How to Fix the Baseball Hall: Research Report Summary by Dean Krikorian, Ph.D,
Introduction: This post/pictorial summarizes a sizable chunk of research I performed on the Baseball Hall of Fame general election process. It took about six months for me to complete in July, 2014. As an avid baseball fan (go O's!), I also have a Ph.D. in group decision-making processes. I used to sit on several doctoral committees at Cornell University. I consider this my second dissertation because has all of the elements of a done dissertation (the best kind): (1) Background; (2) Literature Review; (3) Research Analysis; (4) Research Methods; (5) Statistical Results; (6) Qualitative Results; (6) Simulated Model; (7) Discussion & Recommendations; (8) Conclusions & Future Directions. I roughly followed this format in this summary post w/ pictures illustrating my findings. Link to the 1-page Abstract.
The Dean-gleaned data was open source, mostly from baseballreference.com, wikipedia.com, and other statistical sources (past election data available in tabular format). As an ex-satellite electrical engineer, I learned to solve problems, like those of the Baseball Hall, by applying the following steps: (1) What are the symptoms to the problem?; (2) What event identified the problem?; (3) What is the specific problem?; (4) What is the problem process?; (5) What caused the Problem?; (6) How do we make sure the problem doesn't happen again (Corrective Action)?; & (7) How does one implement the best solution? I tried to follow these steps here.
My main premise is that I uncovered a flaw in the Baseball Hall of Fame election procedure and I know how to fix it with minimal effort. My premise is that a 3-word rule on the 1958 ballot changing "vote for 10" to "vote for no more than 10" made it unfair for subsequent candidates. Votes per ballot plummeted from 9.2 before 1958 to an all-time low of 5.1 in 2012. Logically, the fewer the votes cast, the less likely a player can gain the requisite 75% to gain induction. My fix is relatively simple, based on research in the area of group decision making processes and is validated in this report. My solution is to let each elector have ten votes to distribute equally among their chosen candidates: Vote for 5, each get 2 votes; Vote for 10, each gets one vote; Vote for 2, each gets 5 votes, etc. As a deterrent to vote stacking for a pet candidate, I also require each inductee to gain at least 50% of the overall ballots. In the end this should restore the Hall Induction rules to those intended by the founding fathers. Link to my Hall of Fame Research Site w/ much more details.
Background: What's Wrong With the Baseball Hall? For those of you who care, the Major League Baseball National Hall of Fame was established in 1936 in rural Cooperstown, NY, as the original sports Hall of Fame and 2nd oldest hall in the US. The BaseBall Writers Association of America election is held every beginning of the year. Players retired more than 5 years have a period of 10 years to be elected by the writers in the general election. Candidates need at least 5% to remain on the ballot. Induction requires at least 75% of the overall ballots cast from the 500 or so members of the BBWAA. Past that, would-be inductees are subject to a relatively selective Era Committee that meets every fourth year. My research mainly examines the general election and their current election process, of which I deem (Dean!) unfair. Most of the ensuing report is presented in pictures.
Below is a series of research charts explaining the symptoms of the Baseball Hall induction issue (collected from various sources, most notably each sports' HOF Website). The rules are different between the sports (See Hall of Fame Rules Table), but more importantly Baseball inducts far fewer players than other major sports (See Figure 3.1), The Football Hall opened 27 years after Baseball and now has more inductees (See Figure 3.5). Figure 3.4 illustrates the paucity of recent Baseball Hall player inductees (the blue and purple bars stacked together). Figure 1.5 shows how Baseball inducts fewer players compared to previous years. From 1999-2013, it was 3.15 (~ pi) times more difficult to be inducted in the Baseball vs. Football Hall of Fame and, coincidentally, 3.15 times more difficult to be inducted into the Baseball Hall compared to previous years. Too few players inducted. This is the problem. Some may not consider this as a problem, but I feel so because the rules have changed making it more difficult for today’s would-be inductees.
The Growing Problem of the Baseball Hall: Too Few Player Inductees: In early 2014, two voter-members took it upon themselves to put an exclamation point on the brewing outrage on the Baseball Hall election process by submitting two radically different solutions to the problem, each calling out the farcical nature of the voting procedure. The first voter solution was by Ken Gurnick of MLB.com, who voted for only one player, Jack Morris. The second solution was by Dan LeBatard of the Miami Herald Tribune, who ceded his vote to deadspin.com (Tim Marchman), who subsequently submitted the vote to fan poll, resulting in a 10-person list including the top 5 vote-getters. These are two radically distinctive voting strategies and, while each strategy was meant to raise awareness, both had no impact on the overall results: (1) Morris was given no help (other than one vote) by being the only one on Gurnick's ballot; (2) The bottom of the LeBatard/Marchman list contained players that fell well short of the required 75% of the total vote. But no one seems to get the gist of the most glaring problem, that the Baseball Hall BBWAA voting procedure has an inherent flaw. And this flaw (or statistical inequity) is most glaring in the above two examples, where Gurnick cast one vote and Tim Marchman chose ten – the variable vote. A variable vote occurs when electors can vote for a different amount of players. This is the issue. I thank Ken Gurnick for alerting me to this anomaly.
The 1958 Rule Change or Fly in the Ointment: I found that the problem of too few player inductees could be traced to the rule change allowing variable voting - a simple three-word rule change enacted in 1958, specifying BBWAA electors to vote for "no more than ten" from “vote for ten” candidates. The change was presumably influenced by a depleted incoming class of candidates (See Smith, 1982, The Baseball Hall of Fame). I find it ironic that this seemingly innocuous rule change is so obscure that it is not even listed on the National Baseball Hall of Fame website. But the numbers and below charts tell the story: Votes per Ballot plummeted from an average of 9.2 Votes/Ballot from 1936-56 to an all-time low of 5.1 Votes/Ballot in 2012: More than 4 Votes lost per ballot, on average! Lower Votes per Ballot decreases the probability of induction by adhering to the same 75%-Ballot threshold. Statistical tests show how, after the 1958 rule change, induction became significantly more difficult - especially in recent years. Although they did not know it at the time, this one rule change forever changed the Baseball Hall of Fame and ultimately abbreviated the candidacy of 30 deserving candidates. The below charts demonstrate: (1) Votes per Ballot decreases over time; (2) the number of Voters increases over time; and (3) there is a significant inverse correlation between the number of Voters and Votes per Ballot; beginning in the 1960s.
 |
 |
Quantitative Results: Fewer Votes per Ballot, More Ballots: The above 3-chart sequence underscores the statistical tests conducted using the last 70 Baseball Hall BBWAA election years. Induction was first compared to the Number of Average Votes per Ballot and the Number of Ballots. This first test (Blue Chart) showed that Votes/Ballot was significantly related to Induction (rVI = .248, p <.05, two-tailed, n = 70). This means that the lower the Votes/Ballot, the less likely it is to get inducted in any given year. The second test (Red Chart) compares the # of Ballots to Induction. Although the number of Ballots increased significantly over time, it was not related to induction and near random (r = -.008, p = ns, n = 70). The final test (Red-Blue Chart) compared Votes/Ballot with # Ballots (See Figure 4.4) and showed that Votes/Ballot was significantly related to # Ballots (r = -.776, p <.01, two-tailed, n = 70). This means that the lower the Votes/Ballot, the more Ballots. This is an important finding because it demonstrates a Mediating Variable effect of Votes/Ballot on Induction: Votes/Ballot is significantly related to Induction and # Ballots, which are not related to one another. Thus, Votes/Ballot plays a large role on determining if players get inducted and it serves to mask the effect on an increasing voter pool - as long as Votes/Ballot are not low, then the number of ballots won't have an effect. But if there are low Votes/Ballot, then the Induction process becomes more randomized given the unchecked effects of a growing voter pool. These results show that Votes/Ballot not only is significantly related to Induction, but it mediates the potential random effects of an increased voter pool. Votes/Ballot need to be close to the maximum 10 to ensure a fair election process.
Research Analysis Summary: The Process of the Problem of the Baseball Hall: The ensuing flow chart depicts the problem process of the Baseball Hall. The main problem can be traced to one main issue, the rule change enacted in 1958 that created the conditions for a more difficult to enter hall. Electors voted for less candidates on their ballot and it still took 75% of the ballots for induction. The "Dark Ages of the Hall" commenced from 1958 for the next 20 years, where only 0.5 players/year were inducted (3-Time MVP Roy Campanella was denied thrice). This resulted in a growing backlog to the hall. In summary, the real problem with the baseball hall can be traced to a 1958 rule change allowing voters to "Vote for up to 10 players" vs. " Vote for 10 players." As a result, the number of votes per ballot declined over time, making it more statistically difficult to get inducted simply because there were less votes cast, while the minimum threshold of 75% of the total ballot remained the same. This rule changed thus caused several players to be delayed induction, as noted by NBCSports Writer Tom Posnanski. This delay resulted in a backlog of previously deserving players at the foot of the hall. And this backlog grew over time. This made it even more difficult to get inducted by the baseball hall: From 2000-2013, it is 3.15 (pi) times as difficult to be inducted as a player.
1958 Rule Change Flow Chart:
Vote for "No More Than" Ten (10) Candidates -->
Less Votes/Ballot -->
More Wasted (or Unused) Votes -->
Less Inductions -->
Backlog Grows -->
Era Committees Induct Only 2 Players -->
Backlog Grows More (Waiting Game) -->
BBWAA does not pick up slack, inducting less players as well -->
Worst Backlog Ever (15 deserving players in line) -->
Backlog Will Grow Beyond Worst Ever -->
Perilous situation for would-be player inductees
How to Fix the Problem?: 12 Potential Solutions Review: But how do you fix the problem? The real issue is that proposed solutions fall short. For example, ESPN's Keith Olbermann readily advocates that BBWAA voters should be able to “vote for as many players as they want.” This horrible solution only induces more induction randomness because it increases the standard deviation of votes per ballot, which increases the variance, which increases randomness. Apparently Olbermann is ignorant of such negative statistical implications, as his plan actually exacerbates the “Banana Republic,” of which he most fears. Increasing the number of votes/ballot was also strongly not recommended. I reviewed 11 such solutions in this report, as summarized in Table 2.1.
Best Solution: Tom Posnanski & "The Waiting Game:" The deemed best solution was provided by NBCSports Writer Tom Posnanski, who suggested the minimum ballot% be reduced from 75% to 50 or 60%. The implementation issue is that this would reduce bar of entry and presumably cheapen the Baseball Hall, at least compared to other sports halls that require 75%. It is highly unlikely that the baseball hall will change this late in its game, it was the original sports hall. But the idea does have its merits because Posnanski examined past induction data and found that the main problem of the Baseball Hall was in DELAYING 78 candidates that eventually got in anyway. The Lost Years of players that deserved earlier induction was actually the main problem identified by Posnanski (with due reference to Blogger Tom Tango). It was the waiting game that made it even more difficult to gain induction.
The Growing Backlog: This backlog has actually grown over the past 13 years because the Veterans/Era Committees have only elected two players and the BaseBall Writers Association of America has not picked up the slack, actually inducting relatively fewer despite a surge in recent elections (See Previous Figure 1.5). As a result the backlog to the hall has reached all-time high when factoring in the Wasted (unused) Votes and Lost (delayed) Votes. In the simulation I performed at the end this report, I found that the "Supermarket Line" to the Baseball Hall is the longest it has ever been in 2014 with 30 deserving players in line (204 total years wait). Factoring in for players who got in later via BBWAA (64 Players, 186 years wait) or Veterans/Era Committees (32 Players, 158 years wait), the waiting game affected 126 players, resulting in 548 combined years of waiting. Hence, the problem Posnanski cast as "the waiting game," is much worse than expected. But what can one do about it, especially at it worsens?
How to Fix the Baseball Hall: Weighted Votes Method: My solution is a Weighted Votes Method: Vote for 5, each gets 2 votes; vote for 2, each gets 5 votes; vote for 10, each gets 1 vote; vote for 20, each gets 0.5 votes, etc. This method assures that ten votes are cast on each elector's ballot. This is the original rule of the Baseball Hall of Fame. The Weighted Votes Method assigns selected candidates equal scores based on the number of players chosen on each elector's ballot: Every voter gets ten votes: It is just that they now get to equally distribute all of ten of their votes. As such, choosy writers like Ken Gurnick could cast all ten of their votes for someone like Jack Morris. There are no more unused votes. Every voter has ten votes or the standard deviation of Votes per Ballot is always equal to zero. This method gives all voters the same voice and allows them to determine their impact. This method is based on group decision research using a "Weighted Votes Methodology," as applied to an "in or out" of a Hall of Fame. My solution factors in my doctoral research on group decision processes and applies the method I used in my successfully defended dissertation: "Individual and Group Normative Forces in Decision-Making Processes." This method also avoids "Arrow's Paradox," which mathematically demonstrates of no one best way to rank order a series of candidate by equally weighting each vote on a given ballot. The Weighted Votes Method should solve several problems: (1) voting for as many players as you want, realizing that more selections means less voting impact; (2) mediating the effects of an increasing number of electors; and (3) reducing the recent backlog into the hall, while (4) remaining relatively difficult for marginal candidates to gain induction.
Method Validity: 170-Ballot Sample from 2014 As evidence, I examined a 176-Ballot sample from 2014 to demonstrate how to score, calculate, and test Weighted Votes using a basic spreadsheet. The resultant impact of this 30.8% sample analysis is to induct candidates earlier that eventually got in anyway, while not aiding marginal candidates. This effectively reduces the backlog. I then validated the internal and external validity of the model, statistically, versus the sample and population for (1) Actual Ballot% =B; (2) Sample Ballot%=S; & (3) Weighted Ballot%=W) (Test Results). Internally, the correlation between Weighted Votes% and Sample Ballot% was r=0.952, p<.001, n=170, two-tailed. Externally, the correlation between Weighted Votes% and Actual Ballot% was r=0.946, p<.001, n=170, two-tailed. Interestingly, the correlation between The Weighted Votes% and Actual Ballot% (r=0.952) was actually was more strongly correlated than the Sample Ballot% to the Actual Ballot% (r=.946). The Weighted Votes Model thus provides an excellent replication of the distribution of actual ballot totals because the effects on players are so highly similar (90% explained model variance). This affords us confidence in implementation because the results show that no one type of player is given an advantage with the Weighted Votes Model.
Procedural Fix: Two Slight Word Changes As basic procedural fix, I offer two slight word changes to the current voting instructions on how to implement the Weighted Votes Method and provide sample “scarecrow” instructions on how to easily enact this change, The Weighted Vote or “Ken Gurnick Amendment.” Then I provided an easy-to-implement solution to prevent and deter potential collusion or vote stacking attempts, the Majority Ballot Rule or “Gil Hodges Provision.”
Results: What the Hall Looks Like Now for the fun part. Results indicate that the Baseball Hall has three basic tiers, whether we recognize it or not (in decreasing order of BBWAA % of Ballots): (1) BBWAA First-Ballot Inductees; (2) BBWAA General Election Inductees; & (3) Veterans, Negro League (not included here), or (Now) Era Committee Inductees. The below chart places these groups as Green, Light Blue and Purple (respectively). Today, we also seem to want to stratify the best (Green) players by noting their percentage of votes on the first ballot. Seaver, then Ryan, then Ripken, then Cobb - according to the highest percentages. But this makes no sense because Babe Ruth is ranked 12th in the Green Group. Jackie Robinson barely made it on the first ballot. Joe DiMaggio didn't make it on the first ballot. Scroll down the Green Group and verify the badasses and where your player stacks up. It is interesting that, past the original five inductees in 1936, the next first-ballot inductee was Bob Feller in 1962. What I find most interesting are the Groups in the middle...
The Purple Group was inducted by the Veterans or (Current) Era Committees. Thus, they were not elected by the BBWAA and can be deemed as "Second Class Citizens" in the Hall. They could not get in after 15 (now 10) tries via the writers. But at least the Blue and Purple are in the Hall. There are some interesting stories in the Blue & Purple and Red, Orange, & Yellow (not in the Hall). Perhaps most interesting is the relative wait for these players into the hallowed hall. I figured this out by estimating and factoring in (1) the number of wasted votes due to incomplete ballots and (2) the number of lost votes due to players lingering on the ballot. The Total Years wait is simply the Year they got inducted minus the year they "shoulda" got inducted, according to my simulation. Nellie Fox died before he got inducted, So did Ron Santo (a really tragic story, but what a great man!). They should have been inducted when they were living, foiled by an unfair voting procedure. Joe Torre shoulda been inducted as a player. But this is merely water under the bridge. As noted earlier, this effect of delaying candidates their due entry was first discovered by NBCSports Columnist Joe Pisnanski (w/ due props to Blogger Tom Tango). I found in my analysis that the effect was far greater than expected. In the below chart, the top 22 players delayed or denied entry totaled 322 years: Nearly 15 years wait per player, on average.
This last chart summarizes the overall findings of my research. The Good, The Bad, and the Ugly. The Good are the players that shoulda been in, had the writers filled out their ballots completely. The Red are the most deserving, followed by the Orange, then the Yellow. Many of these players were the victims of so-called "Double Whammy" years of Low Votes/Ballot and a flood of new stellar candidates. In the below chart, the first number in the box is their projected highest score and the second number is their highest BBWAA ballot % year total. The above list contains many recent candidates: TIm Raines, Jack Morris, Lee Smith, Alan Trammell, etc. (save Gil Hodges). But toward the end you see some older candidates popping up, many past their eligibility such as Jim Kaat, Dave Parker, Lew Burdette, Orel Herschiser, Minnie Minoso, & Luis Tiant. These players were dissed from the Hall, victims of an unfair voting system that made it more difficult to gain entry than their predecessors, especially those that played in the 1930s.
The above two Groups are both Fortunate and Lucky. They got in despite garnering only a few votes in their opportunities in the general BBWAA election. All were elected by committee. The Purple group had a few votes cast for them, but not many. Most of these players are from the 1930s, where batting averages were greatly inflated. The same is mostly true for the Brown Group. In fact, these players never got more than 5% in any BBWAA election! Many of these candidates were the subject of cronyism, when the old Veterans Committee let in their buddies or fulfilled the dying wish of Judge Kennesaw Mountain Landis (See Previous Hall Snafus). I made this Group Brown on purpose...In the end, there are those who are Good, Bad, and Ugly, even in the Baseball Hall. Baseball fans everywhere were exasperated with their Hall.
Statistical Simulation: What Would the Hall Look Like Before the 1958 Rule Change
I statistically simulated the effects of this rule change on all previous BBWAA elections using Weighted Votes from all previous hall general elections using a weighting algorithm, which proportionately assigns unused votes (i.e., those on less than 10-vote ballots) to each candidate. My Simulated Baseball Hall reflects the factoring in of ten votes per ballot and the earlier entry of deserving candidates (i.e., Wasted and Lost Votes, respectively).
Check it out at http://linksviewer.com/baseballhall. There are many options in viewing the simulation: (1) playing the movie over time; (2) slowing down or speeding up the process; (3) adding or subtracting groups of candidates. The end result is the below "Simulated In" group of candidates and "Simulated Out" group of less deserving candidates. Given that the latter group is not going to get their "Hall pass" revoked, it seems prudent to let the more deserving group in, or certainly give them their due process in light of new evidence. I hope the Board of Directors take heed to the recommendations in this report. If only the Hall were more fair!
