Baseball Reference and Baseball Prospectus, ugh.

[OleMissCub]

These two websites must be in cahoots with one another to try and share traffic.

BR has OPS and OPS+, but BP does not.

BP has the good metrics, like WARP3, but BR does not.

I always have to go to BOTH websites when I'm researching a player's stats.

[SABR Matt]

BP's metrics are proprietary...they won't even tell people how they're calcualted, let alone give that info out. BR has thus far not focused on sabermetric tools...I'm hoping Forman will be looking that way in the future.

[ChrisLDuncan]

Actually they tell you how EqA is calculated.

[SABR Matt]

well...except for the really basic ones like EqA and whatnot...right.

[Colin Wyers]

WARP itself is scaldingly easy to figure out; it's the components that are hard to figure. Even then, presumably BRAR is just EQR (which we have) above replacement; and BP's replacement level isn't exactly a state secret either. About the only part of the WARP chain that isn't public enough for an enterprising person to derive in an afternoon is the FRAA/FRAR, which are about this close to worthless unless you want to discuss ballplayers prior to 1958.

Same with VORP - enough of the MLVr chain of stats is public knowlege that you could derive it if you really wanted to.

The problem with BP's stats isn't so much that they're proprietary as they're the same stats BP was using back in 1996; Nate Silver and Dan Fox routinely invent new toys that are more useful than half of the stuff in BP's stats reports, and they get shuffled into the archive and lost. Things like PBP fielding metrics and BaseRuns and WPA get completely ignored; it's as though the past decade of sabermetric research outside of what BP themselves have done is completely ignored.

Basically it's similar to the problem that I have with Bill James these days - it's more like looking into a time machine and seeing what baseball research was like than it is looking at the latest research.

The biggest problem I have with all baseball websites is that none of them seem to give a complete picture of a player. If I want to look at batted ball data (ground ball rates, fly ball rates, line drive rates), about the only site willing to give me a fully gory breakdown is Fangraphs. Baseball Reference now gives me Batting Runs, which is hot, but I can't get any fielding metrics better than Range Factor, and their pitching stats aren't much better than ESPN's. I really like Hardball Times, but sometimes they baffle me too. FIP? xFIP? Nothing wrong with either of them per se, but... you're a computer. You mean to tell me you couldn't just give me DIPS? You'll give me GPA but not wOBA? I guarantee you that the CGI scripts are saving a lot less work by using "shortcut" stats than I am; the whole point of GPA and FIP opposed to more complicated measures is that I can figure them easily enough myself.

You can generally find what you want if you go digging far enough, and if nothing else you can find an Excel spreadsheet over on Rally's or Tango's website (or such) that has what you want. I just don't know why it has to be so hard.

[Ubiquitous]

The name is the same but WARP and the defensive metrics get constantly tinkered with.

I've never done it but I would imagine that duplicating WARP would be rather easy and I'm betting after some jiggering you can get it to WARP3 as well.

[KCGHOST]

The thing that aggravates me the most is that there is a single place you can go to see WARP3, Win Shares, OPS+/ERA+, and RCAA/RSAA. You go to BP for WARP3, THT for Win Shares, B-R for OPS+/ERA+ and CBE for RCAA. I suppose the reason is that each site fears loss of readers of their info is accessible elsewhere.

[Ubiquitous]

All except for WARP can be easily duplicated. It is simply just a matter of setting up the formula. I'm thinking THT doesn't do stuff like OPS+ and such because A) Baseball-reference already does it and B)they are selling their own stats and product.

[Pere]

BP's metrics are proprietary...they won't even tell people how they're calcualted,

Wow, I didn't realize this. I largely stopped reading them a few years ago, for other reasons.

How can you judge whether a metric is any good if you don't know what it is?

[Ubiquitous]

How can you judge whether a metric is any good if you don't know what it is?

Because they do tell you what it is. They have had numerous articles published about EqA and their fielding metrics. They have discussed how they turn runs into wins.

[SABR Matt]

They'll tell you (as I said) for some things...they won't tell you anything about fielding metrics and you really have to go digging to find anything about the pitching metrics.

[Ubiquitous]

The Fielding metrics have been published in their annual publications on numerous occasions.

[brett]

WARP itself is scaldingly easy to figure out; it's the components that are hard to figure.

I understand their BRAR and BRAA and such but it does not look like they convert it to games properly. It appears that they use around 10 rarp to equal 1 warp no matter what the run environment. Shouldn't it theoretically (at least in a linear model) be 2x the run environment, or the idealized adjusted run environment?

BBRef has changed their ballpark effects, to look at 3 year effects, but I'm not sure I agree because changes may be due to the ball used (like in '77 and '87?)

Also, BBPro sets the replacement level INCREDIBLY low, like a .250 WP hitter, but also a replacement level fielder. Nobody is that bad. It greatly overrates the long career player.

Also with BBRef, I don't agree with the methodology in averaging ERA+ and OPS+ by calculating the career average on base, and slugging, and league ob% and slugging. This is a mathematical fallacy of averaging proportions. Its like saying that 5/4 (1.25) and 3/2 (1.50) average out to 4/3 (1.33). It brings good pitchers down over the long run and keeps good hitters higher over the long run.

[Patriot]

I understand their BRAR and BRAA and such but it does not look like they convert it to games properly. It appears that they use around 10 rarp to equal 1 warp no matter what the run environment. Shouldn't it theoretically (at least in a linear model) be 2x the run environment, or the idealized adjusted run environment?

BP converts all of the stats into an "ideal league" before calculating WARP, and so the R/W stays constant or close to it. Take Yaz in '68 for example. He has 113 UEQR, which means they estimate that he actually created 113 runs. But his EQR is 135, which is the equivalent in the ideal league.

http://www.baseballprospectus.com/dt//1968BOS-A.php

BBRef has changed their ballpark effects, to look at 3 year effects, but I'm not sure I agree because changes may be due to the ball used (like in '77 and '87?)

By using more years in your PF, you are always muddling things up, as there surely are effects from year to year due to weather and the like, and the rest of the league changes as well. With that being said, though, I don't think there is any doubt that the tradeoff for a larger sample size is worth it. At the very least, if you do use just one year of data, you have to regress. Regression is always something to consider, but of course it is more important if your sample size is smaller.

Also, BBPro sets the replacement level INCREDIBLY low, like a .250 WP hitter, but also a replacement level fielder. Nobody is that bad. It greatly overrates the long career player.

I agree whole-heartedly.

[ChrisLDuncan]

The thing that aggravates me the most is that there is a single place you can go to see WARP3, Win Shares, OPS+/ERA+, and RCAA/RSAA. You go to BP for WARP3, THT for Win Shares, B-R for OPS+/ERA+ and CBE for RCAA. I suppose the reason is that each site fears loss of readers of their info is accessible elsewhere.

What is CBE?