Historical What-if’s . . .
In addition to simply replaying past seasons, one of the common appeals of a game like APBA Baseball is the ability to pit historical players from different eras against each other. What would result if Bob Gibson were on the mound glaring at Barry Bonds? How would legends like Ruth and Gehrig fare against the likes of Pedro Martinez or Greg Maddux?
On the surface, this seems like an easy enough thing to do. All an APBA replayer would have to do is acquire a few data discs featuring the desired teams from the game company, and then use Advanced Draft to set up your dream match-ups. If you aren’t too picky (APBA is a game after all) this alone can be pretty satisfying.
Alternately, individuals could import the desired teams, or players, directly from the Bill James Electronic Baseball Encyclopedia. These days it’s even possible to purchase season discs from specialty card makers like Skeetersoft. ((Skeetersoft even offers a series of normalized league champions – the Dynasties Vol 1 — Vol 10.))
My first Baseball for Windows “replay” of that sort was with the “Old Timers Vol. 1″ disc that was included with my original game purchase. I was very excited about playing the 1976 Cincinnati Reds against foes like the ’69 Mets, ’57 Braves, ’53 Dodgers, and ’34 Cardinals. I started using fatigue and injury options for realism, and, in the beginning it was all quite glorious.
But, when playing teams from different eras against each other, some problems come forward immediately, and, as a season progresses, others rear their ugly heads. For instance, teams from a 8 franchise league played a smaller number of games than modern teams, which folks may remember was the at the root of commissioner Ford Frick’s decision to place an asterisk after Roger Maris’ single season record of 61 home runs.
Having played a smaller number of games directly leads to teams from the the pre-expansion era being tired and worn out before the season ends when playing against teams from the modern era. There’s a utility at Mako Jo’s site that can correct for this particular problem, but this is just one of the many small issues that arises when attempting to play out any kind of serious cross era comparison.
The biggest issue that comes to bear when making cross era comparisons, or playing some kind of old time fantasy match-up baseball, is the extreme range of conditions that have occurred through the years. In 1930 the game was so slanted in favor of offense that the National League batting average exceeded .300 and league slugging percentage approached .450. In 1968 the game was so slanted in favor of pitching and defense that American League batters could only reach a .230 league batting average, and Carl Yastrezmski led the league by hitting .301.
Yaz’s performance in ’68 directly led to a great number of wins for his team, but his individual league leading offensive performance wasn’t far from what the average NL player produced in ’30, making direct statistical comparison of players from these disparate eras nearly impossible. Anyone who’s paid attention since the opening of Coor’s field and the entry of the Rockies into the Majors can also attest to the extreme effects a park can have on a team’s performance. Just as there is a huge difference in the way an era shapes a player’s statistics, their park can mask (or accentuate) their skills.
Fortunately, baseball statistics withstand various forms of manipulation pretty well, and with a few mathematical contortions, it’s possible to essentially put everyone on the same playing field, and compare players across eras and ballparks. Bill James showed one method in his New Bill James Historical Baseball Abstract, while the Davenport Translations featured by Baseball Prospectus offer another alternative. Miller Associates offered a “Normalization” option in player generation with their Wizard software, and a utility offered at the Mako Jo site, called a T3 Normalizer offers yet another method of putting players into a standardized context for APBA.
In 2002, an entire book appeared on this topic, Leveling the Field: An Encyclopedia of Baseball’s All-Time Great Performances as Revealed Through Adjusted Statistics by G. Scott Thomas. By this time I was deeply enmeshed in my own study of the topic, and while I appreciated the fact that G. Scott Thomas had given the subject a voice in print, I didn’t find his methodology entirely compelling.
My disagreement with the author of Leveling the Field was based upon my own long term exploration of different normalization concepts and methods. Back in the early 1990′s I started working on developing my own personal method of comparing professional ballplayers across different eras and accurately determining their peak abilities. Late in the 1990′s I ran my methodology by a good friend who was a civilian statistician for the Department of Defense (not to mention a Senior League Softball National Champion and All-Star Third Baseman) who had once told me of his experience using baseball to teach statistics to young men.
My friend was impressed by the logic and sophistication of the work I was doing, in all honesty, I was floored.
While I have always been very much inspired by the writing of Bill James and his groundbreaking work in helping us all better understand baseball, I considered myself more of a statistical hack than any kind of individual accomplished in mathematics, let alone somebody who could come up with original formulas for baseball analysis.
A Nod to My Influences
It’s been nearly fifteen years since my original work and formulas involving the normalization of players for cross era comparisons, about a decade since I was bold enough to ask a friend to review the work.
Since then my thinking and methods have continued to evolve. I’ve read and at least attempted to understand the dense mathematics of Baseball’s All-Time Best Hitters: How Statistics Can Level the Playing Field, and also Baseball’s All-Time Best Sluggers: Adjusted Batting Performance from Strikeouts to Home Runs, both by Michael J. Schell. Schell is a biostatistician at the University of North Carolina, so he’s far more authoritative on this topic than I’ll ever be, and I do encourage others to read his book (as well as all the others mentioned in this article, excepting G. Scott Thomas’ less than compelling effort.)
I’m also quite sure that I’ll never be as accomplished as Nate Silver ((http://en.wikipedia.org/wiki/Nate_Silver)) or Clay Davenport ((http://en.wikipedia.org/wiki/Clay_Davenport)), just two of the brilliant minds over at BaseballProspectus.com (I’m a proud subscriber) who not only provide comprehensive coverage of current major league ball, but also have an authoritative archive of advanced player statistics including Davenport Translations. The folks at BP also perform the heretofore impossible, predicting the future with their wonderful PECOTA Projections (not to mention Wil Carroll’s injury projections etc).
All those caveats aside, I would like to humbly introduce my own personal method of determining a player’s value, devised for use in producing unique player cards for playing in APBA draft or replay leagues, but of course useful as well in comparing a player’s individual contributions at the peak of their performance. I am currently calling my humble invention “Peak Normalized Bettendorf Transformations”, unless/until I can arrive at a better sort of descriptive moniker.
Essentially this means that the pitching and hitting statistics on the player pages I create will have gone through a proprietary process to adjust for the effect of the era, and even the particular ballpark, in which an individual player performed. This kind of analysis is being done by several sources these days, it’s a process producing what’s variously called “translated”, “neutralized”, or “normalized” statistics. From my perspective, those terms all refer to a relative handful of different methodologies that all attempt to measure the same thing. Everybody is looking for a way to guage a player’s value in relative terms that can be compared across eras. The best I can tell, the various names depend solely on the preference (and imagination) of the statistician who developed them.
The method used to “normalize” the player’s statistics here is, to the best of my knowledge, a unique personal invention. A player’s original pitching and/or hitting statistics go through a number of calculations I’ve programed into a spreadsheet to “transform” them into a standard context. This produces a standardized player record showing each season individually. That’s nothing terribly revolutionary, it produces results similar to what might be found looking at a pitcher’s Neutralized Pitching chart or a hitter’s Neutralized Batting chart at BaseballReference.com, although (obviously – due to differing methodology) my results differ somewhat from what might be found at that excellent site.
After looking over the results of a player’s season by season “transformation”, I then determine a player’s three most valuable seasons, using my player transformation results along with the help of other tools like Win Shares and Pitcher Wins Above Replacement (the latter from BaseballProjection.com). Then with the player’s statistics placed into a standard context (normalized/neutralized/translated/transformed), and their three most valuable individual seasons identified, the three years are then averaged into a single season record to portray a player at their peak performance level.
Every method needs to be assigned a moniker of some sort before being introduced to the public. Being irreverently respectful to my teachers and forebears, I came up with the term “Bettendorf Transformations” as a descriptive for my particular normalization process. You see, in my world, Davenport isn’t a sofa, or even a fine gentleman named Clay, it’s a geographic neighbor of Bettendorf, Iowa.
By coincidence, Bettendorf is the fair city where I received my middle school and high school education at a fine institution named “Rivermont Collegiate“. The school is still today located in what had formerly been the home and mansion of the Bettendorf family, so it was a location that was beautiful, historic, and inspiring. My math teacher for five years of accelerated math in that fine institution was a Grand Master bridge player, so even if I didn’t feel direct inspiration from her in those days, I certainly own Mrs Strohm a debt for teaching me that the math skills she worked so hard to help us to develop, could be applied for game and sport.
Since other statisticians and statistic compilers have already appropiated the terms translation, normalization, and neutralization, I needed some other term for the results of my work. I looked for another descriptive term to describe the process of leveling the field for baseball statistics, and the best I found was “transformation”.
Over the next few months (a process that will potentially last years if I see it thru to the project’s full completion,) I’ll be adding write-ups of individual players that include my personal calculation of their peak normalized value, as well as some biographical information, and a hopefully even a mock up of what their statistics project using Wizard to create their APBA card and BBW cards.