BP redoing WARP

[Baltimoron]

A realistic replacement level and a play-by-play based fielding system. Apparently they are trying to redo their recent WARP scores to get ready for them incorporated into the 09 Annual! Hooray!

Behind the scenes at BP, Clay Davenport has been hard at work revising the Wins Above Replacement Player system, our player valuation metric that covers the entirety of baseball history. Namely, he's incorporating two major changes; first, he's raising the replacement-level floor significantly beyond that of the bottom-of-the-barrel 1899 Cleveland Spiders or a current Double-A player to conform to a more modern definition of the major league replacement level, and second, he's adding a play-by-play based fielding component for the years where it is available.

Alas, the tail end of this research and development is taking place during the chaotic and often stressful period known around these parts as "book season," where our authors and editors are slaving away on player comments for our 2009 annual. The vanguard of Clay's fielding changes are geared towards the book, and as such, the fielding side of things for the years outside of its purview is not yet ready for prime time.

link

[DrungoHazewood]

A realistic replacement level and a play-by-play based fielding system. Apparently they are trying to redo their recent WARP scores to get ready for them incorporated into the 09 Annual! Hooray!

link

Rock on! That really is a huge thing. That puts BP's big numbers on par with most of the other cutting-edge metrics, and their search/lookup tools make this really nice. It should also have a side benefit of improving their derivitive metrics like MORP.

I've always liked BP because I think their writing is good, and many of their contributors are really sharp (like Clay and Joe Sheehan). Nice to see they've taken some constructive criticism from guys like Tango and improved their products.

[davearm]

I'd be curious to know whether they will attempt to back-cast the new WARP measure.

Moving the replacement level threshold should be no biggie, but unless they're using an existing fielding metric with available historical data, I don't see how they could.

Having a better WARP metric moving forward is obviously worthwhile in and of itself, but the best of all worlds would be to have at least a few years of actual data be updated as well.

[Baltimoron]

Having a better WARP metric moving forward is obviously worthwhile in and of itself, but the best of all worlds would be to have at least a few years of actual data be updated as well.

Definitely.

The quote "he's adding a play-by-play based fielding component for the years where it is available" makes me think they will have pbp data going back a few years.

The two companies who sell pbp fielding data are Stat, Inc (MGL used to use this for his UZR) and BIS (Baseball Information Systems). BIS is John Dewan's company and BIS data is the basis for the Fielding Bible, I believe the RZR and OOZ numbers at THT, for the UZR published at Fangraphs and David Pinto's PMR (among others).

Fangraphs looks like they publish UZR back thru 2002, so I'd hope they go at least that far back. I think that might be as far back as BIS goes for fielding pbp data, although the Stats, Inc data may go back further. But there are some suggestion that the BIS provides higher quality pbp data.

[DrungoHazewood]

I'd be curious to know whether they will attempt to back-cast the new WARP measure.

Moving the replacement level threshold should be no biggie, but unless they're using an existing fielding metric with available historical data, I don't see how they could.

Having a better WARP metric moving forward is obviously worthwhile in and of itself, but the best of all worlds would be to have at least a few years of actual data be updated as well.

From the article it seems that they're going to use the play-by-play metric for more recent seasons where the data exists, and use some version of Davenport's FRAR for prior seasons. It'll be interesting to see careers that span both methods to see if there are discontinuities or weird results. I'm sure there will be, but you do the best you can with the data you have.

[Frobby]

While I don't think even the PBP fielding metrics are always accurate in assessing who is a good fielder, they certainly will be better than Rate/Rate2, which I've found to be counter-instinctual and contrary to other sophisticated fielding stats about half the time. And since Rate/Rate2 is calculated in an undisclosed manner, there has never been a reason to trust it. So, this should be an improvement.

[Baltimoron]

While I don't think even the PBP fielding metrics are always accurate in assessing who is a good fielder, they certainly will be better than Rate/Rate2, which I've found to be counter-instinctual and contrary to other sophisticated fielding stats about half the time. And since Rate/Rate2 is calculated in an undisclosed manner, there has never been a reason to trust it. So, this should be an improvement.

PBP systems are only as good as the data on which they are based. Stats Inc had the market, but their former founder John Dewan went out and formed BIS and sold BIS video data. Many find the BIS data of better quality.

But Stats just blew it up again, acquiring Sport Vu

SportVu is an Israeli start up company that has developed a system that creates a multiple angle animated replay of the football pitch and gathers data from three fixed cameras, providing fans with statistics of their favorite players, including good or bad tackles and the distances players run.

They have a mobile app that recreates the game on an under laid soccer pitch and its from where they get work rate and other soccer stats.

In baseball, the big issue has been precision in recording not simply where players start or how hard the ball was or how long the ball was airborne, but more simple problems like "in what zone did a ball land," and rely on a human observer in the pressbox or watching available video angles making a notation on a screen as to where they think the ball went.

Maybe not so much any more.