Joe, for some reason I didn't get your registered voter email....

No problem KJOK, I'll add you. Matt, what does 'nm' mean? I assume you want to be added?

Sorry for the confusion. I put my e-mail in the box thinking that was enough, the "nm" ment no message.

I thought I had registered already.....

Add me in.....thanks

I hope I am missing something here. I seem to be confused on the players that are being posted. As I understand it, these are the players appearing on (i.e., eligible for) the first HOM ballot. Also, as I understand it, the first ballot "year" is 1915, and with the 5-year waiting period, all players whose careers ended in 1910 or earlier are eligible.

If this is correct (and maybe I am missing something), then I would have expected more turn of the century players. Guys like Jake Beckley, Cupid Childs, Jimmy Collins, John McGraw, Lave Cross, George Davis, and Herman Long who, I think, all ended their playing careers in 1910 or earlier.

Maybe there will be a second "batch" of players posted later that are also eligible for the first ballot. Or maybe I have completely misinterpreted what is going on; if so, I apologize.

Rob:

Rob -- Originally, the ballots were going to start in 1915. But we realized that if we started with players whose careers were finished by 1900, we'd still have 30 years of major league ball (if you count the NA as major) to go on, or basically the first 2 generations of players.

We, or more specifically RobertDudek and I, felt that it was important to make sure several of these players get in, before they have to compete with the Ty Cobb's and Honus Wagner's of the world. So we pushed the first election back.

Just to clarify -- It was a voter that actually suggested moving the elections back, I meant to say RobertDudek and I agreed. It didn't come out like above for some reason.

Thanks for the replies. It took me awhile to notice that there were virtually no players posted who played in the 20th century!

On another topic that was raised in one of the position threads (on Joe Start), I have a few meta-questions. Is a voter allowed to consider non-professional accomplishments, such as pre-1871 performance? Second, how much weight should be given to being a pioneer, such as introducing new strategies into the game or being instrumental in new leagues? Third, is it acceptable to shade one's views in favor of players who were captains/managers of their teams? I guess I am seeking more guidance on what exactly are we doing here. Thanks much.

I believe scruff said somewhere that we were pretty much limiting ourselves to on-the-field accomplishments. I think there's a "statement of purpose" around somewhere. As far as managers go, there was talk of doing seperate managerial elections every 5 "years" or so. Again, I'm not sure what the final decision was on that.

On another subject, scruff, when you're putting the profiles of the various players up, could you include a link to their b-r pages? It's always nice to have numbers to throw around in the arguments.

Devin -- Duh!!! What the hell was I thinking? BR Links will be up as soon as I get a chance. I can't believe I didn't do that.

"What is the relationship between pitching and defense in 19th century baseball?"

To me, that is the fundamental question each of us has to answer first before we can rationally evaluate these players. I see the question having two possible answers:

1) It's pretty similar to today. Our modern tools for evaluating these players will work in about the same way. However, the conclusions are startling. The pitchers are doing the work of complete pitching staffs (1870's) or half of one (1880's). We can compare the star players behind them all we want, but these guys are supernovae. At their best, they are Randy Johnson pitching every day; Maddux and Glavine each doing half of the workload. When you prorate their stats to 162 games, the numbers become even more surreal; Radbourn's peak 3 win shares for 1882-84 are 50, 60, 89; scaling his team to 162 games gives him 96, 99, 129!. They rack up potential HOF careers in 3 or 4 years and then flame out. A star like Anson piles up impressive career stats, but when comparing him with Radbourn, its like the Sutton/Koufax quantity/quality debates.

2) It's not similar to today. Our modern tools for evaluating these players will not work unless somewhat modified. The pitchers do not have the impact that are otherwise deduced. However, there is a corollary to that; if the pitchers aren't winning that share of those games, then the defense is picking up the slack. For example, win shares uses a roughly 2:1 ratio between pitching and defense. If the pitcher's value is cut in half, then the defensive values must be doubled to maintain the overall totals. For 1Bmen this won't mean much; for SS and 3B, this can be huge.

In other words, either pitchers are the super-stars or defense is more important than today. I have no answer to this question, and, unfortunately, no time to devote to its study. I suspect each of us will come to a personal decision on this, and as a result, the first election(s) may be the most controversial.

We have to discuss and address this issue while evaluating the individual players.

(Apologies for hijacking the thread, but it looks like the pitchers will be last, and I think this issue is important enough to start thinking about it now.)

The statement of purpose that I read did not address my three questions. Speaking with personal experience of the Baseball Survivor group, I cannot over-emphasize the importance of resolving (to the extent possible) these meta-issues up front. If they are not resolved, they will be revisited ad nauseum as we go forward, and likely some people will have their feelings hurt one way or the other.

My third question pertained to playing captains of the 19th century. As we know, most teams didn't have managers (in the modern sense). A player, typically the best player on the team, was the captain and formed strategy, lineups, etc. I am very willing to give these captains extra credit, especially on successful teams. I guess this can be considered "on-field performance", although much of the contributions are from their teammates.

My two cents on the other issues. I think we need some sort of statement defining who we are trying to honor in the HOM. Something like "the best players of major league baseball played in North America from 1871 to the present day". Thus, Negro Leaguers would be eligible, but not Japanese players (in Japan). Non-professional before MLB 1871 would be eligible, but they would be worthy only if they made significant on-field contributions in post-1871 MLB.

Regarding pioneers, I would extend them a small amount of extra credit, but not enough to elevate a marginal candidate into the HOM. That is, there would be no Homers based in any substantial manner on their pioneering activities. Of course, this is a grey area on which individual voters will have to exercise their best judgment, but I think a general statement up front would be a good idea.

JimD raises the most meaty of all 19th century issues. I think most of us would agree that the split between offense/pitching/fielding in the 19th century probably bears little resemblence to today's game. There are analytical methods that can shed light on this important issue. I hope to look into this issue soon, but I am in the middle of some other research that I have to wrap up first.

One more procedural question. Once we get started, how frequently do we anticipate the "annual" votes will take place? Is it on the order of once a week or so? Back of the envelope math: there will be 97 "annual" votes to cover the 1906 election through the 2002 election. 97 weekly votes is a little less than two years. Does this seem about right?

One more procedural question. Once we get started, how frequently do we anticipate the "annual" votes will take place? Is it on the order of once a week or so? Back of the envelope math: there will be 97 "annual" votes to cover the 1906 election through the 2002 election. 97 weekly votes is a little less than two years. Does this seem about right?

Still another issue that should get an airing is the practice of scaling up 19th century players' stats to 162 game season equivalents. I am all in favor of trying to put 19th century players on a level playing field with their 20th century brethren. However, short seasons can lead to small sample problems, which may become apparent when we scale up performance proportionately to the number of team games.

Of course, if a team only played a handful of games in a season, we would be very cautious before we would scale up all performances to a 162-game schedule. Well, the same concerns may apply to schedules of 80 or so games. If a batter hit .400 over the first 80 games of a season, we'd not jump to the conclusion that he'd hit .400 over a full 162-game season. The same argument applies if the season itself is only 80 games long.

There is no perfect way to do it, but I would propose using results from sampling theory. The confidence (standard deviation) of our estimates of a player's "true" performance over N games is inversely related to the square root of N. By using the binomial distribution for batting average, say, it is straightforward to derive the relevant factor of how much we need to scale down our per game multiplier to reflect our greater uncertainty. The factor depends upon the number of games played, the player's observed performance metric, and the degree to which we want to "shade" our estimates of a player's "true" performance.

For most reasonable values of the above parameters, I find that the multiplication factor should be about 96-98% of the straight-proportional scaling, even with as few as 80 games in a season. Thus, I feel very comfortable with the straight-proportional scaling of Win Shares, with the proviso that maybe around 3% of a early player's win shares are a little suspect. For players who played significant time on teams with far fewer than 80 games in a season, a more explicit shading is probably appropriate.

Very good points, Rob Wood. The National Association stats are going to exhibit this problem to an even greater degree. During the first season, teams played around 30 games against each other. (This starts as a loose association of the best barnstorming teams playing for bragging rights; they played many other games with amateur and semi-pro teams of varying quality, but the stats are for the NA games only.) The seasons increased in length each year, but 30 games is a very small sample (one reason why Levi Myerle hit .492; he was one hit away from being the first and last .500 hitter).

Using the same methodology I utilized for an 80-game season (leading to an estimate for a shading factor of 96-98%), I find that a 30-game season should be discounted at a 93% rate. That is, under my assumptions it would be reasonable to scale a player's performance (win shares) achieved in a 30-game season up to a 162-game season proportionately, and then multiply by 93% (i.e., discount by about 7%).

At first I was surprised how small the discount is under this method (I expected the shorter seasons performances to be more suspect). The reason that the proportionate method is fairly robust is that the "event" that gives the N in the square root of N term is the number of plate appearances, not the number of games. In my formulas I assumed an average of 4 plate appearances per game. 30 games is around 120 plate appearances, which is not quite such a small sample as 30 plate appearances would be. (The multiplier for 30 events vs 162 under my assumptions would be 86%, but this is not correct since events are plate appearances, not games.)

Rob,

Granted, my statistics is a little fuzzy, and hazy, but your 7% "short season discount" strikes me as an odd way to look at him.

If Player X has 50 WS (when scaled to 162 games) in a 30 games season, one way to look at it is to say that we are only confident of this result to within one standard deviation, and therefore we will discount by 7%, and say that he was with relative certainty at least a 46.5 WS player that year.

That method -- it seems to me -- will systematically undervalue the short seasons, though. Standard deviations go it both directions, which means that Player X would have been just as likely to be a 53.5 WS player as a 46.5 WS Player. It would make more sense to me to present a range ("He was a 46-54 WS player in 1871"), and decide which is more likely by extrinsic evidence (team's pythagorean W/L, stats in surrounding years, strength of the team's pitcher, etc.)

Also, another criticism of the 162 game scaling (going in the other direction) is the idea that a player could do in 162 games anything like what he did in 30. Would a pitcher throwing twice a week for 15 weeks be as effective as the same pitcher throwing every day for two weeks straight? Does a player who plays one game on a gimpy ankle still play if he knows he won't have four days to recover before the next game? If a modern day closer can do what he can do one inning a game, should we multiply his numbers by three or four so he can be compared to starters? Do we multiply the righty side of a platoon by three to see what he'd do against right and left handed pitchers? Of course not. So why assume that players would perform the same in games whether they played in them or not?

If David Segui only good for 50 games per year, why should be get more credit for playing those 50 games in 1876 during a 50 game season, rather than in 2001, where he's got to go on the DL every other week? If Roger Clemens and Bobby Mathews both start 36 games per year, why should Mathews get more credit for being in a larger percentage of his team's games?

Don't worry about comparing players from then with now. It's not supposed to be a "time capsule." The idea was to put the 19th century players in a way that we are used to. There is going to be problems with any statistical setup. That's why it's good to have a forum that we can discuss the players: there are different ways to look at the same picture.

John,

It was more of an analogy than an attempt at comparison.

We adjusting to 162 games to make the variously-lengthed seasons comparable. But it's only comparable if there is an assumption that a player can do the same thing over 30 games that he can over 60 or 90. My initial assumption is that it's easier to do something twice a week (1876 stats) than it is to do it nearly every day (1899 stats).

But providing stats normalized to 162 games essentially assumes that a 10 game winner in 1876 equals a 25 game winner in 1899. And I think that degrades the later accomplishment.

John,

It was more of an analogy than an attempt at comparison.

We adjusting to 162 games to make the variously-lengthed seasons comparable. But it's only comparable if there is an assumption that a player can do the same thing over 30 games that he can over 60 or 90. My initial assumption is that it's easier to do something twice a week (1876 stats) than it is to do it nearly every day (1899 stats).

But providing stats normalized to 162 games essentially assumes that a 10 game winner in 1876 equals a 25 game winner in 1899. And I think that degrades the later accomplishment.

I acknowledge that my "back of the envelope" method is just that. It uses one standard deviation of a player's accomplishment in N games as a proxy to the discount we should apply when we scale up his accomplishments to 162 games. I could not think of a better way to look at this issue. Of course, I encourage others to chime in as well.

I do have one comment on the scaling up. Since we are all well aware of 20th century win share figures (seasonal and career), it will be impossible to look at a 19th century player's scaled up win shares and not mentally compare (or judge) his accomplishments in light of our more well-known 20th century standards. Don't get me wrong, I think this is a good thing. But we should readily admit win shares scaled up to 162 game seasons will carry an added punch that may not entirely deserve.

Do not make the mistake of thinking that these are the only games that these teams are playing. They are not playing for a couple of months, or playing a couple of games a week. Their touring schedules include many exhibition games against other teams of various quality. 19th century fans apparently enjoyed seeing these pro teams beat the stuffing out of their favorite local team, giving them a real taste of the difference in quality between the teams.

Harry Wright's Cincinnati Reds of 1869 were the first openly professional team, acknowledging that all players were paid (not fake "amateurs" like the best players on some other teams), and issued a challenge to take on all comers. They barnstormed across the country from Boston to New Orleans to San Francisco and were undefeated in sixty-something games. Scores ranged from 4-2 against the top-flight New York Mutuals (Boss Tweed's team of New York city

>> From what I understand his numbers for the 1874 NA say something like 77% defense, as opposed to the assumption of about 32.5% defense Bill James works with in Win Shares (which is pretty valid for modern times).

This I would like to hear more about. This suggests that the pitching portion of a pitcher's WS will be cut down by factors of 2 to 3. (The hitting portion will stay constant, and pitchers did make significant positive contributions as hitters; Spalding led Boston in plate appearances in 1874, suggesting that he may have batted leadoff at least part of the year; it's certainly doubtful he batted in the 9-slot.)

I would also point out that it means that defensive WS will go up by a factor of 2 or more. The position players WS/G will then increase, perhaps dramatically for the C/SS/3B fielders.

Matt B:

The easiest way to correct the problem is to find out the standard deviation of Win Shares per season.

My point before was in regard to the Hall of Merit only. We should be comparing players from roughly the same time period, so we shouldn't (if we elect wisely) have to worry about comparisons with today's players. I don't think we'll have a situation where a player has to wait over ninety years (such as Bid McPhee) before he's elected.

Joe wrote:

"As for the procedural things -- Rob I agree weekly will probably be the way we go.

After the first election we'll only have to add the new eligibles to our assessment/ballot, so we'll all have our personal rankings, and we'll just need to slot the new guys. Of course our personal rankings may shift as discussion continues, etc. But the really heavy work will be in sorting this first group out, after that it's evaluating how the top new candidates compare to the established eligibles. So I think one week is plenty of time for this."

Gee. Weekly seems pretty often. If I submit my ballot on Friday, how long will it be before you've tabulated them all to see who has won? Can you turn it around in a day or two? You can't start thinking about the next ballot until you know who will be eligible.

Assume a two day turnaround (which seems pretty ambitious to me). That only leaves five days to consider new entrants, re-slot the remaining eligibiles, and re-submit your votes. Plus, if I recall, you have wanted us to include explanations along with our ballots. I hope no one's planning to go on vacation for the next two years!

I would assume that most people will devote most of their attention to the top positions in their ballots, without worrying whether a guy is really the sixth or seventh best candidates. It is only as they move up the list that more critical analysis is necessary. Same thing with positions. If my favorite shortstop hasn't made it in yet, I'm not going to worry about where I put my second favorite on the ballot as much. Once he goes in, I would want to re-examine my second favorite shortstop (maybe 8th on my last ballot), and pay more attention to where he should be placed against other players.

Plus, if my favorite player, who I think is a shoo-in, turns out to be far lower on the ballot than I expected him to be, I'm going to want to marshall some arguments in favor of him. But if half the ballots for next year are submitted by the time I can do any further research, it won't really make a difference.

I'd advocate (at minimum) a week of discussion and politicking, followed by a week of open polls for balloting.

Sure, it would take longer, but as time goes on, and we're no longer picking the best of the best in each election, the gradations become finer and involve closer inspections.

I agree that two weeks is much more sensible!

Agreed Matt, two weeks is much more reasonable. I forgot about the time it will take to actually add up the ballots. Four years seems so far away. It's a whole bachelor's degree or Olympics from now . . .

On the other hand, it'll be just in time to start ranking Gwynn, McGwire, and Ripken on our ballots.

When the pitching WS are posted, would it be possible to see both the "raw" WS (Bill James formula) and the adjusted WS numbers? It'd be nice to see how large the adjustments are.

An example of how the pitching/defense balance will affect the discussions within just the position players. I know that career totals aren't everything, but here is a top-10 position players based on the career Win Shares posted so far. (I know, the NA is not yet included in these numbers.)

567 1B - Anson

Concerning the split between offense and defense in 19th century baseball, I am confident that the "intrinsic" weights allocated to the defensive positions are quite different than they are today. I imagine that the third baseman, left fielder, and first baseman are more important in 19th century ball. Second baseman, right fielder, and center fielder are probably less important compared to today.

Is this a settled issue? Does Bill James adequately address this issue? Are we planning on using the defensive win shares from James, and then scaling them up to reflect the additional importance of fielding (relative to pitching) in 19th century ball?

To my knowledge, Bill James has shown little interest in 19th century baseball. He runs the numbers through his formulae and then says "I don't trust these results", leaving it up to others to explore where he doesn't care to. Which of course is fine; he's not our employee.

All I'm attempting to do is dramatize the situation by doing crude back-of-the-envelope adjustments to show the magnitude of the problem by demonstrating potential impacts. As I've stated before, I don't have the time or resources to do the real work, though I'm extremely interested in the results.

There's been talk of presenting "adjusted" pitching WS. This is fine, but if pitching has less impact, then fielding must have more, and those must be presented too.

I don't doubt that the relationships between the positions have changed. James discusses the shift in the defensive spectrum, 2B vs. 3B, in his Win Shares book, because it occurs during the 30's, part of the "modern" baseball he cares about. I am fairly certain that there is a shift like you mention, LF vs RF, during the 19th century. Before substitutions were allowed (without the consent of the opposing team), the backup pitcher often played RF in case he was needed; presumably he wouldn't have to make too many plays, but could catch and make a strong throw when necessary.

Radical thought...

What if we decided to vote every two years, instead of every year, and doubled the number of players elected off each ballot?

This would result in a much quicker time frame overall, white retaining the moving-through-the-eras approach.

Thoughts?

Craig,

If the intent is to induct up to five people in later years, doubling the numbers would require us to vote for ten people, where the top ten would be inducted.

I don't know how many people are expected to vote, but I'm guess that, if its around 100, it would only take a handful of ballots to get a guy into the top 10.

Increasing the number of inductees allows a more marginal candidate to slip in if he has a small contingent of supporters. Assume 10% of voters love Ross Barnes, and the other 90% think he was a NA flash-in-the-pan. Ten voters give him first place votes, the other 90 don't vote for him at all. That might be enough to get him onto a top 10 list.

Is a guy a HoMer is 90% of the voters think he wasn't worthy of even a tenth place vote?

As it is, the voting process is subject to the Palmiero-gold-glove fiasco where the vote was split 15 different ways and a marginal candidate slipped in with a bare plurality.

It's been scientifically proven (economist named Arrow, I think) that anytime there is more than 2 options, there is no such thing as a majority will. Every vote is biased in favor of how the election is structured, and there is no way to "fix" the problem, because there's no right answer.

The only defense against this kind of problem is, generally, consensus. I am assuming that, for the most part, the people who are inducted will generally be on a majority of ballots somewhere. I may think Roger Connor is better than Cap Anson, but if Anson gets in first, and I only have him, say, seventh on my ballot, I don't really have a valid complaint.

If the top 6 or 8 or 10 guys on the ballot get in, chances are a sizeable chunk (sometimes even a majority) of the inductees will be people I (and everyone else) didn't consider a top 10 candidate.

Personally, I'm opposed to letting anyone other than the top candidate in, because it increases the risk of a marginal candidate with a small, strong following becoming chosen (I'd prefer two elections where only the top candidate gets in to one election where the top two get in.) That, however, is probably not practical.

Generally, a Barnes-like figure who may end up number 10 in total votes, may not end up in the top one or two (or even top five) after 10 individual votes. As Bill James wrote re: Palmiero, have a wide enough field of candidates, and eventually David Duke will be elected to something. The fringe Klansman will get his 12% no matter who he's running against, and the 8 moderate candidates will get 11% each from voters who would put the the other 7 moderates ahead of Duke (their 8th choice).

Very good points Matt. would adding more people to the ballot (say voting for the number elected, plus 10, instead of just 10) help or make it worse? I definitely want to avoid these types of problems.

Jim -- you are definitely correct about RF vs. LF. Robert has done replacement studies, and RF was worse than the modern DH. Not so much on the 1B side, although it was a little better than today, IIRC.

I think a big reason for this is that pitchers weren't throwing 98 and almost everything was pulled. Most of the hitters were RH, so naturally there are going to be a lot more plays for the LF/3B/SS.

I think WS will take this into account though, because if the LF's are making more plays, they are going to get more WS. James considers OF one 'contingent' in splitting out defense and then gives credit to the individual fielders based on the number of plays they made (minus adjustments for errors, assists, etc.). I don't think it's an issue.

Also for 1B he estimates unassisted putouts, etc. If they were making more plays, they'll get more credit. It's a pretty solid system in this respect, lots of self-correcting things like this.

If someone has the time (I don't right now) take the LF's and RF's and divide their defensive WS by their season numbers quoted. I think you'll see the LF's come out significantly ahead. The numbers will be a little soiled by players playing other positions, but that's just another sign that the players were better fielders.

Matt -- would designing the ballot something like 15 people (first year we elect 5), with points going 7-6-6-5-5-5-4-4-4-4-3-3-3-3-3 work better? Getting on the ballot would be more important than where a guy finished, and thus make it harder for a strong minority to push a candidate through.

It would also make it easier to fill out your ballot. You don't have to decide if Roger Connor was better than Dan Brouthers. Vote them 2-3 and they each get 6 points. Or vote them 5-6 and they each get 5 points. That style ballot also shows the extreme end of the bell curve pretty well. I really like this idea, please point the flaws I'm not seeing out. If not, I'm strongly recommending this is how we do it. I don't remember this suggestion on the old threads.

I'm going to set up a few different threads tomorrow if I get a chance. I'll make one for ballot construction. One for a managers/execs/contributors wing also, since I've gotten some email on this. Any other suggestions for threads?

First, you have to decide whether there is a problem. Some earlier comments suggested that we allow people to have extra votes to scatter around their ballot. Extra votes would exacerbate the "vocal minority" problem, but would decrease the "strength of feeling" problem where people with a (say) 7-6-6-5 ballot don't feel that a first place vote adequately expresses how strongly they feel that the candidate was the best (or, by how much he was better than the second place guy), so want some extra votes to show how quantity of feeling.

Neither is the right answer, you just have to know that any choice involves trade-offs, and you have to decide which trade-offs you are making.

Scruff, your suggestion really had two parts: (1) change to ballot structure to 7-6-6-, and (2) increase the ballot to 15 people. I will address each of them separately.

Assume: 100 voters

Note, my link above goes to the 1987 NL MVP election results. Scroll up for AL results.

Looking back a little further, there were several players who got fifth place on the MVP ballot with about the same percentage of the maximum on a 10 point scale as David Justice. The lowest I found was Gary Matthews, who received 70 votes (none for first) in 1984. That's less than 30% of the maximum 240 on a 10-9-8 scale.

http://www.baseball-reference.com/awards/awards_1948.shtml#NLmvp

Wait, I spoke to soon. 1948, Harry Brecheen, who I have never heard of, finished fifth with 61 votes out of 24 ballots, or about a quarter of the maximum on a 10-9-8 scale.

I'll stop now.

Nice work MattB.

I just want to point out that the flatter you make the points distribution, the more the ballot resembles the HOF ballot (completely flat), but without the 75% minimum requirement which insures consensus. Whether or not this is a good thing depends on your perspective.

I would prefer a system which would prevent an organized, vocal minority from imposing their will on the majority. What if we combined a 10-9-8 scoring system with a 51% minimum requirement?

I'm new here but enjoy the discussion. As to ballot configuration, somewhere along the way I saw the the basic idea of the MVP-style ballot was that people (we) are familiar with it. A 10-9-8...2-1 configuration, yes, a 7-6-6-5-5-5...well, clearly that is not a familiar format. Not to mention, half the fun is HAVING TO DECIDE whether Roger Connor is better than Dan Brouthers. Would it make more sense to stick with a 10-9-8...2-1 format with a minimum percent needed to qualify rather than a fixed number of winners? Consensus is a good thing, and so are appropriate standards.

As to adjustments to pitching and defense, I saw the adjustments for defense and frankly my gut tells me that McPhee and Glasscock probably should rate closer to Connor and Brouthers rather than way behind. But I didn't see whether the adjustments to the pitching WS were proposed as halving the raw WS or the adj WS. The latter seems right. Take John Clarkson 1885 for example. Raw WS is 62, adjusted to 162 games is 89, then half of that is 44.5? Is that right? 44.5 sounds better than 31 given two man rotations. Clarkson pitched over 600 innings in just 113 (team) games, worth two-fifths of a modern rotation, plus of course he was effective as hell (adj ERA 162). So 31 doesn't do his achievement justice. Do I have this right? Thanks.

To keep ourselves from getting confused (and I probably am one of the sources of confusion), I propose the following terminology.

BJWS: "raw" Win Shares, calculated using Bill James' formulae as published.

AWS: "adjusted" Win Shares, takes BJWS and adjusts them to a 162 game season. Important to normalize 19th century due to constantly changing schedule lengths. When we get to 154 game schedule, this adjustment will act like a 5% sales tax.

MWS: "modified" Win Shares, recalculates WS using modified formulas to change pitching/fielding balance.

So what I meant by my post of "8:06 p.m., July 17, 2002" was: that I'd like to see the AWS (to compare with the AWS published on the position player threads), in addition to the MWS (if that is what is indeed published for the pitchers).

ChapelHeel (Chapel Hill, NC?):

Maybe this belongs in the ballot structure test, but I have another procedural question. Now that we have decided to have pre-1900 players on the first ballot, have we decided how many people are going to be elected in each ballot? I see in Joe's original article that he proposed 5, 5, 4, 4, 4, 3, 3, 3, 3, 3, 2, 2, ..., but this was when the first ballot contained pre-1910 players. So I think we need to scale back how many are elected given that we'll now have 10 more elections. (Sorry if this issue has already been discussed/decided, but if so I cannot find the record of it.)

One reason that I am posting this question now is that it is somewhat related to the issues pertaining to voting tallies and possible minimum percentage requirements. Thanks.

One other thought on ballot tallying. The flatter the point system the less room there is for "strategic" voting, by which I mean giving a player more points than they deserve in order to maximize his chances of getting elected. Or, probably even worse, not voting for someone deserving in order to get your personal favorite elected instead.

Saying the same thing, a 10-9-8-7-6-5-4-3-2-1 ballot will encourage people to tailor their ballot in such a way as to try to have the greatest possible personal impact on the results. (The difference between 10 points and 1 point is very large, whereas the difference between 1 point and 0 points is negligible.) This is something that is very hard to avoid and something we should think about and discuss.

Concerns about ballot manipulation seem a little far-fetched. You're only going to have bona fide ballot manipulation on a border line candidate, obviously, so I did a little computation on that assumption. Say you've got a border line candidate--not a shoo-in but a consensus top ten. Say he is placed on every one of 100 ballots but in the following distribution from 1st to 10th: 0, 0, 1, 5, 9, 15, 19, 21,17, 13 = 358 points. Based on the David Justice discussion, this is worth 4th or 5th place at best, if all the other votes are widely distributed, but probably more like 6th or 7th place. Now let's say a minority of eight voters all place this individual first, and he gets one less vote each at 3rd through 10th. Now the total is 402 which probably moves him up one spot.

This analysis also suggests that getting 501 votes (50 percent plus 1 of maximum among 100 ballots) is pretty slim. Maybe that's good, maybe that's what is wanted. Chances of anyone getting 75 percent plus 1 is of course slimmer yet. I would guess votes on this first ballot will be pretty widely distributed.

So anyway, what about a run-off procedure? A nominating ballot to reduce from 25 or whatever to 10. With 10 then a 75 percent requirement would be fair. With 25 and no Ruths or Cobbs or Wagners, I'm not sure if you can get a 50 percent consensus, much less 75.

On AWS: I don't think anyone is making the assertion that these numbers are an "extrapolation" of the players performance. In other words, that if he had played 162 games, this is what he'd have done. (This is particularly ludicrous on the pitchers.) I believe the intent is to simply scale the numbers to a constant season length. This shows the impact of the players numbers, and makes comparisons easier. This is a "value" oriented number, as opposed to an "ability" oriented number, which is what you might get if you folded in factors like career rates, performance in adjoining seasons, pitching inning limits, etc. to get a "reasonable" extrapolation.

As an extreme example of this, consider Al Spalding in 1872. I have come up with an estimate for him of having earned approx. 51 WS in a 48 game season. This scales to 172 AWS for the 162 game season. Before anyone howls in protest, consider this. Spalding had an 188 ERA+, comparable to Randy Johnson last year at 184. Spalding also tied for 5th in RBI. Imagine having Randy able to pitch everyday and also hit like, say, Pujols (29WS); Randy pitched 1/6th of the Arizona innings and earned 26 WS. 6 Randys plus Pujols adds up to 185 WS (or 55 WS in 1872). The impact or value comparison is not that far off (particularly if you consider how crude my 51 WS estimate is).

I'm not saying that Spalding could do 172WS over a 162 game season. I'm saying that that number approximates his value to Boston in the 1872 season. (At least it does if pitching as important in 1872 as it is today.) There is a difference.

There is always the possibility of MVP ballot manipulation. One of the most famous is the AL MVP vote in 1947 where Ted Williams lost to Joe DiMaggio by one point. It was later discovered that one writer had left the Triple-Crown winner (Ted) off of his ballot entirely. A 10th place vote would have made it a tie.

The current HOF ballot is not subject to that kind of manipulation because it is essentially an independent referendum on each candidate. Is he worthy? Yes or maybe. (No becomes the alternative only when the player is in his last year.) The flaw is due to the restriction on the number of Yes votes allowed, and only comes into play when there are too many qualified names on the ballot (like in the 40's).

Sorry for cross-posting this (I just put it on the 1B discussion but note that there has been no discussion recently):

Where does Cal McVey fit in--I don't think I saw him as a C or a 3B or a 1B. He was a dominant player in the NA and deserves consideration to the same degree as Spalding, White and Barnes.

My comment re ballot manipulation was based in part on the assumption of 100 or so voters, significantly more than the 1947 MVP vote. I suppose a group of 10 to 15 could have the same impact here, but the analogy also fails in that in 1947 it was pretty obvious you had a two man race with only one winner. Here you've got a 25 man race with multiple winners.

The bigger question, I think, is whether we are indeed electing a fixed number or whether there is a fixed threshhold. (Has that been decided?) I'd rather get that right than worry too much about manipulation.

Another thing I worry about is that we all get carried away with our enthusiasm for this project, the result of that being that we have an unnatural desire to elect. Thereby lowering our standards.

Scruff's declared intention is to get a Hall the exact same size as the Hall of Fame's player-only representatives... which I believe is 215 by 2002?

I think this is a good idea, and takes care of the lower/higher standards question. I am concerned that we get the number of electees in each year right, or at least reasonable, since we can back-weight the numbers (older players are still eligible down the road since no one loses eligibility).

jimd mentioned that One of the most famous is the AL MVP vote in 1947 where Ted Williams lost to Joe DiMaggio by one point. It was later discovered that one writer had left the Triple-Crown winner (Ted) off of his ballot entirely. A 10th place vote would have made it a tie. In fact, three writers also left DiMag off their entire ballot, an equally ridiculous decision. But at any rate, I don't see how this is "manipulation".

There are two principles we are supposed (as I understand it) to be following here: one, "serious voters only, please!"; and two, you have to be willing to stand up and defend your decision. I think this takes care of the "manipulation" question. If you don't believe that someone is one of the ten best candidates, and you can and will defend that decision, then by all means vote that way and "manipulation" be damned. Likewise, if you think Ozzie Smith is #1 on your 2002 ballot, vote that way. "Manipulation" is what people shout when a ballot doesn't go the way they wanted (or more legitimately, when it's decided that we won't count ballots and we'll just "declare" a winner :)

On the other hand, if you won't defend the decision, or give spurious reasons... "I just don't think so" isn't good in my view... then we can throw those votes out.

Does that sound reasonable?

On July 18, jimd wrote:

"To my knowledge, Bill James has shown little interest in 19th century baseball. He runs the numbers through his formulae and then says "I don't trust these results", leaving it up to others to explore where he doesn't care to. Which of course is fine; he's not our employee."

Recently on the SABR-L board was this posting. I think it's pertinent to our discussion here:

"Subject: Re: Win Shares and the Hall of Fame

I want to weigh in with a proposal on how to structure the HoM.

It has been suggested that we use ?biannual? elections beginning in 1906 to reduce the time it takes to produce the HoM. That makes sense, up to a point. Here is my compromise to reduce the number of elections from 100 to 70.

How about we start with biannual elections from 1906 through 1966, our first 31 elections. In 1967, the HoF returned to annual elections, so I suggest we do the same at that point, through 2005.

A major aim of the HoM is to be able to contrast it with the HoF elections. In the end we want to have the exact same number of players as the HoF has. How many is that?

There are 189 men in the Hall elected for their play in the major leagues. Add to this 17 men elected for their play in the Negro Leagues, 3 pioneers (Spalding, G. Wright, Cummings) and two managers who may have been good enough to be elected as players (McGraw, Griffith). That makes a total of 211 players up to 2002.

My compromise will require 70 elections. Assuming biweekly elections, that covers 140 weeks (about two years and eight months) and takes us into April 2005. The Hall has now entered a period of fewer electees, so lets assume two players per year will be elected in the next three years. So I think we?re aiming for a membership of 217 players in the end.

How many should we elect each year? Someone pointed out the potential for lesser players making the HoM if too many were elected in one year. With this in mind, I suggest we elect no more than five in any one election. I also agree with the ideas embodied in Joe?s structure of electing more players at first (to take in the backlog) and in the end (to account for expansion).

Here is a proposed schedule for inductions:

1906-12 (4 elections)?5 each time?total 20 players

Having spent the better part of the last year of my life enthralled in the Baseball Survivor group exercise (weekly votes to determine the ranking of the game's all-time greats), I can personally attest that the "voting manipulation" issue needs to be addressed up front.

The number of voters has little relevance, nor does restricting the ballot to "experts". Nor does a requirement that you justify your vote. We don't really expect every voter to accompany each vote with a long message detailing his/her reasons, do we?

The point is that every ballot will likely allow each and every voter to potentially affect the outcome of the election (this is a good thing), especially if we use a wide point spread. Once you admit to that possibility, you automatically allow the opportunity for voters to misrepresent their views in order to effectuate an outcome they prefer. This is especially true if there will be a fixed number of selections each ballot rather than a minimum vote tally requirement.

As someone has already pointed out, this is related to a famous theorem in economics which essentially proves that there is no voting system that cannot be manipulated by misrepresentation of voter preferences. The upshot is that great care needs to be paid to how to structure the ballot, determine how the ballots will be tallied and the election requirements.

If anybody thinks this is not a potentially important issue to address, they haven't thought very hard about it.

"As someone has already pointed out, this is related to a famous theorem in economics which essentially proves that there is no voting system that cannot be manipulated by misrepresentation of voter preferences. The upshot is that great care needs to be paid to how to structure the ballot, determine how the ballots will be tallied and the election requirements."

Actually, quite the opposite. Since every voting system can be manipulated, I don't think it's really worth much thought. A 10-9-8 voting system is the most intuitive. For any other system, the question is, does it solve a problems without creating any new ones? I'd be dubious of any one that does. If I think Babe Ruth is a shoo-in, what will stop me from burying him 6th or 7th on my ballot to give more points to more marginal candidates that I like? Changing the voting scheme to 7-6-6 may solve that problem, but doesn't allow for strength of feeling among the mass of voters that won't act "strategically". If two candidates stand above the rest, why should one have to share six points with the greatly inferior #3 candidate. Maybe we should allow bonus points so we can structure the ballot based on our preferences . . .

It is just a fact that ballot structure will effect the results. You cannot decide on a correct ballot structure without first deciding what result you want. Want Ross Barnes in? Make voting 14-9-8-7.

Want him out? Have a run-off between the top two vote-getters in every election, and only let in the top one.

None of these formats are "wrong". They will just all give you different answers. John McCain won all the "open primaries" (where Democrats could vote) but lost all the "closed primaries". Different states had different rules that led to different results. As long as your choices aren't binary, there cannot be a "correct" majority will.

"Strategic" voting is not necessarily wrong, either. People often for for their second favorite candidate in order to avoid having their least favorite candidate win. ("I'd prefer Nader (Buchannan) become President, but if I vote for him, it's really a vote for Bush (Gore), so I'll vote for Gore (Bush) instead.") Is the "correct" answer for everyone to vote for their first choice? No, a complicated voting structure just allows manipulation by those who think about it long enough.

The best answer, I think, is a 10-9-8 format. Everyone will vote stategically (but they would anyway), and the "right" candidates win.

Anyone interested in learning how to manipulate voting results should be interested in reading the above link.

I hesitate to continue the debate with Matt since he has obviously made up his mind (presumably based upon a great deal of thinking about the issue). However, the issue is still important to discuss. I think Matt and I agree on the underlying situation, but not on the ramifications for the HOM.

We all agree that how the ballot is structured (etc.) can, and probably will, affect the HOM results. In fact, as Matt points out, certain schemes are more likely than others to lead to certain HOM results. Some voting schemes are more manipulatable than others. Some schemes allow a voter to reflect strength of preference while others do not. Unfortunately, as a general rule, schemes that can reflect strength of preference are the most manipulatable.

I conclude that we should have discussions up front as to how we "balance" these competing objectives. I for one do not want to be part of a multi-year exercise that is bedevilled by unintended consequences of its core rules.

Reserving the right to retract this view at a later time, I would be in favor of each voter agreeing not to vote strategically. I believe this is in the spirit of what Joe and Robert and the other HOM progenitors had in mind.

I don't have a lot of time to comment right now, but great discussion guys!

"Reserving the right to retract this view at a later time, I would be in favor of each voter agreeing not to vote strategically."

I cannot say strongly enough how much I agree with this. I think this should be part of the 'oath' or whatever you want to call it. We should all agree not to do this.

Let's use Barnes as the example.

I don't think anyone could possibly think Ross Barnes was the best player of the 19th century, even the people that want him in. I mean the guy played 2 1/2 years after the age of 26. Calling him the best player of the 30 year period is pretty ludicrous on any level. He had a great 6-year run, but that's it. If you count career value at all in your evaluation, you'd have to rank a few players ahead of him at least.

As such, his supporters should not strategically place him first, just to override people that don't think he's worthy at all. That would be dishonest in my opinion. What we are asking with the ballot, very specifically I might add, is to rank the X (currently 10, subject to change though) best players in their opinion, in order. Use career, peak, whatever you think, but be honest in your rankings.

On the flip side, no one should completely discount Barnes' 6-year run, just because most of it was against weaker competition. Reasonable deduction for weak competition sure, but not complete disregard for his achievements. We're asking people to consider that period, so they should.

I'll get the schedule for inductees as it stands right now, along with the rationale up in the next few days (maybe it'll be another article. I urge everyone, especially the new people with questions, to go back through the prior discussions on the archive, much of the questions asked here have been discussed previously. I have no problem answering, but if you went back through, you might get an answer quicker, my time has been tight of late and I haven't been able to check up here as often as I should.

Rob wrote:

"I hesitate to continue the debate with Matt since he has obviously made up his mind."

Aw, gee. Where's the fun of that?

Honestly, when this discussion started a few months ago, the consensus seemed to be for a "bonus system" where we could allocate more points to top selections in some years, or spread them out over several candidates in others when no candidate stood out. I objected to that scenario because it seemed to manipulatable.

Now, the consensus appears to be for a "flatter" vote. That method seems less subject to manipulation, but also less flexible in terms of expressing preferences. I agree with the other poster who felt that voting just wouldn't be as fun if it doesn't matter who you vote second or third, because it wouldn't matter anyway -- they'd both get six points. A flatter voting scheme would also result in more "bunching" of scores, with little differentiation between the top groups of vote getters (and would be more likely to result in ties).

Actually, as I write, I am starting to wonder if there might be a correlation between "bonus points" supporters and those who prefer "peak value", and between "flat voters" and those who prefer "career value." I don't know why I think this, but the idea sound somewhat appealing.

Really, "manipulation" is just a dirty word for "strength of feeling". If a life-long Democrat registers Republican in Louisiana just to vote against David Duke in the primaries, is he being "manipulative" (bad) or "stategic" (neutral), or is he merely expressing his opinion that who does NOT get elected is more important (in this case) than who DOES get elected.

I personally, have no intention to do anything other than list ten names in my personal order of preference and have whatever point values assigned to them that the consensus determines is accurate. I doubt there will be large conspiracies for or against any particular candidate.

My point on the ubiquity of manipulation possibilities was merely intended to promote a simple 10-9-8 balloting structure that, I think, combines the twin goals of "strength of feeling permission" and "large point manipulation avoidance."

Scruff wrote:

"I don't think anyone could possibly think Ross Barnes was the best player of the 19th century, even the people that want him in. I mean the guy played 2 1/2 years after the age of 26. Calling him the best player of the 30 year period is pretty ludicrous on any level. He had a great 6-year run, but that's it. If you count career value at all in your evaluation, you'd have to rank a few players ahead of him at least."

I also would not have Ross Barnes at the top of my first ballot. Without looking too far down the road, though, I could see that those who consider him a top tier second baseman might have him third on their lists (of second basemen), and move him to the top after the first two are enshrined. If Bill James were voting, he would never be on the list, as he does not put Barnes in his list of top 125 second basemen. After several iterations, it does seem likely that Barnes might be one groups' top 2B choice (and ballot topper in some cases, based on peak value), while others would omit him altogether.

The same can happen for Negro Leaguers, due to the problems in comparing them to White Leaguers.

The more I think about it, the more I think "manipulation" is a red herring. "Come up with the system that decreases the chance of manipulation" is sort of putting the rabbit into the hat. You come up with a different answer than "Come up with a system that allows voters to best express their true preferences."

Anyway, I spent a few minutes trying to come up with an example, so here is my mock election. I got three voters, I will call them Scruff, John, and Matt (not their real names). Assume that the top 3 vote-getters get in.

"Scruff" votes solely based on total adjusted Win Shares, as calculated through the charts listed in each category in these weblogs. This is his ballot, which looks like a potentially reasonable actual ballot:

Cap Anson

Thanks Matt for showing the example; I think it does a good job of illustrating some issues. Okay, I'm willing to let go of the strategic voting issue. For the time being, let's suppose that every HOM voter takes an oath promising to not vote strategically (details to follow).

Now, even under that scenario, it is not clear what is the "correct" voting scheme. It depends on what we are trying to capture here, and on our underlying assumptions about the inputs (voter "preferences" over HOM worthies).

There are several questions pertinent here. Do we want each voter to express his/her opinion on the HOM worthiness of each candidate? Or do we want each voter to express his/her opinion on the relative value of all candidates on the ballot? These are related but different questions. Also, do we fix the number of selections per ballot (which I kinda like) or let it free float based upon minimum vote requirements?

An idea that intrigues me is to give each voter X number of points. The voter can allocate those points however s/he wants among all the players on the ballot (pile them up on only a few players or spread them out over several players). Either the top N vote-point getters are selected or all those who receive more than Y vote-points are selected, or some combination of the two. I guess this is similar to the bonus point scheme Matt mentioned above.

The reason this may be attractive is that MVP-type ballots (e.g., 10-9-8-7-6-5-4-3-2-1) seem more suitable to expressing your opinion about who was the single best player in the league (MVP, Cy Young). But it is not immediately clear that this is the best way to express your opinion as to who belongs in the HOM.

Returning to my original point, even abstracting from the strategic voting issue, a 10-9-8-7-6-5-4-3-2-1 ballot seems to me to be far too steep than appropriate. Giving a player 1 vote will very likely be operationally equivalent to not voting for him at all. And it will do a lousy job of reflecting my preferences for who is a Homer.

There is no way that the 10th guy on my ballot will be 10 times less "worthy" than the 1st guy on my ballot. Let's look at the opinion's of Matt's three hypothetical voters above. I would hazard a guess that each voter has more like a 2-1 ratio of worthiness from his top to his 10th best player on his respective list (Anson vs. Glasscock, Brouthers vs. Richardson, or O'Rourke vs. Stovey). 3-1 at the outside. That's my intuition anyway.

One last thought. I am wondering if anyone thinks the following experiment would be illuminating. Draw 1,000 random numbers from a normal distribution. Look at the distribution of the averages of the top 10 draws over many trials. My half-baked idea is that the 1,000 represents how many major leaguers are (essentially) on each ballot. 10 represents how many are really Homer candidates. The spread of these 10 "values" (such as career WS) may reflect a typical voter's worthiness scale over his/her top 10 candidates. If I can figure out how to run the experiment, I'll try to do it tomorrow.

P.S. Maybe the experiment should be to draw "true" values for the 1,000 players, and then draw 100 sets of noisy values around these values to reflect the 100 voters' perceptions. Then, over many trials, look at different voting schemes to see which type of scheme does the best at reflecting the underlying truth.

First, I would certainly support the concept of a "loyalty oath" though I am not comfortable that "we" would ex post facto decide that a voter was being "strategic" because of who they did or did not vote for. This sounds a bit too much like the Supreme Court deciding afterwards which votes to count and not count.

Second, Matt's analysis was extremely helpful in clarifying the issues. I find the 7-6-6-etc. system very uncomfortable mostly because I LIKE splitting hairs down the list, but also, as Matt points out, because the vote tally bunches up so much. In a perfect world, there might be a clear split between winners and non-winners.

I also prefer the 10 point tops versus 14. I would expect that anybody concerned about voter manipulation would prefer 10. Also no one player initially will be 1.4X better than the next. 10-9-8-etc. seems to me to do the best job of forcing all of us to split hairs, and a hairsbreadth more often than not represents the real differential among the players.

My biggest concern, frankly, is avoiding the very problem that has given rise to the HoM in the first place--weak candidates sliding through. I would much rather have a minimum vote requirement than 5 automatic winners. Matt's analysis shows that the 5th player in each of the three scenarios had 53.3 percent (10-9-8-etc.), 47.6 percent (14-9-8-etc.) and 61.9 percent of the maximum. In the case of the 10-9-8-etc. system I'd like to see the winners get something MORE than 53.3 percent of the max. I don't know if that's 55 percent or 60 percent--67 percent might be too high. But a minimum. With a failsafe, which is, if nobody gets elected, then have a run-off of some number (maybe the top five) with only one winner regardless of percent.

This could of course result in vastly underpopulating the HoM. But how can we avoid allowing weak choices to slip through--especially given our natural enthusiasm for electing people?

Rob,

I see what you are saying about the 2-1 ratio. In fact, the difference from Anson to Glasscock in WS (567 to 360) is less than 2-1, and the ratio of Brouthers to Richardson in WS/162 is even less.

My biggest problem with 7-6-6-5-5-5-4-4-4-4 is that it does not distinguish between 6 and 10 at all.

Perhaps a compromise position is a 20-19-18. . . 13-12-11 ballot.

It preserves the distinctions between all ten positions, but also adds a 10 point "on the ballot" bonus. If two players have the same vote total on a 10-9-8 scale, the one that appears on more ballots would win on a 20-19-18 scale. These are the results based on my example above:

To comment on a couple of points raised by Rob. He wrote:

"We don't really expect every voter to accompany each vote with a long message detailing his/her reasons, do we?"

Well, yeah, we kinda do. Not so much a LONG message. But at least a sentence or two indicating the direction of the reasoning.

Rob also wrote:

"I for one do not want to be part of a multi-year exercise that is bedevilled by unintended consequences of its core rules."

Yes, a central issue. Which is why we began thrashing out this structure over six months ago. I want to reiterate the idea that new readers here should go back and read the old threads.

One direction we could take is to have voters assign a point value to every player on the ballot. Earlier, a consensus indicated we would use a 100 player ballot for each election. We could adopt a structure similar to the following. On each ballot, voters will assign:

10 points to 3 players

On Bid McPhee and Al Spalding: (or fielding vs pitching, again)

I did some more rough work with the National Association numbers and Al Spalding. Estimated AWS in the established format look like:

805 - 172,140,125 - 684 Al Spalding 5.81 seasons (detailed breakdown not available) His rate is, what, 138.55?

Babe Ruth has an adjusted total around 810 (756 BJWS). As incredible as Spalding's adjusted numbers may seem, they are not out of line for Randy Johnson if he pitches every day and can hit. 6 years of that is about as valuable as Babe Ruth's career.

If we believe that pitching numbers from the 19th century have the same interpretation as they do today, we are forced to the conclusion that the position players are not that important, at least not until they can pile up long careers. Using BJWS (or AWS), the MVP every year before 1899 is a pitcher; no position player ever gets within 10%. The position players would need a "Cy Young" style award to get any recognition. Peak value discussions will be about pitchers. It is a pitcher's game, except they have a tendency to burn out quickly (get hurt?) and have short careers.

The alternative to this picture is that fielding is much more important than it is today. Which was the point of my list above, where Bid McPhee was number 5 in career DFWS (Double Fielding Win Shares). These are not calculated with any precision, they are simply based on two principles: 1) run prevention is as important as run scoring, and 2) if the pitchers aren't preventing the runs then the fielders are. So let's double the fielding win shares and see what happens (This changes the balance of pitching:fielding balance from 2:1 to 1:2).

McPhee and Glasscock are the players with the long careers and high defensive WS totals who are most likely to benefit when reallocating their team's wins due to any shift of responsibility from pitching to fielding. It's a see-saw; if pitching goes down, then fielding goes up.

By the way: using DFWS, Spalding has a total of about 485, comparable to that block of Brouthers, McPhee, Kelly, Hines, and Glasscock.

BTW, that original voting discussion is not currently accessible in the archives. There is a problem with the link.

I missed the first round of voting procedure discussions, please forgive me if this reflects thoughts presented before.

First, how will the procedure ultimately be determined? Do we have a commissioner or blue ribbon panel who will decide, or is this something we will vote on?

How many registered voters are there presently, and will others be able to join in after we get started?

I propose a system using a 10-9-8-7-6-5-4-3-2-1 ballot where only players who appear on 51% of ballots submitted are considered for enshrinement. This effectively gives every voter a "no" vote on any player not on their list and should avoid undesirable ballot manipulation results. If 49% of the voters rank Ross Barnes (not to pick on Barnes, we could call him Chick Hafey) #1 while the majority leave him off their ballot, Barnes does not get in regardless of his placement in the point tally. I would not require voters to list more players than they feel worthy of election - if it is their opinion that too many 19th century players have been already been selected, I would accept their blank ballot as a "no" vote on everybody.

I recognize that when there is a large pool of worthy candidates the possibility would exist that not enough players would be named on 51% of the ballots. In that case, a run off election could be held.

Great stuff Eric, thanks for the info!

I have to write my Bootleg column tonight, so I don't have a ton of time, but let me address the one thing that's got everyone questioning the system.

As this was designed the top X number of people in the voting (we'll leave how the voting is constructed alone for now) would be enshrined. X will depend on the year of election.

We set up a spreadsheet. The number of people elected in a given year is based on the number of 'team seasons' leading up to the election, of course we'll take time to catch up, not just elect our backlog of 19 HoMers for the 1906 election.

I'll outline the whole formula here, some of the final numbers are subjective.

Team seasons used in the calc:

1871-75: 0

One note: the reason I think the minimum standard won't work is because it's arbitrary, just like the 75% rule for HoF elections. There's absolutely no way to know what number should be right. We'd just be pulling a number out of thin air, like the 75% number.

DanG: the system I devised fixes that.

Essentially, what I am doing is figuring three values:

* Runs that result from events charged only to the pitcher, like HR, BB and SO.

* Runs that result from events charged only to the fielders, like errors, outfield and catcher assists and DPs.

* Runs that result from everything else, namely hits.

The split is the pitcher-only runs and half the shared (hits) runs versus fielder runs and the other half of the shared runs. It scales nicely; here's the split for some seasons:

1874 NA 71.3% fielding

"The pull rates are worthless, since batters do not pull flyballs."

This might be true today, but when pitchers weren't throwing very hard and you could call your pitch high or low, I bet pull rates on everything were through the roof. I'd be very surprised if that wasn't the explanation.

Thanks for posting the pitching numbers Charlie -- there's a very nice progression there.

scruff proposed the following schedule of elections per year:

1906: 5

>> Thanks for posting the pitching numbers Charlie -- there's a very nice progression there.

Charles Saeger, it's also a tease. For this project, I'm on the edge of my seat, dying to know whether the changes are evolutionary or revolutionary (big changes when high impact pitching rules changes happen). Inquiring minds want to know more. Encore, encore.

>> A yes/no referrendum is exactly how the HoF problems arose.

scruff, I strongly disagree with this statement. "approval voting" is definitely appropriate for this kind of election, like the HOF. They required a very high standard of approval (75%) to insure that the writers were in strong agreement about a candidate's worthiness. The HOF's problem arose from putting an artifical limitation on the number of "Yes" votes. Combine that with a ballot overflowing with worthy candidates and deadlock developed. Add an old-timer's committee commissioned to induct some people pronto with inadequate research and then you had a big problem (unqualified people selected).

Please do not interpret this as me lobbying for a HOF style ballot here. I am quite curious about how the MVP ballot works out for this, and I think it should be the 14-9-8... scale for that reason. I'm agreeable with any variant on that that people come up with. My favorite was 4 bonus points for each candidate elected; i.e. 14-9-8... when electing one candidate, 14-13-8... when electing two candidates, 14-13-12-11-10-5-4-3-2-1 when electing five candidates.

Just wanted to say that I am currently running some experiments (simulations) that may shed additional light on the ballot structure issue. I'll post the results later this afternoon.

Okay, as promised, here are some results I hope will prove interesting. I performed a bunch of simulations into the impact of the voting structure. Let me say that in all of the simulations there were 1,000 players (think of this as the "universe" of potential MLB players), 100 voters, each voter voted for 10 players (except for one of the schemes, see below), and 5 players were elected.

I did 1,000 trials and averaged the results over the trials. In each trial, I drew 1,000 true values for all the players in the universe. The draws were from a normal distribution with mean 250 and standard deviation 75. Then for each of these 1,000 players I drew 100 perception values, one for each voter, where the draws were from a normal distribution with mean equal to the player's true value and standard deviation of 35.

Then voters voted according to their perception values. They voted for their top 10 players (except for one scheme). Votes were given different points according to what scheme was followed, and then the top 5 players who received the most total points were elected. I then compared how well the elections matched the truth (the top 5 players with the highest true values). I tried 5 different voting schemes and measured the results with 3 different metrics.

The five vote tally schemes. V1=10-9-8-7-6-5-4-3-2-1. V2=1-1-1-1-1-1-1-1-1-1. V3=14-9-8-7-6-5-4-3-2-1. V4=20-19-18-17-16-15-14-13-12-11. V5=5-0-0-0-0-0-0-0-0-0. I could easily test other voting schemes if people want.

The three metrics from toughest to easiest. M1 simply counts how many of the top 5 best players (from the true values) were elected. M2 rewards different numbers of points to top-5 matches taking into account the rank both of the vote tally and the actual true values, giving more importance to electing the top players. Top vote getter gets 10 points if he was top true value, 8 points if 2nd true value, 7 if 3rd, 6 if 4th, and 5 if 5th. 2nd vote getter gets 7 points if 1st true value, 9 if 2nd, 7 if 3rd, 6 if 4th, and 5 if 5th. 3rd vote getter gets 5 if 1st, 6 if 2nd, 8 if 3rd, 6 if 4th, and 5 if 5th. 4th vote getter gets 5 if 1st or 2nd true value, 6 if 3rd, 7 if 4th, and 6 if 5th. 5th vote getter gets 5 if 1st, 2nd, 3rd or 4th true value, and 6 if 5th. (I apologize for all the details, but I think it is better if I present everything.)

The third metric M3 ignores where the player ranks in the vote tallies, but does care about the rank according to true value. If the top true value player is elected, that is worth 10 points. If the 2nd true value player is elected, that is worth 9 points. And so on all the way down to 1 point for the 10th highest true value player.

In the results below, I present the metrics as percentages of the total possible points. Hopefully this will make sense to people, but if not, feel free to ask for clarification (though I will not have access to the site for a couple of days starting tomorrow).

____ M1 M2 M3

Great stuff Rob, I'm glad someone's actually trying to objectively judge the voting system. It was one of my biggest hopes when I took this public (it originally started with just me doing everything for my own benefit, then I started MostlyBaseball, and so and so on).

Please give the 7-6-6-5-5-5-4-4-4-4 or 7-6-6-5-5-5-4-4-4-4-3-3-3-3-3 systems a try as well, if it's easy to do. Thanks! Also, if you are going w/5 electees, could you go with 14-13-12-11-10-5-4-3-2-1 voting as well? Thanks again!

I must say that I'm still up in the air, but I agree that the 10-down to-1 voting is in the early lead, although I'm hoping for the 14-13-etc. 5- down to -1 system to be the best.

You know, I'm on board with whatever voting procedure you guys come up with. And, sure, agonizing over the structure is important, not to mention part of the fun.

But one thing to remember is that the BBWAA voting is not the problem with the "real" HOF. The Veteran's Committee is the problem and there is no comparable structure or procedure contemplated here (thank goodness). That said, I don't understand the need to elect the same number of HOMers and HOFers. Now I recognize that this is not the same as saying that we are allowing the real HOF to establish the standard of excellence because they haven't established one. We will. Thank goodness again. I do feel that determining what and where that standard should be established is what the process is all about, not something that is arbitrarily determined ahead of time. That's no more logical than a 75 percent (or whatever) consensus requirement.

Marc has an excellent point. How many mistakes have the Writers actually made? I don't think very many (wish I had my _Politics of Glory_ here to check).

Marc has an excellent point. How many mistakes have the Writers actually made? I don't think very many (wish I had my _Politics of Glory_ here to check).

The bottom 5 by the writers, if you look at it from the POV of career WS:

181 Dizzy Dean

The BBWAA has been way too tough, their errors are errors of omission, because of the flawed ballot scheme and an arbitrary 75% of yeah/nay votes with no criteria to go on.

Because of the flawed BBWAA structure, the Veteran's Committee became a necessary evil, and their errors of commission come into the play.

The whole thing is just a mess, mainly because there was no foresight into structuring the initial rules.

I agree that the "whole thing" is a mess, the "whole thing" meaning the HOF's response to the BBWAA's failure to elect in some early election cycles. But the BBWAA over time has elected worthy players (mostly) and rejected unworthy ones (mostly, though how they missed Wahoo Sam Crawford is being me). The HOF responded by establishing a Veteran's Committee that elected unworthy players (mostly, at least early on). All of that is beyond challenge. That is why there is a HOM and I salute those of you who have given birth to this process.

But not to confuse ourselves. When we say the BBWAA has been "way too tough," what do we mean? Take second base as a case in point. Who that the BBWAA rejected should they have accepted? How many of the Vet's Committee choices should have been selected by the BBWAA? (The HOF roster is included here for reference.)

Rogers Hornsby BBWAA

<i>

Marc wrote:

"That said, I don't understand the need to elect the same number of HOMers and HOFers."

I think one of our tasks here is to show people exactly what HoF choices are errors and exactly which players ought to be there in their place. The HoM has no chance of replacing the HoF in the public consciousness, but our choices can be a way of framing the discussion to focus on the best overlooked Hall candidates. Hopefully, in some way, what we do here will have influence to help correct the existing injustices in the Hall's process of honoring baseball's greats.

As a guess, I think we will differ with the HoF on 10-15% of their 211 players, (21 to 32 guys).

Regarding BBWAA errors of omission, I would at least add Goslin and Mize to Crawford and Vaughn.

Reagrding the worst BBWAA choices. I think the worst are 1)Pennock, 2) Maranville, 3) Hunter.

DG

Marc wrote:

"That said, I don't understand the need to elect the same number of HOMers and HOFers."

I think one of our tasks here is to show people exactly what HoF choices are errors and exactly which players ought to be there in their place. The HoM has no chance of replacing the HoF in the public consciousness, but our choices can be a way of framing the discussion to focus on the best overlooked Hall candidates. Hopefully, in some way, what we do here will have influence to help correct the existing injustices in the Hall's process of honoring baseball's greats.

As a guess, I think we will differ with the HoF on 10-15% of their 211 players, (21 to 32 guys).

Regarding BBWAA errors of omission, I would at least add Goslin and Mize to Crawford and Vaughn.

Reagrding the worst BBWAA choices. I think the worst are 1)Pennock, 2) Maranville, 3) Hunter.

DG

I can't find the comment now about rushing Brouthers, O'Rourke and Rusie to eligibility (referring the fact that each appeared in a few games after the turn of the century and long after their first retirement). I don't have strong feelings about this either way, but I did discover an interesting little twist. Maybe somebody has already commented. Anyway, Sam Thompson retired after the 1898 season, then made a brief "comeback" (8 games) in 1906. If we are simulating a 1906 HOF election, then we would have no way of knowing early in the year that Thompson would do that. If this is a simulated reality, he would be on the ballot; if we want to acknowledge that it is not (now) 1906, then maybe not.

In Brouthers' case, it might be fun to see how we all rate Anson-Brouthers-Connor, head to head. In O'Rourke's case I don't really care either way, ditto Thompson. In Rusie's case, it might again be fun to see how we stack him up against the other leading pitchers who retired in the '90s--Clarkson, Keefe, Radbourn, Caruthers, Mullane, Galvin, Welch--but on the other hand you could also argue that he was more of a comtemporary of Kid Nichols who didn't retire until 1906.

I guess in the balance I'd rather consider these guys based on their original retirement, against their peers, just as I would have preferred that Minnie Minoso go against other players of the '50s. It feels right.

Dan wrote:

>Regarding BBWAA errors of omission, I would at least add Goslin and Mize to Crawford and

I don't mean to speak for Scruff (or anyone else) who was in on the HOM ground floor. Having said that, let me chime in with why I think there needs to be some firm constraints on the voting. If we allow each voter to express his/her opinion on who belongs in the HOM without a pre-determined number of electees for each ballot, then the selection rule (e.g., minimum requirement such as 75%) will likely have a great deal of influence on who gets elected, and how many people get elected. So we wind up endlessly debating the selection rule (as distinct from the ballot structure). I don't think there is anyway around this issue and no way in the world that we can agree on the best selection rule.

As a way to "close" the system, Scruff has suggested that we fix the number of Homers equal to the number of HOFers. In this case, essentially we would be replacing the worst HOF selections with the best non-selections. And we would have era representation built-in from the start, something I am a strong proponent of. Though I personally think that there are too many HOFers, I have no problem with this approach.

To sum up, I think the system that Scruff has set up (i.e., pre-determined number of selections in a series of annual ballots starting in 1906 or so) is not perfect. But I cannot think of anything better, taking into account all the inherent problems associated with identifying the game's greatest players.

Splitting Win Shares between pitchers and fielders.

I'm probably making a few mistakes here, and this is my method, so there's no one to correct me, but here it goes.

1) Calculate the league strikeout value.

We need a strikeout value that takes into account hit prevention before we go anywhere. Therefore, we need the league hitting rates and XRuns to compute this value.

((0.50*1B)+(0.72*2B)+(1.04*3B)+(0.30*Err)-(0.098*(AB-H-SO-(0.60*Err))))/(AB-HR-SO)+0098

For the NL in 2001, this value is 0.174. This value is now chi, or X, for lack of a better shorthand.

2) Calculate pitcher-only runs.

There are some events we know to be pitcher-only, save for park and possible psychological effects of a decent defense. We need to find out the value of these events.

(1.44*HR.pkadj)+(0.34*(BB-IBB+HBP))+(0.25*IBB)-(X*SO)

HR.pkadj is park-adjusted Home Runs, which is just Home Runs allowed times the appropriate park adjustment.

For the 2001 Atlanta Braves, this value is 197.

3) Calculate fielder-only runs.

Likewise, there are events that are primarily the responsibility of the fielders.

(0.37*Err)-(0.32*DP.1b)-(0.50*A.of)-(0.32*A.c)-(LgR/PA*(AddlDP.1b+AddlA.of))

LgR/PA is the league Runs divided by league Plate Appearances. For the 2001 NL, this value is 0.124. AddlDP.1b is the team rate of Double Plays by first basemen over and above the league average rate. AddlA.of is the same for Outfield Assists.

Alternately, you could use OCS for A.c and add 0.18*OSB, where these are available. For the 2001 Atlanta Braves, this value is -17. For recent season, this value is typically negative, due to low error and high double play rates.

4) Calculate the shared runs.

This is Hits Allowed. We need to calculate the league value of a hit, where doubles and triples allowed are not known:

((0.50*1B)+(0.72*2B)+(1.04*3B))/(H-HR)

We'll call this LgH. For the 2001 NL, this is 0.564.

Then, we need a value for outs that reflects the hit-prevention value of a Strikeout:

(0.09*PO+0.008*SO-X*SO)/(PO-SO)

We'll call this LgPO. For the 2001 NL, this value is 0.48.

Now, the actual figure.

(H-HR)*Park-S*LgH-(PO-SO)*LgPO

Park-S is Bill James's Park-S adjustment. For the 2001 Atlanta Braves, this value is 524.

5) Split the shared runs.

This is where style and taste enters the system. However, the bland version, which I use, is just adding half the shared runs to the fielders, and half to the pitchers.

That's a thoroughly arbitrary split.

You could be named "Chris Dial" and blame hits allowed mostly on the pitchers, or named "Mike Emeigh" and blame hits allowed almost entirely on the fielders. If you are, you could move this 50/50 split to something like a 60/40 split either way, or a 75/25, or 81/19, or whatever. I use 50/50, which gives results in line with traditional values.

IAE, the 2001 Atlanta Braves now have 459 runs charged to the pitchers and 245 runs charged to the fielders.

6) Sum the league totals, and divide by league innings pitched.

Nothing special, but we need league rates. We'll call this LgP or LgF, as needed.

7) Figure the replacement rate for each value.

The replacement rate is 1.52*LgP*TmIP. Substitute LgF for the fielders. You might understand where I am going with this ...

8) Subtract the team value from the replacement rate to calculate Claim Points.

Very Bill Jamesey. For the 2001 Atlanta Braves ...

The 2001 NL is charged with 8699 runs to pitchers and 4128 runs to fielders in 23074.1 innings. The Braves played 1447.1 innings. Working the math, the Braves pitchers earned 370 Claim Points and the fielders earned 148 Claim Points.

9) Apportion Defensive Win Shares between the pitchers and fielders based on Claim Points.

I figure the 2001 Braves had 161 Win Shares on the defensive side, easily the best in the league (the Mariners had more in the AL, who had 165). Based on the above proportions, I figure the Braves pitchers earned 115 Win Shares (2nd in the majors behind the Yankees, who had 120) and the fielders earned 46 Win Shares (6th in the NL). The Braves had 28.6% of its defensive Win Shares going to fielders, the NL had 32.2%. Note even the total defensive Win Shares were above average, as a typical NL team had 40. Even as a percentage of total Win Shares, the Braves fielders came out ahead (a normal 88 win team in the 2001 NL would have 44 fielding Win Shares).

Per Scruff's request, I have run another 1,000 trials of my vote structure tally simulator. See my July 24 (7:39 pm) post above for the details. Scruff asked if I could try out three more possible voting schemes. To ensure comparability, a ran another batch with the original five plus the new five: V1=10-9-8-7-6-5-4-3-2-1. V2=1-1-1-1-1-1-1-1-1-1. V3=14-9-8-7-6-5-4-3-2-1. V4=20-19-18-17-16-15-14-13-12-11. V5=5-0-0-0-0-0-0-0-0-0. V6=7-6-6-5-5-5-4-4-4-4. V7=7-6-6-5-5-5-4-4-4-4-3-3-3-3-3. V8=14-13-12-11-10-5-4-3-2-1.

I calculate three different metrics (M1, M2, M3) to evaluate how well each voting scheme performs (see above post for details). Each metric is expressed as a percent of possible total points, so the ideal is 100% on each metric.

V1: 69.5 89.4 98.4

Andrew S. asked:

"Is it my imagination or will our voting schedule result in the election of more 19th-century players than are already in the Hall (by defintion, at the expense of later players). If I'm right about this, isn't this problematic? Even if I'm wrong, why are we starting these elections so early chronologically?"

Our number of 19th century players will be about the same as the HoF has. We feel this is appropriate representation for them, despite the fact that we know that "modern players are superior to their predecessors".

Players like O'Rourke competed for real championships at the highest level of play available. We feel these early greats are worthy of honor, being the best of their era.

If we started with a later date, there would be voters who entirely discounted the achievements of the 19th century greats. We think that is misguided thnking. They would have smaller representation in our hall than we feel they merit. So I think the HoM is set up deliberately to ensure the representation we feel is appropriate for them.

DG

I think the biggest advantage of starting with 1906 is that 19th century players will be judged against their peers. It had to have been impossible to evaluate the 19th century players against Ruth, Cobb, Johnson, et al by 1936, which is why the HOF ended up punting.

If anything I'd argue that we could start earlier since it is a little bit of a stretch to argue that players like Brouthers, Ewing and Thompson are "peers" of Deacon White, Barnes, George Wright, Al Spalding and the others from the NA and pre-NA. Just comparing pitchers pre- and post-1893 is tough enough.

I personally want to consider NA performances as best I can, as well as pre-NA performances, and that will be difficult enough. But I really like the idea of considering such players as players, rather than copping out and calling them "pioneers." 1906 isn't too bad, though, in that as a practical matter we are talking about players who started their careers, at least, in the 1880s. Most who started in the '90s will not yet be eligible. But comparing the '70s vs. the '80s is tough enough, so maybe if we started right in 1900 or 1901 we could help ourselves just a tiny little bit. I don't think you'd commit to 5 electees at that point, however.