In my first two parts of this series, I looked at the stats to be examined when evaluating hitters and projecting Runs and RBIs. Today, I am going to take a look at projecting plate appearanes (PAs) and stolen bases (SBs).
With projections, I find the most error and room for improvement is with playing time. An owner needs to have a hitter playing and accumulating as many PAs as possible. The more PAs, the more counting stats they can hope to generate.
In order to get an accurate projection, an owner needs the expected PAs per game from a player. First start with the number of games each batter is expected to start. If it is a regular player, 150+ games is normal. Part timers and those that are injury prone, less. I feel that before spring training, every player’s projected talent/stats should be known. The key from then on is to concentrate on figuring out playing time (injuries, batting position, platoon roles, etc). Once an idea of the numbers of games is found, determining the number of PAs is fairly easy.
A team’s OBP predicts the average PAs per game decently (r-sqaured > 0.9). Here are the projected number of PAs per game per player knowing the team’s OBP:
I will use Ryan Howard, sorry for the jinx last week, again as an example. The Phillies had a 0.321 OBP in 2011. Howard should expect ~4.23 PA per game. He started 152 games, so he would be projected to have 617 PA (actual = 644).
The above values for OBP are for the average hitter in a lineup or the 5th spot. Another adjustment for PA is the batting order. Those hitters that bat first will have more PA. Here are the average number of PA gained over a 162 game season depending on the batting order position:
Since Ryan played in 152 games, his total would need to be adjusted for 16 more PA or 633 total. Without taking into account a few pinch hit ABs on his days off, the total PAs are close (633 vs 644). Again, the key is to get a good idea of the number of games to be played.
SB predictions are a mess. I just found that there was just too many variables to make it simple (speed, attempts, chances, team and manager philosophy, position in the lineup, etc). I have looked at too many different methods and have found a just few different points that were worth while.
1. Players will continue to steal no matter if they are good at it or not (see B.J. Upton). Past attempts are a great indicator or future attempts.
2. If the player is fast just doesn’t matter (see Jeff Francoeur). It just matters if they think they can steal bases.
3. The more the PA, the more the attempts. This is just looking at more opportunities. Players on good teams or at the top of the lineup will steal more because of the increase PA.
After way too much head scratching, I will just use SB attempts (SB+SC) per PA. I chose PA instead of opportunities at 1B because the values are about identical. I ran a correlation of (SB+CS)/PA vs SB/(1B+BB+HBP) and the r-squared ended up at 0.97. I am looking for a simple process.
Also, some information seems to indicate that players in the 1st spot will attempt steals more than those in other lineup positions. Some fellow Fangraph writers, that are better at database searches than me, are looking into the subject. I will let you know when they find any information.
With this procedure, I will begin to look at several different groups of players. I will start with 2012 2B valuations.
Print This Post