there is no good reason to prefer one to the other
2) Dividing OPS+ by PF is a lot simpler than the Willie Davis method, which given 1), is a good a reason to prefer one (Palmer) over the other.

[AstrosFan]

Okay, I thought you were choosing the Palmer method for accuracy reasons, but it appears you're going with it due to it's ease of calculation, and that it is not really less accurate than BB-Ref.

Players who play in pitcher's parks show up as better in the Palmer method, while players who play in hitter's parks show up as better in the Baseball Reference method, as far as I can tell.

[fenrir]

which one would you say is more accurate?

[RuthMayBond]

anyone know why they are different?? in some cases pretty large. i was looking at the 1961 season for mantle and seen that his ops+ on baseball reference is 206 but in the new baseball encyclopedia it's 210. Less than 2% difference for one season is "pretty large"?

[Tango Tiger]

Is it more accurate to say
TB = 1b+2b*2+3b*3+3.8 HR
or
TB = 1b+2b*2+3b*3+4.1 HR
?

That's what discussing the accuracy of OPS+ of two highly similar versions (each flawed at the core, by design) is at this point.

[Bench 5]

I happened to notice that the OPS+ figures cited on BBR for Barry Bonds changed from earlier this year (around May) to the present. I saved a copy of his stats from earlier this year and compared them to the current listing. Bonds was credited with a top season OPS+ of 275 in 2002. Now that same season is listed as 268. There have been other changes for several other years. I looked at a few old-timers like Ruth and Cobb and noticed some minor changes of one or two points in a couple years. Not nearly as many changes as the seasons below. I suppose they used updated park factors causing the changes?

Here's the comparison for Bonds.


Nov May
Year Age OPS+ OPS+ Diff
1986 21 103 103 0
1987 22 114 114 0
1988 23 148 147 1
1989 24 126 125 1
1990 25 170 170 0
1991 26 160 161 -1
1992 27 205 205 0
1993 28 204 206 -2
1994 29 183 182 1
1995 30 168 168 0
1996 31 188 187 1
1997 32 170 170 0
1998 33 178 177 1
1999 34 155 162 -7
2000 35 188 191 -3
2001 36 259 262 -3
2002 37 268 275 -7
2003 38 231 231 0
2004 39 263 260 3
2005 40 174 177 -3
2006 41 156 156 0
2007 42 170 NA NA
-18