The debate in today's Rockpile inspired me to do some research into the area of lineup consistency, as it relates to team performance. The following is the research I have performed over the past few hours. Should more be desired, I can produce a followup analysis based upon community feedback at a later date.
I've always been frustrated by the assumed significance of lineup consistency, in regards particularly to the application of regularly changed baseball lineups to be symbol of poor management. It seems like all too easy an answer, a pedantic accusation of desperation to apply any blame to the construction of the team's offense independent of the individual performances themselves. It has become one of the most common criticisms of Jim Tracy during the 2010 season. This investigation will take a look at two small but significant sets of data: lineup consistency throughout Rockies history and lineup consistency in recent playoff teams. The intent of this is to highlight any noticeable patterns in team history or in the (recent) history of successful teams. If it were reasonable for me to include every playoff team of all time, I would, but I obviously can't. However, further depth can be attained if desired in the future.
My specific stance on the issue going into the study is that the consistency of a lineup is ultimately far more irrelevant to a team's success than is the actual skill and talent of the team, therefore I expect there to be very few patterns related to the two factors in any of the cases looked at here. This, of course, will be far from conclusive evidence for any perspective, but it should serve as an adequate reference point to contextualize the discussion, as these are numbers I find it hard to believe that anyone who takes a stance on this issue has looked at and considered when forming their opinion.
Two notes for this study:
- "Lineup", in this case, is being defined as the 8 NL players or 9 AL players that make up the positional alignment during any given game, and NOT the batting order. In other words, this is the definition Andrew Martin used for "lineup" during the discussion, and not my own.
- In case it wasn't clear from the above note, this does not include pitchers for obvious reasons.
Numbers and commentary after the jump.
PROCESS and DATA 1
I begin with presenting a table of each year in the history of the Colorado Rockies as it relates to the number of lineups used, in direct contrast with the winning percentage of the team.
- Winning %: Wins/Games
- Lineups: The number of individual positional alignments used during given season.
- Games: Games played during season.
- New Lineup Rate: Games/Lineups. This statistic is necessary only to create proportionality between seasons, as the Rockies have played several seasons that were not 162 games long for various reasons. It establishes "lineups per game", and can in other words, be interpreted as the average number of games that the team went during each season until a new lineup was used.
Bolded seasons indicate playoff seasons.
|Colorado Rockies||Winning %||Lineups|| Games
||New Lineup Rate|
As you can see, the Rockies have experience a wide variety of results in every category. Of most important note initially is the division of the New Lineup rate, which ranges from as low as 1.39 games between new lineups (this season so far) and as high as 3.11 games between new lineups (1996 season). The Rockies had a .512 winning percentage in 1996 and a .510 so far in 2010, and incredibly small distinction in between the incredibly large distinction in lineup consistency. Though the smallest imaginable sample size in this comparative study (two comparisons), this provides an example of my assumption. But is it merely an exception to a more consistent pattern that would contradict my stance?
Here is a table that sorts some of the results for easier contextualization. The NLR division is arbitrary, but the easiest way to sort it. Perhaps the two top divisions could be combined, but I've left it as is for now.
|New Lineup Rate|| Total
||Avg. %||Max||Min||Playoff App.|
|3 and higher||2||.523.5||535||512||1 (50%)|
|2.00 – 2.99||5||.512||568||453||2 (40%)|
|1.99 and lower||11||.453.9||510||414||0 (0%)|
Here are the easy deductions that can be made from this:
- The Rockies have tended to use inconsistent lineups far more regularly than the same consistent one during their history.
- The Rockies average winning percentage has improved at each level, though the 2 season total at 3 games or higher NLR is shifty.
- With the exception of the small sampled 3 and higher NLR, the ranges of best and worst results results between any given bracket is relatively large.
- The Rockies have never performed significantly well when using less than going at least 2 games between every new lineup.
- In the largest sample size, the team has never made the playoffs when using the team at a rate of 2 games or less per new lineup.
- A fascinating but meaningless little pattern: the minimum of 3 or more is equal to the average of 2-2.99, and the minimum of 2-2.99 is just 1 point off from the average of 1.99 or lower.
These results do suggest a historical pattern for the Rockies, which may inspire justification for your concern: the Rockies have never made the playoffs in a season of this type, and they are currently traveling at their lowest repeated defensive lineup ever, slightly below 2005's mark of 1.4.
However, two important issues to raise before jumping to fear:
1. Cause and effect: which is which here? Do teams have poor seasons because of inconsistent lineups, or do teams have inconsistent lineups because of poor seasons? While there will never be a clear answer that illuminates the easiest place in this circle to place the origin, the very appearance of this circular relationship is significant. A team can, and may often have an exaggerated NLR because of significant injuries or because of significant underperformance, both of which are likely to have a causal relationship with a disappointing winning percentage as well. Therefore, these results could make perfect sense even without a causal relationship of several lineups to poor seasons. I could not find an efficient way to compare rate of lineup chance to rate of injury and see if one or the other had a more profound effect if it was because of managerial decision or due to unavoidable roster constraints..
2. How do we know the Rockies' pattern fits the bill for what makes a team successful? In order to produce a small but reasonable (in regards to my time and energy and yours) comparative presentation, which will be highlighted by the study below.
PROCESS and DATA 2
Here I have produced a similar table, comparing the winning percentages and the NLR for every team to make the playoffs for the last three years.
|Recent Playoff Teams||Winning %||Lineups||Games||New Lineup Rate|
|2008 Red Sox||.586||108||162||1.5|
|2009 Red Sox||.586||84||162||1.92|
|2008 White Sox||.546||74||163||2.19|
|2007 Red Sox||.593||72||162||2.25|
Here we see an incredibly wide variety of successful teams. The average playoff team this season had a winning percentage of 575.58% and used just below 83 lineups. The Rockies were below the average winning percentage both playoff appearances, and also used fewer lineups than average both times. We see successful teams with far more NLR than the Rockies have had in their entire history, and also far less. Being that the average is around 83 and the median is at 81, it seems we have a somewhat even spready here (though the extremes at the "more lineups" end of the spectrum are larger).
So, over the past three years, has there been a distinct pattern between playoff success and numbers used? I interpret the results to suggest no. Looking at the winning percentage column, the results of each season seem to be spread out significantly, with both absolutely downright amazing teams and teams that barely sbuck their way into the playoffs scattered about alike. I see no pattern here.
Final Notes and Potential Improvements
One interesting pattern with the second chart I noticed was that the NL and AL teams, with some exceptions, tend to be bias themselves towards different ends of the spectrum. AL lineups seemed more likely to have more lineups than NL ones. This could be related to the presence of the AL DH, giving them one more position to be inconsistent, or it could just be altogether a coincidence.
I also find the observation that certain managers tend to be prone to different styles entirely to be fascinating. All three of Charlie Manuel's seasons on this are in the top 5 most consistent lineups, whereas two of Mike Scioscia's three on this list represent the 2 most extremely inconsistent.
The results of this study could suggest entirely different results if more time were put into archiving further playoff years, or perhaps giving a look at the comparison between winning percentages and consistency in the lineup across the entirety of MLB over a few given seasons.
In conclusions, these modern results, I feel, back up my position. But they do not provide long term suggestions of a relationship one way or the other.