Backstage
Menu
Statting Lineup
Blog Posts
For those that enjoy reading, seeing visuals, and having immediate access to data sources that are used. The frequency of posts varies greatly and merely depends on when I have an idea come to mind and when I have the time to address it. Feel free to contact me with any blog ideas or desires.
Tomorrow, (Sunday, December 4th), the results from the Hall of Fame's 2023 Contemporary Era Committee ballot will be announced. The ballot consists of 8 players that made their primary impact after 1980, and you can view the players on the ballot here. With this in mind, I thought I would take a break from my Player Value research to see what my Hall of Fame predictive model thought of these candidates, as well as share my own thoughts and provide the current version of Player Value for each candidate's career. I'll go over some details of my Hall of Fame predictive model and its use on the 2022 ballots first; feel free to skip ahead if you'd just like to see the model's and my thoughts on the 2023 Contemporary Era ballot candidates. Hall of Fame Predictive Model Overview I first introduced my model and used it on the 2022 BBWAA ballot here. I entered the model into the 2021 Fall USCLAP competition during my final semester in college, and it ended up finishing in 2nd place. You can view the winners here, and the official report here. The report dives into the nitty gritty of the model, if you are interested in predictive modeling and learning those fine details. As a quick(ish) summary, the model is intended to only be used on position players that finished their careers after 1957, and that did not use steroids or have some other obvious scandal that is the primary deterrent of the Hall of Fame induction. There are some players, such as Barry Bonds and Pete Rose, that are statistically pretty obvious Hall of Fame inductees, but that are left out because telling the model that someone of their caliber isn't a Hall of Famer would confuse it, since it only considers their performance on the field. While I could have keyed in a "character clause" predictor in the model to handle this, I felt it easier to just exclude the players in question. Gold Gloves and other awards are fairly important predictors of whether a player is in the Hall of Fame, and earlier years (i.e. pre-1957) lacked many of these awards, so the model would unfairly judge these earlier players. Since pitchers are judged on an entirely different basis than position players, it didn't make sense to predict them using the same model, so pitchers are excluded as well. Lastly, Negro League players and stats are not incorporated in the model. While these leagues were basically equally competitive as the Major Leagues at the time (as seen by the dominance of early transition players like Jackie Robinson and Roy Campanella), they played far shorter seasons and have less recorded statistics, so these players' stats and model predictions would be flawed when considering their career statistics. The model uses 5 classes of predictors:
Defensive statistical averages per season, such as putouts per 162 games, were considered but not used due to their lack of predictive power. In fact, most of a player's defensive Hall of Fame value is only encompassed in their Gold Gloves. Only their fielding percentage and range factor per game differences from league average ended up being predictive from the career defensive statistics. Generally, the most important predictors for the various submodels (listed below) ended up being a player's All-Star seasons, career runs scored, career singles, and career RBI. In fact, a simple decision tree model can be run to visualize these important predictors: This simple decision tree model predicts anyone with at least 7 All-Star seasons and 1,208 runs scored as a Hall of Famer, an assertion that's right every time based on the dataset. It also predicts anyone with at least 7 All-Star seasons, less than 1,208 runs scored, but more than 1,239 singles as a Hall of Famer, an assertion that's only right about half the time based on the dataset. Anyone with fewer All-Star seasons, runs scored, or singles is predicted as not a Hall of Famer, an assertion that we can see is nearly always correct. This simple decision tree model is not as accurate as my actual Hall of Fame model, but it's surely better than a coin-flipping approach and is very easy to interpret and helps us visualize that All-Star seasons and runs scored are the preeminent predictors of a player's Hall of Fame fate. Again, this was just an illustrative example - this is NOT my actual Hall of Fame model. My initial model was completed just before the results from the 2022 Golden Days Era Committee ballot were announced. You can check out those candidates here. That ballot consisted of 9 players (7 position players and 2 pitchers) and 1 manager. Jim Kaat, Gil Hodges, Tony Oliva, and Minnie Minoso would all go on to be inducted into the Hall of Fame. The initial model performed really well, with an AUC of .9817. AUC stands for Area Under the Curve, and is basically a measure of model accuracy on a scale of 0 to 1. An AUC of 0.5 represents a random guess, coin flip approach. The higher the AUC, the more accurate the model, and my model's AUC was quite high. The simple decision tree model that I displayed above had an AUC of .8105. We technically care about the test AUC, which is the AUC of the model on the test set, meaning the data/players that the model was not trained or developed on. The training set are the players used to essentially teach the model, and then the test set are the players used to evaluate the model's accuracy. My Hall of Fame predictive model is an ensemble model of 4 different submodels:
With 187 players in my dataset overall, I placed 141 in the training set and 46 in the test set. Of the 46 players in the test set, 17 were Hall of Famers, and thus 29 were not. My model correctly predicted 28 of the 29 non-Hall of Famers, asserting that Dave Parker should be a Hall of Famer (personally, I agree!). It correctly predicted 15 of the 17 Hall of Famers, stating that Lou Brock and Alan Trammel were not up to par. The initial model was designed for predicting future players on each year's BBWAA ballot, not the Era Committee ballots. The Hall of Famers and non-Hall of Famers that the model was trained on were players that already had their BBWAA fates decided, and any candidate on an Era Committee ballot would have been rejected by the BBWAA already. Because of this, simply using the same initial dataset that trained the initial model and then predicting the 2022 Golden Days ballot players is a flawed approach that results in none of the players being predicted as Hall of Famers. For any player such as Ken Boyer that was in the training set, the model will predict them as a non-Hall of Famer since that is exactly what the model was told when training on the data. While it could predict players in the test set still fine, all of the Golden Days candidates in the test set were not predicted as Hall of Famers. The proper approach is to remove the 7 position players - Dick Allen, Ken Boyer, Gil Hodges, Roger Maris, Minne Minoso, Tony Oliva, and Maury Wills - from the data and retrain the model on this adjusted dataset. Doing this worsens the predictive accuracy of the model down to .9477. While this is worse than the initial model's AUC of .9817, it is still great overall. Nonetheless, this reduction in accuracy foreshadows how the model thinks of these players. Removing the information that told the model that these players weren't Hall of Famers made it worse. After applying this adjusted version of the model to the players on the 2022 Golden Days ballot, only Dick Allen was predicted as a Hall of Famer. This isn't too shocking, as Allen's bWAR of 58.7 is larger than those of the players' who ended up getting inducted - Hodges at 43.9, Oliva at 43.0, and Minoso at 53.8. Allen is also 23rd all-time in career OPS+ at 156, tied with Frank Thomas and the most of any player that isn't active, used steroids, banned from baseball, or simply archaic (sorry Pete Browning and Dave Orr). His OPS+ is higher than that of both Hank Aaron and Willie Mays. Outside of Jim Kaat - whose 16 Gold Gloves are the 2nd most all-time by a pitcher (and thus isn't handled by the model) and has 287 wins with 2,461 strikeouts - I personally wasn't too sold on any of the players being inducted on last year's era ballot, but Allen would have been at the top of my consideration. As I wrote about previously, when applying this adjusted version of the model to the players on the 2022 BBWAA ballot, only David Ortiz and Todd Helton were predicted as Hall of Famers. Of course, Ortiz was inducted his first year on the ballot with 77.9% of the vote. Helton received 52% of the vote in his 4th year on the ballot, an increase from his 44.9% received in 2021. When applying the Hall of Fame predictive model to the 2023 Contemporary Era ballot, there are 2 approaches we can take:
Albert Belle, OF Years: 1989-2000 Teams: Cleveland Indians, Chicago White Sox, Baltimore Orioles Accolades: 5x All Star, 5x Silver Slugger, 3x RBI Leader, 2x SLG Leader, 1x Runs, Doubles, HR Leader Key Stats: 381 HR, 1,239 RBI, 389 Doubles, 1,539 G, 6,676 PA, .933 OPS, 144 OPS+, 3,300 TB Player Value: 312.15 Total, 244.60 Batting Value, -2.52 Baserunning Value, 70.05 Fielding Value Photo courtesy of WKBN 27 Model Talk: The initial combined ensemble model gives Belle a Hall of Fame probability of .2379, which rounds down to 0 and thus predicts him as not a Hall of Famer. The FDA submodel is particularly unimpressed with Belle, giving him a probability of just .0008. The GLM model isn't too fond of Belle either, giving him a probability of .2145. The model averaged neural network holds a similar stance, with a probability of .1667. However, the SVM model does think Belle should be a Hall of Famer, giving him a probability of .5695. While the final combined ensemble model's AUC was .9664, the GLM and FDA submodels were actually more accurate after the training set updates, with AUCs of .9800 and .9811, respectively. The SVM and neural network submodels were still worse, however, with respective AUCs of .9054 and .9391. Given that the most accurate submodel gives Belle the lowest probability, and the least accurate submodel gives Belle the highest probability, we can conclude that the model doesn't like Belle too much. It did give him a higher combined probability than Hall of Famer Alan Trammel (.1039), however. Since the ensemble model isn't actually the best in this case, and since the 4 submodels have varying accuracy, another approach for the final probability is to compute a weighted average based off of the accuracy of each submodel, rather than using a simple average. That is to say, weight the GLM and FDA predictions more heavily since they are more accurate, rather than treating them equally as the SVM and neural network predictions. This alternative approach makes the ensemble model's AUC now slightly higher at .9706. This alternative approach has a pretty minimal effect. Each of the submodel probabilities are the same, but Belle's new final ensemble probability is now slightly lower at .2321. Again, this is due to the more accurate submodels giving him lower probabilities of being a Hall of Famer. What if we update the training set with the 2022 Hall of Fame results and then retrain the model? The resulting ensemble model is worse, with a lower AUC of .9286. The updated FDA submodel has an AUC of .8948, the updated GLM submodel has an AUC of .9206, the updated SVM submodel has an AUC of .9246, and the updated neural network submodel has an AUC of .9206. So, the FDA and GLM submodels got worse, as did the neural network submodel and the ensemble model overall, but the SVM submodel actually became more accurate with the 2022 Hall of Fame results. This updated simple average ensemble model gives Belle a probability of .2430, slightly higher than without the updates but still not enough to be predicted as a Hall of Famer. The FDA submodel probability is .0022, the GLM submodel probability is .2115, the SVM submodel probability is .4680, and the neural network submodel probability is .2904. Lastly, if we use the weighted average ensemble model with the training dataset that includes the 2022 Hall of Fame results, Belle's new ensemble probability is slightly higher at .2450. Still not high enough to be predicted as a Hall of Famer. In this case, the now less accurate FDA submodel is weighted less while the now more accurate SVM model is weighted more. The FDA submodel liked Belle the least, and the SVM submodel liked him the most, so the increase here makes sense. Interestingly enough, in this case the weighted average ensemble model is just as accurate as the simple average ensemble model, as both had an AUC of .9286. Predictive Model Verdict: Not a Hall of Famer My Thoughts: Maybe you don't think Belle's career totals are that impressive, and maybe you're wondering why I included his career games played and career plate appearances under his "Key Stats". The answer is context. Belle played in just 12 seasons, including his first 2 seasons when he played in just 71 games combined. So out of 10 real full seasons, he was an All-Star half of the time and a Silver Slugger half of the time. He also finished in the top 10 in MVP voting for 5 of those seasons, and in my opinion was robbed of the 1995 MVP by Mo Vaughn, who he bested in basically every offensive category (you can see for yourself here). He hit 30+ HR in 8 of those seasons, and hit 28 and 23 in the other two. He had 100+ RBI in 9 of those seasons, and recorded 95 in the other one. He hit 30+ doubles in 9 of those seasons, and hit 23 in the other one. Consistently 30 doubles, 30 homers, and 100 RBI per season? I'll take that. In terms of the 255 players used in my predictive model dataset, Belle ranks 6th in doubles per 162-game season with 40.9, behind 4 Hall of Famers (Medwick, Greenberg, Hafey, Herman) and Nomar Garciaparra. He ranks 7th in RBI per 162-game season with 130.4, behind 6 Hall of Famers (Gehrig, Greenberg, DiMaggio, Ruth, Foxx, Simmons). He ranks 3rd in HR per 162-game season with 40.1, behind 2 Hall of Famers (Ruth and Kiner). Looking at the LF JAWS leaderboard, he ranks 10th in MVP shares behind Barry Bonds, Pete Rose, Manny Ramirez, and 6 Hall of Famers. I will note that WAR doesn't like Belle's peak quite as much, as his 7 year peak WAR of 36.0 ranks just 29th all-time among left fielders (but ahead of him are 16 HoFers, Rose, and Bonds). That is all to say that Belle had a tremendous peak. But it wasn't just that he was great for a decade and then slowly panned out; his career was abruptly cut short at the age 33 due to a hip injury. I can't emphasize this enough, as I feel it is frequently overlooked when discussing Belle's case. Kirby Puckett and Roy Campanella had career-ending injuries at 35 and are in the Hall of Fame. Ralph Kiner had a career-ending injury at 32 and is in the Hall of Fame. Heck, even Ross Youngs was done by 29 due to illness and is somehow in the Hall of Fame. So, why not Belle? When a player's great career is suddenly cut short, I prefer to give him the benefit of the doubt. Belle's total Player Value of 312.15, under the current version, ranks 66th out of the 4,737 position players since 1974, which puts him in the top 1.4%. His Batting Value of 244.60 ranks him 59th, which is the top 1.25%. Clearly Player Value thinks Belle's peak was sufficiently great! My Opinion: Put Him In Don Mattingly, 1B Years: 1982-1995 Teams: New York Yankees Accolades: 9x Gold Glove, 6x All Star, 1x MVP, 3x Silver Slugger, 3x Double Leader, 2x Hit Leader Key Stats: 2,153 H, .307 BA, 442 Doubles, 222 HR, 1,099 RBI, 7,722 PA, 1,785 G Player Value: 205.42 Total, 66.50 Batting Value, -0.33 Baserunning Value, 139.25 Fielding Value Photo courtesy of NBC Sports. Model Talk: The initial ensemble model gives Mattingly a combined probability of .2787, rounding down to a non Hall of Famer prediction. The FDA model has him at just .0331, while the GLM, SVM, and neural network models are slightly more positive, giving him respective probabilities of .3066, .4377, and .3375. If we use the approach where we weight the submodels based off of their accuracy, Mattingly's new ensemble probability becomes .2749, even worse than before. Again, the GLM and FDA submodels were the most accurate (highest AUCs) and they gave Mattingly the lowest Hall of Fame probabilities. If we use the model that was retrained on the training data that includes the 2022 Hall of Fame voting results, the simple average ensemble model gives Mattingly a notably higher probability of .4351, but this is still too low to be predicted as a Hall of Famer. The FDA submodel gives him a probability of just .0076, the GLM submodel gives him a probability of .4479, the SVM submodel gives him a high porbability of .7002, and the neural network submodel gives him a solid probability of .5849. If not for the FDA submodel's tiny probability (which now has the lowest AUC and is thus the least accurate after the 2022 updates), then Mattingly might have been predicted as a Hall of Famer by the simple average ensemble model. Lastly, if we use the weighted average ensemble model with the training dataset that includes the 2022 Hall of Fame results, Mattingly's new ensemble probability is slightly higher at .4384. Still not high enough to be predicted as a Hall of Famer. In this case, the now less accurate FDA submodel is weighted less while the now more accurate SVM model is weighted more. The FDA submodel liked Mattingly the least, and the SVM submodel liked him the most, so the increase here makes sense. Predictive Model Verdict: Not a Hall of Famer My Thoughts: Mattingly also retired early at age 34, but due to a more gradual deterioration via back injuries rather than due to a sudden career-ending injury. His 9 Gold Gloves are the 2nd most by a first baseman in history, behind only Keith Hernandez (who I also think should be inducted and been included on this ballot). Besides these two, every other players with at least 9 Gold Gloves is in the Hall or is still having their fate decided. I think the many Gold Gloves are what drives Mattingly's case for me, while still not being bad offensively. Despite a shorter career he still amassed at least 2,000 hits and 1,000 RBI, won a batting title, led the league in OPS in 1986 when he finished 2nd in MVP voting, and led the league in doubles 3 times. Only 9 players that spent at least 50% of their time at first had a higher career batting average than Mattingly's .307 with as many plate appearances. Of those 9, 7 are in the Hall of Fame, Todd Helton is still awaiting his fate (I think he should be in), and the last is Stuffy McGinnis, whose .307 career batting average is contextually not as impressive given that he played in an earlier era where higher batting averages were more common. Mattingly may not have been a powerhouse offensively, and WAR may disagree with his defensive ability, but he won 9 Gold Gloves nonetheless and like Dale Murphy below is another player that can bolster the lackluster Hall of Fame membership of players from the '80s. Mattingly's total Player Value of 205.42, under the current version, ranks 155th out of the 4,737 position players since 1974, which puts him in the top 3.27%. At least under the current iteration, this suggests that Don is more 'Hall of Great' territory. His Fielding Value of 139.25 ranks him 127th, which is the top 2.68%. My Opinion: Put Him In Fred McGriff, 1B Years: 1986-2004 Teams: Toronto Blue Jays, Atlanta Braves, Tampa Bay Devil Rays, San Diego Padres, Chicago Cubs, Los Angeles Dodgers Accolades: 5x All Star, 3x Silver Slugger, 2x HR Leader Key Stats: 493 HR, 2,490 H, 1,550 RBI, 1,305 walks, 1,349 R Player Value: 247.50 Total, 180.52 Batting Value, -0.57 Baserunning Value, 67.55 Fielding Value Photo courtesy of Sports Illustrated. Model Talk: The initial ensemble model gives McGriff a solid probability of .6322, rounding up to a Hall of Fame prediction. The FDA submodel loves McGriff, giving him a .9019 probability. The GLM submodel thinks otherwise, giving him just a .3492 probability. The SVM and neural network submodels also support his candidacy with probabilities of .5418 and .7357, respectively. If we use the approach where we weight the submodels based off of their accuracy, McGriff's new ensemble probability becomes .6330, slightly higher than before. The FDA submodel was the most accurate (highest AUC) and it gave McGriff the highest Hall of Fame probability, thus the increase. If we use the model that was retrained on the training data that includes the 2022 Hall of Fame voting results, the simple average ensemble model gives McGriff a notably lower probability of .3710, which is now low enough to not be predicted as a Hall of Famer. The FDA submodel gives him a much lower probability of just .0694, the GLM submodel gives him a probability of .4100, the SVM submodel gives him a probability of .5260, and the neural network submodel gives him a probability of .4785. The FDA submodel went from loving McGriff to hating him, and became much less accurate in the process. Lastly, if we use the weighted average ensemble model with the training dataset that includes the 2022 Hall of Fame results, McGriff's new ensemble probability is slightly higher at .3733. Still not high enough to be predicted as a Hall of Famer. In this case, the now less accurate FDA submodel is weighted less while the now more accurate SVM model is weighted more. The FDA submodel liked McGriff the least, and the SVM submodel liked him the most, so the increase here makes sense. Predictive Model Verdict: Hall of Famer, ignoring last year's results My Thoughts: McGriff may not have won an MVP, but he finished in the top 10 in voting 6 times. The only first basemen with more top 10 MVP finishes are 5 HoFers (Gehrig, Thomas, Murray, Killebrew, Ortiz), 2 future HoFers (Pujols, Cabrera) and Freddie Freeman. Tied with McGriff with 6 top 10 MVP finishes are 3 HoFers (Mize, Terry, Bagwell), 2 active players that I think are likely future HoFers (Votto, Goldschmidt), Ryan Howard, and Andres Gallaraga. McGriff's 493 home runs are tied with Lou Gehrig for the 12th most by a first basemen and the 29th most across all positions. The 11 first basemen with more homers are 7 Hall of Famers, 2 future Hall of Famers (Pujols, Cabrera) and 2 notable steroid users (McGwire, Palmeiro). Of the 28 players with more homers, every single one is either in the Hall of Fame, used steroids, still active, or not yet eligible for the Hall of Fame. Fred McGriff has the most home runs of any "clean" player that has been thus far rejected for the Hall of Fame. I personally think reaching 500+ home runs should automatically qualify a player for the Hall, granted that they didn't use steroids, and historical voting seems to reflect this rule. The fact that McGriff has been excluded due to 7 homers is absurd, especially when we consider the 1994 strike-shortened season. In 1994 the Braves (like all other MLB teams) played a shortened schedule of 114 games, of which McGriff played in 113. In those 113 games, McGriff hit 34 home runs, good for a pace of about .3 homers per game. Across a normal full 162 game season, that's 48.6 homers. McGriff played in 113 out of 114, or 99.12% of his team's games. That brings the hypothetical full season total down to 48.17, which we'll round down to 48. That's 14 more HR than his actual 34 in 1994. With this short hand math, we estimate McGriff would have had 507 career home runs, if not for the 1994 strike. McGriff's 1,550 RBI rank 47th most all-time. You can take a look at this list and see that all of the players ahead of him are either in the Hall, used steroids, or haven't been on a ballot yet (Beltre and Beltran). He ranks 15th in RBI among first basemen, with the usual HoF/steroid/not yet eligible suspects ahead of him. His 2,490 career hits also rank 15th among first basemen. He hit 30+ home runs in 10 seasons, a feat achieved by just 21 players. Besides Carlos Delgado, every other player that has done this is either in the Hall, will be in the Hall, or used steroids. The advanced metrics don't like McGriff as much. His WAR of 52.6 ranks 30th all-time among first basemen, below the Hall of Fame positional average of 65.5 (which would rank 14th). His JAWS of 44.3 ranks 31st all-time among his position, also below HoF average. But should the clean guy with the most HR and RBI not in the Hall be excluded, especially if he would have reached the essentially automatic qualifier of 500 HR if not for a strike? I think not. McGriff's total Player Value of 247.50, under the current version, ranks 119th out of the 4,737 position players since 1974, which puts him in the top 2.5%. His Batting Value of 180.52 ranks him 94th, which is the top 1.98%. My Opinion: Put Him In! Dale Murphy, OF Years: 1976-1993 Teams: Atlanta Braves, Philadelphia Phillies, Colorado Rockies Accolades: 7x All Star, 2x MVP, 5x Gold Glove, 4x Silver Slugger, 2x HR and RBI Leader Key Stats: 398 HR, 2,111 H, 1,266 RBI, 350 doubles Player Value: 166.45 Total, 220.28 Batting Value, 1.48 Baserunning Value, -55.31 Fielding Value Photo courtesy of Baseball Egg. Model Talk: The initial ensemble model gives Murphy a probability of .5079, barely rounding up to a Hall of Fame prediction. The FDA submodel isn't too high on Murphy, giving him a probability of .2701. The GLM submodel is slightly more favorable at a .3986 probability. The SVM submodel is a virtual toss-up with a .4943 probability. The neural network submodel is a big fan of Murphy, giving him a .8687 probability. If we use the approach where we weight the submodels based off of their accuracy, Murphy's new ensemble probability becomes .5043, slightly lower than before. The SVM and neural network submodels were the least accurate (lowest AUCs) and they gave Murphy the highest Hall of Fame probabilities, thus the decrease. However, these submodels were still accurate enough and still gave Murphy high enough probabilities to still merit a Hall of Fame prediction overall. If we use the model that was retrained on the training data that includes the 2022 Hall of Fame voting results, the simple average ensemble model gives Murphy a slightly higher probability of .5125, which is high enough to be predicted as a Hall of Famer. The FDA submodel gives him a probability of .1462, the GLM submodel gives him a probability of .5207, the SVM submodel gives him a high probability of .7555, and the neural network submodel gives him a solid probability of .6274. Lastly, if we use the weighted average ensemble model with the training dataset that includes the 2022 Hall of Fame results, Murphy's new ensemble probability is slightly higher at .5153. Still not high enough to be predicted as a Hall of Famer. In this case, the now less accurate FDA submodel is weighted less while the now more accurate SVM model is weighted more. The FDA submodel liked Murphy the least, and the SVM submodel liked him the most, so the increase here makes sense. Predictive Model Verdict: Hall of Famer, regardless of last year's results My Thoughts: Murphy has the accolades worthy of a Hall of Famer, but his cumulative career totals are somewhat lacking. We can't make any type of career hits, homers, or RBI arguments for Murphy like we can with McGriff. But he did win 2 MVPs, which only 5 center fielders have done in history. The other 4 are future HoFer Mike Trout and 3 HoFers (Mantle, Dimaggio, Mays). Not a bad crowd. WAR does disagree with his winning of these MVPs, however, favoring Gary Carter instead in 1982 and John Denny or Dickie Thon in 1983. There have been a total of 32 players that have won multiple MVPs in history, and just 11 have won at least 3 with only Barry Bonds winning more than 3. Of the 31 other dudes, 23 are in the Hall of Fame, 3 used steroids (Bonds, A-Rod, Juan Gonzalez), and 3 are future Hall of Famers (Pujols, Cabrera, Trout). The remaining 2 are Bryce Harper - who is still active and likely a future Hall of Famer as well - and Roger Maris. Murphy has nearly 800 more hits, 100 more HR, and 400 more RBI than Maris, as well as 4 more Gold Gloves. Their cases are very similar, but Murphy was able to stay around a few seasons longer than Maris was, and was better defensively (at least in terms of awards; Rfield has Murphy at -34 and Maris at 45). Amongst center fielders, Murhpy's 398 home runs actually track pretty well, ranking him 8th. Ahead of him are 5 HoFers (Mays, Griffery Jr., Mantle, Dawson, Snider) and 2 players whose Hall of Fame fates have yet to be truly decided in Andruw Jones and Carlos Beltran. In general, the 1980s are underrepresented in Cooperstown. Greats like Darryl Strawberry, Dave Stewart, Keith Hernandez, Dwight Gooden, and Dave Stieb have all been excluded. Sports Reference's Adam Darowski shared a split of Hall of Famers by their debut year on Twitter. The 1980s have just 16, compared to 22 from each of the '50s and '60s, and a whopping 46 from the '20s. Murphy was another one of the '80s greats, and inducting him could help begin righting this wrong. Murphy's total Player Value of 166.45, under the current version, ranks 212th out of the 4,737 position players since 1974, which puts him in the top 4.48%, not quite Hall of Fame caliber. His Batting Value of 220.28 ranks him 66th, however, which is the top 1.39%. I certainly think that the offensive side of Player Value is currently more accurate than the defensive side. My Opinion: Put Him In So the model says to put 1 to 2 guys in, and I'd put all 4 in given the chance. What can I say, I'm a "big Hall" guy. The following players weren't predicted by the model since they're pitchers or used steroids, but here are my thoughts on their Hall of Fame cases: Curt Schilling In my hypothetical 2022 ballot, I highly emphasized that I thought Schilling should be in Cooperstown. Straight from that earlier post: "Historically, certain career marks have been guarantees for induction. One such milestone is 3,000 strikeouts, which only 19 pitchers have done in history. Of these, 2 are active players (Max Scherzer and Justin Verlander), 1 is not eligible for the ballot yet (C.C. Sabathia), and 1 used steroids (Roger Clemens). Of the remaining 15 pitchers with 3,000 or more strikeouts, 14 of them are in the Hall of Fame and the other is Curt Schilling. Schilling's 3,116 career K's are good for 15th all-time, more than Hall of Famer John Smoltz's career total in about 200 less innings, and just 1 less than Hall of Famer Bob Gibson's career total in about 600 less innings. Schilling's career WAR of 79.5 is 26th best among starting pitchers and the most of any starting pitcher not in the Hall of Fame, with the exception of Clemens. Schilling also rocks an impressive 6 All-Star game seasons, 3 World Series, and a World Series MVP. While he never won a Cy Young award, he did come in 2nd place three times and in 4th place once. People like to rag on Schilling's character, which is admittedly deplorable, but... [t]he Hall contains the best baseball players in history, and Curt Schilling is clearly one of them and therefore should be inducted." Max Sherzer and Justin Verlander have since passed Schilling in K's to now rank him 17th all-time, but the point still stands. Schilling has no connection to steroids and absolutely should be inducted as one of the game's great pitchers, regardless of how objectively awful of a person he is. The current version of Player Value has Schilling at 188.64, ranking him 29th among the 6,077 pitchers since 1974. His Pitching Value of 237.97 ranks 14th. Barry Bonds My hypothetical 2022 ballot also included Bonds. I'm not going to hash out his case all over, but feel free to click the link above under Schilling to review what I previously stated. The short of it is that I'm generally against steroid users in the Hall of Fame, but make an exception for Bonds who was clearly a Hall of Famer prior to his steroid use and was statistically significantly better than his steroid counterparts. That was my stance for his final year on the BBWAA ballot, which included 394 voters. The 2023 Contemporary Era Committee will consist of just 16 voters. I still think Bonds ought to be in, but I'd rather the first undeniable steroid user that is inducted to be voted in by more people via the BBWAA ballot. Plus, Bonds just had his chance last year on the BBWAA ballot; other worthy and steroid-free candidates on the ballot have had to wait longer for their next chance at the Hall. The current version of Player Value has Bonds at 1,201.08, easily the most of any player since 1974. His Batting Value of 1,015.49 also ranks 1st, while his Fielding Value of 166.82 ranks 77th. Roger Clemens My stance on Clemens is basically exactly what I stated for Bonds above. Statistically, obviously a Hall of Famer, but his steroid use calls him slightly into question. Nontheless, I would have put him on my BBWAA ballot last year. However, I think the larger BBWAA ballot should sort out the steroid users before we let just 16 (or really, only 12) people determine if they should be inducted. The current version of Player Value has Clemens at 555.94, ranking him 2nd among the 6,077 pitchers since 1974. His Pitching Value of 518.04 ranks 1st. Rafael Palmeiro Palmeiro is in the same boat for me as Bonds and Clemens, he just wasn't on the 2022 BBWAA ballot. He is clearly a Hall of Famer when you ignore the steroids, clinching the "automatic" qualifiers of both 500+ home runs and 3000+ hits. Given his steroid use, there are more preferable guys to use for the ballot's limited number of spots. The current version of Player Value has Palmeiro at 329.68, ranking him 61st among players since 1974. His Batting Value of 198.59 ranks 76th. I emphasized the current version of Player Value because it is far from complete, but still, it's not that bad or wrong as is. Most Batting Value since 1974? Barry Bonds. Most Baserunning Value since 1974? Rickey Henderson. Most Fielding Value since 1974? Ozzie Smith. Most Fielding Value among pitchers since 1974? Greg Maddux. Most Pitching Value since 1974? Roger Clemens. My Hypothetical 2023 Contemporary Era Ballot:
The actual 16-person committee members that will vote for this ballot was announced recently, and includes former Braves Hall of Famers Greg Maddux and Chipper Jones, both of whom were teammates with Fred McGriff from 1993 to 1997. Frank Thomas spent 2 years with Alberte Belle on the Chicago White Sox in 1997 and 1998. Lou Whitaker would have been a great candidate to benefit from the committee's makeup given the inclusion of TIgers teammates Jack Morris and Alan Trammel, but alas. Lee Smith and Ryne Sandberg (and Greg Maddux) played some with Rafael Palmeiro on the Cubs in the late '80s before he really burst onto the scene. Needless to say, I think the committee's makeup will certainly benefit McGriff the most. As usual, I'll end with a file dump - My USCLAP paper on my Hall of Fame predictive model:
A PowerPoint presentation I gave on the model:
Dataset of players used:
Datasets for 2022 Golden Days, 2022 BBWAA, and 2023 Contemporary Era ballots:
R files for the initial model and 2022 BBWAA predictions, 2022 Golden Days predictions, and 2023 Contemporary Era predictions:
Adjusted datasets that you'll need to run the 2022 and 2023 era ballot R files above:
And that should be everything you need to dig deeper into or replicate the results. Thank you all for reading! Looking forward to sharing more Player Value findings soon, as well as my model's predictions for the upcoming 2023 BBWAA Hall of Fame ballot. Statting Lineup Newsletter Signup Form:
If you'd like to receive email updates for each new post that I make, sign up for the Statting Lineup newsletter using the link below: https://weebly.us18.list-manage.com/subscribe?u=ab653f474b2ced9091eb248b1&id=3a60f3b85f
0 Comments
Leave a Reply. |
Statting Lineup Newsletter Signup Form:
|