Value Rankings

[Chadwick]

I have been playing around with value rankings, tinkering with various formulas. Personally, I give a lot more credence to the depth of a player's value than to the width or breadth of it. I think a player's greatness is determined more by being able to say "at his best, he was this good" than by saying "he was this good over this period of time."

In other words, Sandy Koufax, Dick Allen and Vern Stephens were great ballplayers, Jake Beckley, Tony Perez and Jim Kaat were not.

Pick you value system - win shares, WARP3, TPR - but what do you think of this formula?

T5 = Combined Value of Top 5 Seasons
T10 = Combined Value of Top 10 Seasons
CV = Career Value
SV = Value based on this System

SV = (T5 + T10) + ([CV - T10] * .25)

Essentially, this counts a players' five best seasons twice and reduces everything he did outside his 10 best seasons to one-quarter its value.

I plugged win shares into this system and Sandy Koufax came out just ahead of Don Sutton. Ed Konetchy was significantly higher than Jake Beckely.

What are some classic high peak/low career vs. low peak/high career match ups to further test this with?

I know this is slanted towards peak value. (It's supposed to be.) I'm not at all sure it's as slanted as Bill James' formula used in his NHBA rankings, however.

[SABR Matt]

I think it's important when determining a career "Greatness" for a player to define exactly what it is you're trying to determine (this is true of any work in which you are creating a statistic...you need to sit down and specifically define what you mean to show).

I define "Greatness" as the player's potential to significantly enhance the fortunes of the teams for which he plays. What this breaks down to in my mind is as follows:

Teams require players who can be counted on for a good deal of production over a large period of years. The explosive peaks will be talked about later, but every championship calibar team begins with "the rock"...the three to five good solid players that can be built around. Paul O'Neill, Andy Pettitte, Mariano Rivera, Bernie Williams, Derek Jeter, Tino Martinez...the guys you can plug into your roster and feel comfortable with.

This to me doesn't translate into "consistancy" necessarily, because teams don't build their rosters in a way that accounts for fluctuations in annual performance or for the unlikely event of catastrophic injury...they build rosters making certain assumptions about what to expect from the players they sign/acquire over a period of a few years and fill in the gaps from year to year. This is why I don't like any methodology that uses the concept of the "peak window"...if a player happens to get injured in the middle of his prime, that does NOT make him less valuable/great when his career is viewed in retrospect. Players are generally inconsistant. It is exceedingly rare that a player gives you a constant production rate over many years. All players wobble some.

So this factor...the "building block" greatness, is IMHO based on career value. When the day is done, if a player created 250 wins, his teams acquired 250 wins they didn't have without him.

The second element is the "depth" factor. Because it is innately true that players performing further away from average have a much larger impact on the season to season performance of their teams. Barry Bonds turned a 70 win Giants team into perennially a 92-95 win genuine contender ALL BY HIMSELF...reserve the steroid converstaion for a later time and look at that on face value...single players having truly great seasons definitely make and break franchises (if Beltre hits 40 HRs last year, the Mariners probably win at least 80...not 69).

But again...I refuse to dip into "X # of years" as defining peak greatness. That's not even the point anyway. A player who has 4 truly brilliant seasons but they don't happen all at once...has contributed significantly to four contending teams most likely...it doesn't matter when those contributions occured.

I choose to measure peak performance by taking all wins created above twice the average win creation rate per unit of playing time. about 95% of all players in a given season will be between 50% and 200% of average production...if you're above 200%...that's an important statment...no matter when it occurs.

The third element I use has to do with efficiency. If a team can get 150 wins in 10 years from a player...or it can get those 150 wins in 20 years...it's better to get them in 10..both because that 10-year player produces something more like dominant peak performance (already measured above), and because in the remaining 10 years...that team has the opportunity to find another player who can produce some wins.

What is a fair assumption for the expected production of the other players who take the playing time efficient players do not? Many systems prorate performance to a standard career or a standard average season...I do not. Thart's not really an accurate representation of the value of being efficient. The time the player in question isn't playing the game can be assumed to be played by a LEAGUE AVERAGE player. So I prorate player performance to a standard career length by either (a) (if the player had more PA/defensive Games/IP than my standard career length) prorating down (career length of 10,000 PA, player creates 150 wins in 12,000 PA...prorated win total = 150 * (10/12) = 125 prorated wins) or (b) filling in missing playing time with the league average win-producing rate.

There's a fourth element I use called upward variabilitythat's designed to correct for the negative influences of a player's decline phase on his career scoring rate and therefore his prorated win total...

But this is what I mean when I say you should define what you're trying to find.