Monday, February 02, 2009

Elias Updates

I've found all of the errors in last years rankings, and this year's should be much much closer, if not dead accurate. To summarize where some of the errors came from and what has been done to fix them:
  • Suspended/bereavement/restricted time does not count towards the DL adjustment. All of this time is ignored. In addition, players on those lists on August 31 are not included in the rankings. Notable among the bereaved players was Casey Kotchman, who was not scored in this year's rankings.
  • Duplicate Names: These are a pain. Rather than use the unique ID's that are available, I opted to index the players by name, trying to manually adjust. This led to incorrect rankings for Ramon Ramirez, Tony Pena, and Edgar Gonzalez. Luckily, the Chris Youngs and Ryan Brauns didn't create any problems. I've got Macros to help identify them by middle initial. This will be solved next year.
  • For whatever reason, when importing stats into Excel, I had problems with Jarrod Saltalamacchia's super long name and had to input his numbers by hand. His stats were in six different places (07 AL hitting/fielding, 08 AL hitting/fielding, 07 NL hitting/fielding) and I simply typed in his stats wrong in one of those places, throwing off his ranking.
  • I've already mentioned disabled list time in a previous post, but the DL adjustment, rather than being an estimation like last year, is perfect. In addition, I've also learned that not all retroactive dates are included in the press releases. The best way to calculate DL time is to find out that a player is on the DL and then go to the game logs to find out how many days he missed.
  • Relievers: I was furthest off on relievers. This is for a multitude of reasons.
  • First, the incorrect DL adjustment for players over 60 days masked a lot of problems. I had assumed this was the primary reason, and with so many players, they were bound to be off.
  • Second, I was using the wrong stats! I had correctly identified them and matched up the denominator in the previous rankings, but in setting up the spreadsheet, I left winning percentage included. Winning percentage is not used. This was inexcusable. For what it's worth, those who know the formula and read the Freep article probably got a good laugh, because I cited winning percentage as an area where Kyle Farnsworth could pick up Type B status. Whoops!
  • Finally, for some reason I was DL adjusting the IP/H category. This led to the overrating of some relievers with around 60 days of DL time. This is how Jamie Wright ended up as high as he did. Again, this was another boneheaded mistake, but it is caught in time for next year's rankings.
As for the upcoming season, the tentative plan is to have more frequently updated rankings with an embedded spreadsheet somewhere on this site. If I were to guess, I would say they would start around May.


Zeyes said...

Thanks for all your work on this! Quick question regarding "winning percentage is not used"...were you talking about relievers only, or all pitchers? Or did I misunderstand altogether?

Eddie B said...

Just relievers. It is used for starters.