Wednesday, January 14, 2015

How Optimal Was The Cardinals 2014 Lineup

My simulator is a good tool for looking at optimal batting orders.  It can play tens of thousands and even millions (with a smaller set of games) of games while keeping the opponent static which can give you a good idea which lineups win more games.  Originally, taking an idea from the "straight arrows" over at The Book Blog, I decided to take the 2014 Dodgers most common lineup, that one can find over at the Baseball Reference website and compare how well that lineup does against tens of thousands of other possible lineup permutations.  Now I decided to move on to the St. Louis Cardinals mostly because I like doing this and the Cardinals have a very analytical friendly blog not one that spends endless hours discussing TV shows, taco trucks and their personal problems.  I did set some filters to cut down on the permutations, like I only looked at lineups where the pitcher hit 8th or 9th and where the best hitters didn't hit 8th or 9th and the worst hitters didn't hit 3rd or 4th etc...  I wanted to see how well one of the Cardinals most common lineups compared to what the simulator thought was the most optimal lineup.  Of course you can't expect any manager to be implementing the most optimal lineup but you also don't want him to be giving away fractions of wins each game that can add up over a 162 game season.  So down below, I list the Top 10 lineups along with the most common 2014 lineup in a table sorted by Wins/162 games.  I also list a table showing the frequency of where each player batted in each of the top 50 lineups.  I love the batting order frequency tables as they show you which players have a few dedicated positions in the order they should be hitting in and which players are more versatile in giving you an optimal lineup.  I used 2014 final season stats as the input projections for each hitter because that is for the most part similar to the data that the manager went by.  And I did all the simulations with the opposing team using a right handed starting pitcher.

Before we begin, here is the common Cardinals lineup that I compared against


                                                      Top 10 Lineups
RankLineupWins/162 Behind

And how did the Cardinals common lineup fair? Wel, it finished 2.07 wins/162 games worse than the top ranking lineup according to the simulator. Good thing that lineup wasn't actually used 162 times but it kind of gives you a general idea of how many wins the Cardinals manager may have been leaving off the table during the entire season if you trust the output of the simulator. When I did this for the Dodgers, Mattingly's most common lineup was only 0.84 wins/162 games off from the most optimal.

           Batting Order Frequency Table

Player by Player Analysis

Leadoff:  Matt Carpenter.  Matheny bats Carpenter leadoff and the simulator agrees that that is a good spot for him but he would be a little better off hitting second not first.  Good

Second:  Kolten Wong.  Matheny bats Wong second and the simulator doesn't think that is one of the top three places for him to hit.  It is not a totally glaring mistake as the simulator does have Wong hitting second in the 9th most optimal lineup but a better selection can be made here.  Below Average

Third:  Matt Holliday.  Matheny bats Holliday third and pretty much nails this one.  The simulator thinks that Holliday is a good third or fourth hitter in this lineup.  While Holliday does have some batting order versatility and could even slot into the second and fifth slots without deserving to be taken out behind the shed.  Great

Cleanup:  Allen Craig.  Matheny bats Craig cleanup and as you know Craig had an awful season.  To Matheny's credit he probably didn't know Craig was going to be this bad and stay this bad as long as he did but nonetheless batting Craig cleanup was not a good choice and if done enough times would be extremely costly.  The simulator liked Craig batting 8th, in front of the pitcher.  Bad

Fifth:  Yadier Molina:  Matheny bats Molina fifth and the simulator does not like him there at all.  The simulator likes the slow legged and power hitting catcher batting seventh or sixth at best where he can knock in runs and not get knocked in himself.  Bad

Sixth:  Matt Adams.  Matheny bats Adams sixth and the simulator thinks that is too low.  Mostly because the simulator already knows how bad Craig is.  The simulator likes Craig batting cleanup or in one of the other traditional power hitting slots (third or fifth).  Bad

Seventh:  Jhonny Peralta.  Matheny bats Peralta seventh and maybe he didn't know at the time he was building these lineups that Peralta would be one of his better hitters in 2014.  A look up above at the 'Batting Order Frequency' table and you really see how versatile Peralta is in this lineup.  Peralta hits leadoff in three of the four top optimal lineups and the simulator think the third, fourth and fifth slots are his best.  Good

Eighth:  Peter Bourjos.  Matheny puts the all glove not stick Bourjos eighth.  From above the simulator thinks that spot should be reserved for Allen Craig.  The simulator touts leadoff or ninth, which in and of itself is a second leadoff spot, as the most optimal placement for Bourjos.  Bourjos is leadoff in 2 of the top ten lineups and bats ninth in 4 of the top ten lineups.  There is often efficiency in batting the pitcher 8th if you have the right personnel to do it.  Bad

Ninth:  Pitcher.  The simulator has the pitcher batting 9th in 84% of the Top 50 lineups.  Good

So overall Matheny did not do a very good job with this lineup but to his credit the simulator is using end of season (or after the fact) input projections.  If Matheny was hitting Craig cleanup when it was obvious he wasn't very good then he is greatly to blame, otherwise he gets some slack.  How much slack is up for debate.

No comments: