Twenty Wins Count Double
When I was laying in bed one night, I thought of a way to measure reliably the impact of 20-win seasons on Hall of Fame selection. To state that more carefully, I thought of a way to measure reliably the impact of 20-win seasons compared to wins in Hall of Fame selection.
We start with the assumption that the only thing that counts for Hall of Fame selection is Wins. Nothing else matters; ERA, Cy Young Awards, Pitching for Championship teams, throwing no-hitters; none of that counts. We are measuring simply the extent to which the career win total, by itself, predicts Hall of Fame selection.
Always another phrase to be added. We are measuring the extent to which the career win total, by itself, predicts Hall of Fame selection among pitchers who won 150 to 299 career games, beginning in 1900 and ending no later than 2015. There are 192 such pitchers. 43 of those pitchers are in the Hall of Fame; 149 of them are not. Within that group of 192 pitchers, the Win total, by itself, is 82.79% accurate in predicting Hall of Fame selection.
Now what exactly does that mean?
I’ll get back to that.
The parameters of 150 to 299 wins are not arbitrary. We are studying the impact of 20-win seasons for pitchers within the range of wins in which some starting pitchers are in the Hall of Fame, and some are not. Since all pitchers with 300 or more wins are in the Hall of Fame, and no starting pitcher with less than 150 wins is in the Hall of Fame, including those pitchers in the study would not add ANY information to the study. It would change the numbers; it would change the percentages, etc., but it would not change the conclusions in any way; it would merely be adding "dead data", useless data, into the study.
If two players are compared and the pitcher who has more Career Wins is in the Hall of Fame and the other pitcher is not, then the system is RIGHT.
If two players are compared and the pitcher who has more Career Wins is NOT in the Hall of Fame, but the other pitcher is, then the system is WRONG.
If the two pitchers are both in the Hall of Fame, or if neither one of them is in the Hall of Fame, then the system is neither right nor wrong.
"The System" at the moment is just the pitcher’s win total, but please accept that; it becomes a little bit of a system later on. Call it the process if that works better for you. The system has a right-wrong record in regard to any pitcher. With regard to Tommy John, for example, the system’s right-wrong record is 0-43. There are 43 Hall of Fame pitchers in the group, all of whom have fewer career wins than TJ, so the system is wrong 43 times, and silent in regard to the other 149 pitcher-to-pitcher comparisons.
Dizzy Dean had 150 career wins but IS in the Hall of Fame, so with regard to him, the system has a right-wrong record of 0-149. There are 148 pitchers in the study who had more wins than Dizzy Dean but are not in the Hall of Fame, and there is no starting pitcher within the study who has fewer wins and is in the Hall of Fame, so the system is 0-148. There is also another pitcher who had 150 wins, the same as Dean (Rube Benton). When two players are the same but one is in the Hall of Fame and the other is not, the system assumes that they should also be the same in regard to Hall of Fame selection, therefore the system is wrong in that case, so 0-149.
The leading winners within the study are Tommy John, Bert Blyleven, Robin Roberts, Ferguson Jenkins and Jim Kaat Blyleven is in the Hall of Fame, so comparing him to the 149 pitchers who are NOT in the Hall of Fame, the system is 148-1. He has more wins than 148 of those pitchers, so within those comparisons the system is right in selecting the one who is in the Hall of Fame, but he has fewer wins than Tommy John, so the system is wrong in that case. 148-1.
Robin Roberts is also 148-1 (that is, the system is also 148-1 in regard to Robin Roberts), since there are 148 pitchers with fewer wins than Roberts who are NOT in the Hall of Fame, but one pitcher with more wins than Roberts who is also not in the Hall of Fame. Ferguson Jenkins is at 148-1, and Jim Kaat is 148-1.
Addie Joss, with 160 wins, is 22-127, since there are 22 pitchers with fewer wins than him who are not in the Hall of Fame—meaning that the system was right in comparing the two and choosing between them—but 127 pitchers with more wins than Joss who are not in the Hall of Fame, either. Sandy Koufax is 35-114.
Again, not judging the rightness or wrongness of THE HALL OF FAME SELECTIONS; merely evaluating the rightness or wrongness of Career Wins as an indicator of who WILL be in the Hall of Fame. But taking all 192 pitchers, the system is right 10,609 times and wrong 2,205 times, for a right/wrong percentage of .8279.
To take the case of Rube Waddell, who is in the Hall of Fame with only 193 career wins (some sources use a different number). . .but to take the case of Rube Waddell, in regard to him the system is 89-60—in other words, more right than wrong (.597). The logic of it twists and turns and is hard to keep straight in your head, and it raises a question: why is a pitcher whose win total is low for a Hall of Famer shown with a right/wrong percentage of .597? Since a pitcher with 193 wins is NOT usually a Hall of Famer, shouldn’t the system show that as more wrong than right?
But the overall percentage is .828. For Waddell, it is only .597; thus, Waddell is dragging the percentage downward. We’re not evaluating whether 193 wins is "usually" a Hall of Fame number. That’s what we do in lots of other studies, but not this one. In this one, we are evaluating the extent to which the win total is a controlling factor in Hall of Fame selections, which is a very different question. The fact that pitchers with 193 wins usually do not go into the Hall of Fame means that Waddell’s selection is dragging down the frequency with which Hall of Fame selection is based on the career win total. That’s what we’re studying here—not Waddell, or Koufax, or Tommy John, but the extent to which the career Win total dictates Hall of Fame selection. If there was no relationship between career wins and Hall of Fame selection, then the chart would center at .500, but since there IS a strong connection between the two, the chart centers not at .500 but at .8279.
OK, moving on now toward our goal of measuring the value of a 20-Win Season in determining Hall of Fame selection. Let us suppose that we add one "win" to each pitcher’s career win total for each 20-win season. In other words, Tommy John, with 288 career wins and 3 twenty-win seasons, is now credited with 291 wins—but Robin Roberts, with 286 wins but 6 twenty-win seasons, is now credited with 292 wins. That means that the comparison of Tommy John to Robin Roberts is no longer "wrong"; it is now right. That means that the system is no longer 0-43 in regard to Tommy John; it is now 1-42—and also, that the system is no longer 148-1 in regard to Robin Roberts; it is now 149-0.
With regard to Dizzy Dean, Dean’s four 20-win seasons move him from 150 to 154, moving him ahead of seven non-Hall of Famers, so Dean moves from 0-149 to 7-142, just based on the fact that we are crediting him with one extra career win for each 20-win season. Addie Joss, who also had four 20-win seasons, moves up by the same amount, going from 22-127 to 29-120. Koufax moves from 35-114 to 42-107. Dennis Eckersley, with only one 20-win season, moves DOWN from 105-44 to 101-48, while Dazzy Vance, with three 20-win seasons, moves up from 103-46 to 104-45.
Overall, however, many more pitchers "improve" than decline; rather, the system improves its sorting capacity in regard to many more pitchers than those on whom it loses ground. It’s actually remarkable how much more accurate the system becomes by adding just one win for each 20-win season. Overall, the system moves from 10610-2204 (.828) to 10726-2088 (.837). This is strong evidence that 20-win seasons are significant in Hall of Fame selection—not that we needed such evidence, but if we did, we now have it. If a pitcher with 220 wins and five 20-win seasons was not more likely to be elected than a pitcher with 220 wins and one 20-win seasons, then giving weight to the 20-win seasons would not improve the accuracy of the system.
Well, suppose that we credit the pitcher with TWO extra wins for every 20-win season?
Once we do that, Tommy John is passed not only by Robin Roberts, but also by Ferguson Jenkins. Fergie had 284 wins and seven 20-win seasons, so he is now up to 298, the same as Robin Roberts, while Tommy John is at 294. Thus, the right-wrong log for Tommy John, which started out at 0-43, is now 2-41. Dizzy Dean, with 8 credited wins for four 20-win seasons, has now passed 13 other pitchers, making his contribution to the count 13-136, rather than 0-149. Addie Joss moves up to 36-113, Koufax to 45-104. Rube Waddell moves up to 103-46. Bob Lemon, who started at 114-35 but had seven 20-win seasons, is now up to 129-20, passing fifteen different pitchers who had more than 207 wins, but less than seven 20-win seasons.
Overall, with the addition of two points for each 20-win season, the won-lost percentage of the system moves up to .845. This proves that the value of a 20-win season in Hall of Fame selection is equal to at least two wins.
Well, what if we give the pitcher three points for each 20-win season? That increases the accuracy of the system to .852.
If we add 4 points for each 20-win season, it increases to .861.
At +5, it goes to .866.
At +6, it improves to .869—the upward pressure easing, but still moving upward a little.
If the weight given to a 20-win season is greater than the importance of 20-win seasons in Hall of Fame selection, then these numbers would go down, the system getting less accurate with additional weight. So what IS the point of maximum effectiveness?
It’s 24 points. I’m going to say that it is 20, for convenience—20 win seasons count double—but it is actually 24. The data levels off at 15 wins, but makes small increases until 24 wins:
+15 .8925
+16 .8923
+17 .8932
+18 .8948
+19 .8961
+20 .8959
+21 .8961
+22 .8957
+23 .8965
+24 .8970
Above 24 points for a 20-win season, the system begins to lose accuracy, just a little bit; at +30, it is back where it was at +15, still essentially the same but losing a little bit of accuracy with each step. So if we SAY that a 20-win season has double its win value in Hall of Fame selections, that’s essentially as accurate as saying 24 wins, but easier to work with.
To give you a couple of real-life examples.
Catfish Hunter had 224 career wins, which is not usually a Hall of Fame number, but he had five 20-win seasons. Giving him 20 points for each of those, that would put him at 324 "points". 324 points IS often or usually a Hall of Fame number—about half and half. Other pitchers with numbers around there are Red Faber (334), Ted Lyons (320), Stan Coveleski (315) and Jack Morris (314). Jerry Koosman had 222 wins but only two 20-win seasons; Tim Hudson had 222 wins but only one 20-win season. Frank Tanana and Dennis Martinez had 240+ wins apiece, but no 20-win seasons. Those pitchers move from equal or ahead of Catfish to far behind him, as Hunter is propelled upward by the adjustment.
Bob Lemon had only 207 career wins, but had seven 20-win seasons. That puts him at 347. 207 wins is not usually a Hall of Famer—but 347 almost certainly is. The only guy around 347 who is NOT in the Hall of Fame is Tommy John.
Moving the article now into it’s closing phase; I still have about 10 points to make, but they are disorganized.
*I have worked for more than 40 years to find "weights" to put on Hall of Fame type accomplishments. It is something that has always fascinated me, but it’s mostly guesswork; systematic guesswork, organized guesswork, but .. . .making things fit. This is the most scientific approach to the problem that I have found. That raises a question: could we use this more scientific approach on other, similar problems?
Well, yes and no. You COULD, in theory, use an approach like this to blend in strikeouts, shutouts, wins above .500, Cy Young Awards, Runs Saved or other pitcher’s accomplishments, moving us from 90% accuracy to 95%, maybe 99%, I don’t know.
You could do it, but it wouldn’t be quite the same. Whereas every starting pitcher piles up wins and almost all the Hall of Fame candidates win 20 games, not everybody wins Cy Young Awards. That makes it different. Strikeouts, you’d have to use decimal-point weights for each strikeout. That makes it different.
On the more general point, this works as well as it does for Starting Pitchers because starting pitcher selection to the Hall of Fame is dominated by one category: Wins. Selection at other positions is not dominated by a single category to the same extent, so you wouldn’t be starting at 83%; you’d be starting at 60% or 65%, building upward from there. That makes it different.
You could do hits plus home runs; that is, you could start with hits as an indicator of Hall of Fame selection, and then measure the value of one home run compared to one hit. But you wouldn’t be starting at 83%, because hits are that highly determinant or Hall of Fame selection.
We want EASY ways to judge where a player is in the Hall of Fame chase—simple, straightforward ways. The more things you study, the less straightforward the calculation becomes, so the less useful it becomes.
**I see now an option that I didn’t see until I had finished the study. I expected the weight to be given to each 20-win season to be in the range of 5 to 10 points; I didn’t expect it to be 24. But now that I know that it IS 24, or 20 let’s say, here’s what I could have done.
I would bet that this would work as well as it does here, or better, if, instead of counting 20-win seasons and multiplying by 20, we gave each pitcher one point for each WIN in a 20-point season. In other words, Jack Chesbro gets 41 points added for his 41-win season, Ed Walsh gets 40 points for his 40-win season, Dizzy Dean gets 30 points for his 30-win season, and Sandy Koufax gets 27, 26 and 25 for his three Cy Young seasons. Koufax and Tommy John each won 20 games three times, but Koufax won 78 games in those three years; John won 63 (22, 21, 20). I would bet that giving THOSE as the bonus points would work as well or better than the method I tested here. But testing that theory would take several more hours of work, and I kind of need to move on.
***You might be tempted to read this as "Bob Lemon had 207 career wins, but in the Hall of Fame process it counts as 347 wins, because each 20-win season is worth another 20." That’s not exactly true, because including these points changes the standards. The pitchers in this study averaged 197 actual wins, but they average 238 points—a 20% increase. The "absolute" standard for a Hall of Fame starter is 300 wins, but it is more like 360 "points". So Bob Lemon’s 347 points are not equivalent to 347 "wins"; they’re actually equivalent to about 285-290 wins.
****You might be tempted to read this as "20 win seasons have a large impact on Hall of Fame selection because 20 wins are magic number for pitchers." That is SORT OF true, but it’s not QUITE true. It is likely that if you picked some non-magic number—let’s say 17 wins. It is likely that if you did the same study with some non-magic number, you would still find that there was an impact similar to (but smaller than) the impact we have measured here.
We are measuring two things here. One is the "magic number effect"—which is not a legitimate way to evaluate players—but also the "big seasons effect", which is a legitimate Hall of Fame consideration. We’re giving more credit to pitchers who have big seasons; we’re making 24—11 count more than 13-6 plus 11-5. That’s legitimate; big seasons are what win pennants.