31 March 2014

Paper Orioles 2014: Print Those Playoff Tickets!

A Paper Robin

Each year I use a variety of different win projection tools to keep track of the Orioles’ progress over the year.  The point of this is not to chain yourself to a certain methodology or have any belief in the whole “why even play the games” silly mantra people love to stick to men made of straw.  The point is to give you a rough estimate of where the team is and the general direction we may expect the team to go in. 

A few weeks ago we looked at general projection models and how actual results compared to them.  About 70% of all results come in a near 20 win window around those initial projections.  That may seem useless, but when you compare a team projected to win 85 games versus one who is projected t win 75 games, you wind up with historical evidence pointing toward the 85 win projection club taking their division about 12% of the time.  The 75 win projection club should take it about 2% of the time.  With that sort of difference in past performance, it is difficult to say that these projections are meaningless.

The Systems:
  • mJS (Jon Shepherd model) – This model looks forward to the remaining games on the schedule to develop weighted game scenarios based on opponent frequency and talent levels on different teams.  Talent levels are determined by using the ZiPS projection system and the performance tool on managers that we introduced last week.  Wins are determined by expressing talent level in terms of neutral team win expectations and then applying a log5 function for specific teams.  I do not change the ZIPS projection, so this model is only informed on the current season by accumulated wins and losses only.  That information does not affect the actual projection of remaining games.
  • PWE (Pythagorean Win Expectation) – This model uses runs scored and runs given from the season to date in order to project what will happen in the future.  The generalized equation is: Expected Winning Percentage: (runs scored^2) / (runs scored^2 + runs given^2)   
  • mBP (Baseball Prospectus‘ model) – This model is run by Baseball Prospectus.  They use their own projection model, PECOTA, and, I assume, actually account for in season performance.
  • Parcells (Bill Parcells model) – This model assumes that the season will continue on as determined by your current season record.  As Parcells once said, You are what your record says you are.  Of course, we know this not to be true as a definitive rule because wins are an imperfect measure of a team's ability, but it will be treated as such in this model.
  • fWAR – This model uses the accumulated team fWAR to project future results.
    Additionally, I will also include a league-wide season ending win expectation for the entire league as well.  Remember, this early in the season we will see some strange numbers at times.
 The PWE, Parcells, and fWAR models rely on data, so there will be no preseason model.  If you insist though, each should probably be expressed as 81-81 projected records because these models would see no difference between any teams.

Baseball Prospectus' Playoff Odds tool is quite down on the Orioles.  It sees the AL East as such:
Tampa Bay 87.8 wins
Boston 85.4
New York 82.1
Toronto 81.2
Baltimore 78.0
A common misunderstanding when people look at those numbers is thinking that those are predicted season ending tallies.  They are not.  They are season projections.  Specifically, they are the 50th percentile projections.  Half the team the team performs above that levels and half the time they perform below that level.  In other words, you would probably expect it to take at least 95 wins to take the AL East crown and 92 or 93 wins to secure a Wild Card slot.  PECOTA sees them as having a 13.6% chance to play in the post season.  Only six teams have a worst shot: Minnesota (7.8%), Houston (1.5%), Miami (4.5%), Chicago Cubs (8.1%), Arizona (12.8%), and Colorado (12.7%).

The mJS projection model is a bit more optimistic.
Boston 92.3 wins
Baltimore 85.8
Tampa Bay 83.6
New York 82.5
Toronto 76.9
That 50th percentile for the Orioles would get the club the play-in Wild Card game at home in Baltimore.  They would be hosting the Rays, but other teams in the conversation would be the Yankees, Royals (81.7), Angels (82.8), and Athletics (82.9).  Below is the projected playoff picture for each league:

American League
East: Boston Red Sox
Central: Detroit Tigers
West: Texas Rangers
WC1: Baltimore Orioles
WC2: Tampa Bay Rays

National League
East: Atlanta Braves
Central: St. Louis Cardinals
West: Los Angeles Dodgers
WC1: Washington Nationals
WC2: San Francisco Giants

Let's play ball!


Anonymous said...

Just a helpful thought in weighting to-date current year performance against preseason projections to form an updated projection. Property and casualty insurance actuaries often use a technique called Bornhuetter-Ferguson. If you aren't already familiar with it, I would bet you could find a way to apply it.

Jon Shepherd said...

In years past, I did similar manipulation of statistics. I might play around with it to see if I can do a simple team wide adjustment.