An interesting experiment I'm giving a shot this year is projecting win-loss records using the PECOTA projections.
I'm going to try it using Bill James' Pythagorean Expectation formula, The formula basically uses runs scored and runs allowed to estimate how many games a baseball team should have won, or should win.
Here is a summary of how I will do this. I provide the details on each step below.
- Estimate runs scored using projected lineup and PECOTA stats
- Estimate runs allowed using projected rotation and bullpen, and PECOTA runs allowed for that staff
- Use Pythagorean Expectations formula to estimate winning percentage.
- Apply that winning percentage to a 162-game schedule.
To get the runs scored, I'll use PECOTA hitting projections along with the very excellent Lineup Analysis tool at Baseball Musings to project runs scored for what should be the team's opening day lineup. I'll use last years lineups and common sense (according the the average baseball manager anyway) on what the lineup should be, not what the Lineup Analysis tool gives as the most efficient lineup. The tool gives Runs per game, so multiply that by 162, and that gives you a runs projection based on the projected stats.
For pitchers, my approach will be somewhat similar to the one taken here. For the pitcher player pool, I'll use the projected depth charts from RotoTimes. I'll use the starting rotation(including spot starter, if specified), closer, and 3 best projected bullpen arms. I'll then use the runs allowed projections and innings pitched projections from PECOTA to estimate total runs allowed. Because I'm not using the full bullpen, the numbers will have to be adjusted to equal a full season of pitching. I'll just use a multiplier on the runs to equal approximately how many innings the entire staff should pitch.
After I have the projected runs scored and allowed, I'll plug them into the Pythagorean Expectations formula. Note that I will be using an exponent of 1.82, because this is deemed to be more accurate than the traditional square. There are more accurate ways to get the exponent (see Pythagenpat), but since my source data is probably wildly off from actual seasonal performance, I'll just stick the the simple 1.82 .
The Pythagorean Expectations yields a winning percentage, so I'll just multiply that by 162 to get estimated wins by the team in question.
Before I get any emails attacking my process here, I realize that this is probably not the best way to estimate what the team records will be. I've tried to make the process somewhat simple to allow me to estimate all the teams in the majors. These will more than likely be off. Remember the process is building on projected player stats, which are then projected out even further. It doesn't take into account bench players, injuries, other bullpen arms, and the like. So don't bet anything on what I give here. But, I'm giving it a shot. It's just for fun.
I'll tackle a division per day until I finish. I have a final exam tonight, so I probably won't be able to post the first division until Tuesday night. Check back then, or subscribe to the feed.
