I hope to answer questions like that in this post, as I try to explain exactly how to calculate the formula.
The first thing done is to establish the pool of players. This is seemingly mundane, but people were probably surprised to see Tony Giarratano included in last year's ratings, since he hadn't played since 2005. It appears as if Tony G made the rankings because he finished the season on the disabled list. The rankings appear to rank every player who ends the season on the active roster or disabled list of a team. The American League rankings include every player finishing the season on an American League team and the same for the National League.
A position must then be assigned to the player. This is determined by finding the position where the player played the most games in the past two years. The players are then grouped into the following positions
1B-2 (Also Includes National League pinch hitters and designated hitters)
DH-2 (American League only; also includes players who pinch hit or pinch run in most of their games, eg. J.J. Furmaniak)
If you are following at home, you should have position sets for each of the two leagues right now, and you should also be wondering about that number after each of the positions.
That number will represent the statistical set on which players at that position will be judged. These statistical sets came from mlbtraderumors.com. This would have been impossible to figure out without those statistics available.
1. C: PA, BA, OBP, HR, RBI, FPCT*, Assists*
2. 1B, OF, DH: PA, BA, OBP, HR, RBI
3. 2B, 3B, SS: PA, BA, OBP, HR, RBI, FPCT*, Total Chances*
4. SP: Total Games (Games Started + 0.5*Games Relieved), IP, Wins, Winning Pct., ERA, K's
5. RP: Total Games (Games Relieved +2*Games Started), IP**, Wins+Saves, IP/H, K/BB, ERA
*Defensive statistics are only accumulated at the player's designated position.
**Innings Pitched are given just 75% of the weight of the other relief categories.
Players are then ranked in each of those statistics within their position, based on their totals over the last two years. The leader of each category gets N-1 points, while the last place finisher gets 0 points, where N is the amount of players in the category. As an example, I am going to assign Elias Rankings to Willie Mays, Barry Bonds, Ted Williams, Hank Aaron, and Babe Ruth based on their career numbers. Here is how they rank, relative to eachother:
Now, here are the points they accumulated in each statistic:
Out of a possible 20 points, Ruth earned 12, Mays earned 6, Bonds earned 11, Williams netted 8, while Aaron ended up with 13. This means their Elias Rankings would be:
Hank Aaron: 65.000
Babe Ruth: 60.000
Barry Bonds: 55.000
Ted Williams: 40.000
Willie Mays: 30.000
There are a couple of things to notice. First, your Elias Rating is dependent on the people you are being ranked against. Bobby Kielty had a score of 31.646 last year, but he was in no way better than Willie Mays. This is because Mays has been compared to the all time greats, while Kielty was compared to outfielders, designated hitters, and first basemen who finished last season on an American League team. While that is an extreme example, it illustrates the point that a player can have an inflated Elias Ranking when players at his position in his league are particularly weak in a given season. Second, there is no boost for blowing out a category. I've already alluded to Alex Rodriguez last season, but he led all thirdbasemen in home runs over the 2006-2007 stretch. Even if he had hit 20 less home runs, his ranking wouldn't have changed. He still gets just one more point than the second place finisher. This is probably the biggest flaw of the system, with the use of fielding percentage being in close contention.
So the players at each position in each league are ranked in those categories. That is seemingly, simple, but there are plenty of questions to be asked. Looking at rate stats such as batting average and on base percentage, these are typically rounded when presented. I have learned that no stats are rounded in these rankings. While on TV, .3001 and .2999 both come out to .300 hitters, the .3001 hitter will finish above the .2999 hitter in the statistic. When players tie, they share the points in a category.
Back to the process, once each player gets a score, they are grouped together with the other positions in their group. For example, the shortstops, thirdbasemen, and secondbasemen are grouped together. The top 20% of players in each group are Type A free agents, while the next 20% are Type B free agents.
That brings us to the one thing I am stuck on. Players get an adjustment in their counting stats for time missed on the disabled list. I have thus far been unable to figure out how this is accounted for. If anybody wants to take a shot at that, please let me know and I'll send you the numbers I have. As soon as the disabled list adjustment is quantified, we should be able to figure out this formula.
I wanted to give thanks to Tim Dierkes of mlbtradrumors.com, Bill Ferris of detroittigersweblog.com, and Murray Chass of murraychass.com who each helped me to crack this mystery.