What follows is an abridged, time-phased, date stamped listing of most of the e-mails that were passed around during the discussion of a suggested change in our scoring method. The players are:
At the end are some comments that were submitted after the discussion was closed. I count the very last entree as the single most significant of all. Thank you, Pete.
July 21st/4:44pm (EDT)
I haven't spent a whole lot of time looking into the system you use to rank the contestants in the game, but I've had the feeling for a while that there's something intrinsically wrong with the scoring. I haven't looked at everyone's picks, but looking at Dominick's original picks, he picked TB last in the ALE and CLE 1st in the ALC. I keep asking myself how someone who made such misjudgments be rewarded with a 1st place ranking at the moment.
Again, without having gone through this too deeply, my reaction is that some people are getting rewarded or penalized (unfairly) depending on how close or not close each respective division race is. I don't think it should make any difference at all how many GB a particular team is in each division as long as you pick the correct order they come in, or close to it. For example, again using Dominick, in the NLC he has MIL, StL, CHC, CIN, HOUS, & PITT. He doesn't even have the correct division leader, yet he has only 6 Pts for the NLC, while I picked CHC, CIN, StL, MIL, HOUS, PITT and I have 9 Pts for that division when other than CIN being 2 slots too high I have the rest of the order right.
I wonder if you'd get a much different result if you took out the GB distortion by simply assigning a point value to every team you picked in the right order. For every division with 5 teams, assign 4 points to every team you pick in the right spot, 3 points if you're 1 spot off, 2 points if 2 spots off, 1 point if 3 spots off and 0 points if 4 spots off. For the NLW, use 3,2,1,0. For the NLC, use 5,4,3,2,1,0. The highest number of points wins. I think you could have a max of 122 points...if you picked every team in the right spot, you'd have 20 teams in the 4 divisions with 5 teams where you could get 4 points max for each (for 80 pts), the 6 team division could get you 5 points for each (30 pts) and the 4 team division 3 points for each (12 pts).
I guess you would need some sort of tiebreaker system. I could see the GB feature being used as a TB in some fashion.
What do you think?
In 28 years of running this silly thing I have had a lot of questions about how the game is scored but this the first time anyone has ever challenged the scoring.... I am working on a response which will include some rationale and some history as well a brief scoring critique shared with me by Tom Tippett of Diamond Mind Baseball a few years back.
If you or Jim or John have anything that they would like to add to the mix, please let me know, or perhaps you would like to write directly to Bob.
Just thought that I'd let you know that all is not well in Mudville.
A New Challenge....
I think the scoring system is good. I know of nothing better, though I know that you have looked into it a lot more than I. I remember only vaguely the Tom Tippett discussion about scoring. One possible fault in the scoring system is that we all get bailed out from time to time when we are way off on a team but it doesn't hurt us due to another team's performance. I'm not sure how that could be fixed, or even should be fixed to maintain the tradition of the game.
Some points that come to
- Of course the system he suggests doesn't distinguish between being 1 game off or 20 games off. Those who picked Washington last (which they are by 8.5 games) should benefit more than those who picked Baltimore last (which they are by 0.5 games).
- We've always played the game like this. To maintain tradition, it should remain. To make a change that dramatic might be the equivalent to awarding the football team with the most yards gained instead of points scored with the win. Minor tweaks, fine, but Robert's change would be major.
- Your Apple spreadsheet would revolt!
- Teams are often separated in the final standings by only 1 game, something that can be determined by the bounce of one ball. I think a system needs to distinguish between that and being 10 games off.
- In the NLC he must have more than 9 points (or you are doing him a favor)! In the NLC, there are 3 teams in a close battle for first and 3 teams in a close battle for last w/ an 8 game spread in between. Dominick correctly picked the 3 teams that are playing well (.561 or above) in the top 3 and the 3 teams that aren't playing that great in the bottom 3 and is being rewarded justly in my opinion. Robert and I each picked Cinci in the top 3 and are justly penalized for it in my opinion.
- In the NLC, under Robert's system, transposing the Milw and St L teams (one game diff) is just as important as transposing Milw and Cinci (8 game diff). That would be unfair. Milw is clearly outperforming Cinci this yr. But Milw might tie St L w/ the bounce of a ball tonight.
Anyway, those are some of my thoughts. Feel free to pass some of my "points" along.....
July 21/19:24 Thank you very much John. You just made my day worth while. I think that I have considered all of the points that you have made plus a little about Randy's and Dennis' games and Pete Palmer and Steve Mann's game as well as Tom Tippett and ESPN. I'll try to finish in the morning and I'll share it with each of you.....
Thanks too to Nelson and Jim....
July 21/17:41 Great comments, John. I agree on all points. I think we all agree on our scoring system vs Schaffer's system. John, has illustrated the point well. You can certainly have different scoring methods. But what ever the rules, you predict based on the rules. When I started football predicts we just picked the winner and I had my odds on who was favored. Thus it became a coin flip unless my Odds were way off (put the form out before odds were available). Became a coin toss game thus switched to the current system. As I recall the predictors cum % rating for that 1st year was under 50%!!
Nice that someone out there
is paying enough attention to recommend a different scoring system
- congrats Gerry.
July 21/18:14 Good to see that there is enough interest in the game to raise questions about the scoring system. My opinion is that there is no perfect scoring system, and reasonable people may disagree about what is best. However, the operator of the game gets to decide.
My opinion is that the current system is best.... How many games a team is away from the predicted finish is a fair measure. If a team finishes in first by 10 games and I correctly predict the first place finish that is a more significant prediction than if I correctly pick a team first and they win by 1 game. In the second scenario, there was no significant difference between the 1st and 2nd place team over the course of the season. Thus, there ought to be a minor difference between a predictor who picked the 1st place team to win and one who picked the 2nd place team to win, when the teams were only a game apart over 162 games.
Just my thoughts.
July 22/10:13 (pdt)
Thanks for responding Bob. I really appreciate any critique or suggestions that you may have to offer.
Over the years we have had a lot questions regarding the scoring of the game, usually because some folks can't figure it out on their own. I usually point them to the Top3 page where the scoring is illustrated.
Let me start this message off with a little history because in the end that is what this is all about. The game was started by a sixteen year old high school junior in 1955. While the game has changed some over the years, the scoring has remained essentially the same since the beginning- points based on games out of their predicted position- low score wins. Dumb response: That's the way it's always been. A more complete response: We've got a lot of years invested in the existing method. It is well understood and it works.
Your comments about the penalty applied for teams that are many games out of their predicted order of finish is significant and I also think important. One of the first things that I do every Spring is to try define where all of the big gaps will be in each division. For years I could depend on New York and Boston in the AL East and it always seemed to serve me well, but not this year. This time I have missed on at least two teams in five of six divisions and I am paying the price. I was certain of Detroit and Cleveland in the AL Central, but then nothing is certain, is it?. I too have also missed on Washington, as have so many. I think that it is a significant miss and those 50 or 60 that got it right get rewarded- most are on page one and they'll probably stay there. Also the current gaps between Milwaukee and Cincinnati as well as Texas and Seattle is significant and should be scored that way. I simply can not accept your argument on this one Bob.
I know of two "spin off" games from our game and in each case the new game was scored the same. The exception would be the second of those two games which introduced a new tie breaker which I adopted into our game within the next year or two.
I also know of two other games that were developed independently and in each case they had their own scoring system. The first was begun, ironically in 1955, by couple of guys that were just beginning their professional careers in sabrmetrics. You might recognize their names: they are Pete Palmer and Steve Mann. They were interested in how their preseason predictions would compare with the many annual baseball predictions that were so popular at the time. Their scoring is the reverse of what you are suggesting with one additional and significant difference. They square the differences; that is to say that if you picked a team three places out of their actual position, Tampa Bay for example, the score is nine rather than three and some cases the penalty in Pete's game is greater than it is in our game.
Eight years ago Pete sent me the entire history of his game and I was able to find within that data several annuals that were also included in our game over the same period. From that I was able to do a comp on the overall ranking of that common data. The rankings were remarkably close. Pete's game continues today- 54 years and counting- but it is also closed.
The other game was not really a game at all, but a simple computation that ESPN did with the preseason predictions made by their own staff and, as far as I know, it lasted only for as long as the ESPN staff made those predictions.
One person that had access to all three games was Tom Tippett who was, at the time, the President and creator of Diamond Mind Baseball, a baseball simulation game for home computers. He would use the DMB seasonal player disc to run 50 annual simulations on the coming season and submitted his preseason results along with a narrative to ESPN where it was posted on their site. At the end of the year Tom would collect key predictions- he used our site to collect what he had missed earlier- for certain annuals and writers and compare them to the DMB results. Tom was also familiar with each of the different scoring systems and he told me that he liked ours the best. Perhaps he was just being nice.
Those other games and Tippett's post-season analysis all have two things in common. They appear to be relatively simple and they are, in my opinion, difficult to mechanize.
On the other hand, when I bought my first home computer- a little Apple //c- I had a functional spreadsheet ready accumulate player predictions and compute player scores and rankings in less than two months and it was my very first experience with a computer. The current spreadsheet now at work in my iMac, while much bigger and faster, still retains that original design and it still works.
Granted, I have not given your suggestion a lot of thought but at first glance, I cannot imagine how to mechanize it. I suspect that it would take a very complex algorithm that I am not capable of writing. I also believe that your game shares a couple more common elements with those games described above. They are designed to be small and computed just once at the end of each season.
By now you already know what I think, but let me first close the loop that I opened a few moments ago. First a couple of questions: Where would I begin? Certainly not at the beginning. Even if all of the original data existed, and it doesn't, I don't have enough years left to rewrite 49 years of history. If we choose to begin next season, how do I tie it all together? What you have suggested is an entirely different game and would not match up in any way- well, yes it is baseball- with what we have been doing for 49 years. Finally, how do I explain this to those many faithful players that have become accustom to the way that the game is been played for so long? I can't.
I know that you put a lot of thought into what you have suggested and I appreciate your sharing it with us but It just won't work. It is much more than simple tradition. Bob, it just will not work.
July 22/13:15 Great response Gerry. You should include this exchange of e-mails with Schaffer on your web site history. Right along with the Hollingworth articles or where ever you see fit........
Thanks for taking the time to respond in such great detail. I guess my principal knock on your scoring system is that we were all asked to submit our predictions for the order of finish for each division in baseball regardless of record (or obviously, games separating the teams). However, the scoring system accommodates for exactly that - the games that separate the teams. Do you see the illogical progression here? If we're going to be judged by a certain criteria, then we should submit our predictions of those same criteria...no?
Can you explain to me the significance of Dominick's 1st place in the standings and his prediction of CLE winning the ALC, with CLE now occupying last place and his prediction of TB in last place in the ALE now occupying 1st place?
I'm just having a hard time understanding exactly what the scoring system represents. I appreciate the historical significance of the game and the time and effort that has gone into it over the years, but maybe you could explain to me in simple words what the folks on the 1st page are getting rewarded for as opposed to the last page, for instance. Correct me if I'm wrong, but a player could have several divisions nailed exactly, but one team like the Padres completely out of whack and for that he could conceivably be farther out of the lead than someone who is wrong on 5 divisions but only with tighter margins in each division? Is that generally speaking correct?
Thanks again and don't mistake my questions for lack of interest in the contest...
I will take another look at your data and I am considering creating a special page- a copy of the Top3 page- that will let you compare your prediction to a couple others. Hopefully it will help you understand. in the meantime, here is a copy of the game objective that went out with the game invitation that you responded to on April 4th. It's all there, Bob. More later (tomorrow)
ps: I never responded to
directly to Bob's July 22nd e-mail and I regret that, but Nelson's
response (below) covers it very well.
WELCOME TO THE 49th ANNUAL BASEBALL PREDICTS
OBJECTIVE: Predict the correct order of finish in each of the six divisions for the 2008 Major League Baseball season. Scores are determined by taking the number of games each team finishes out of its predicted order of finish. For example, if you picked New York to win the AL East in 2007, you would receive 2.0 points because the Yankees finished in second place, two games behind the Red Sox. The scores for all 30 teams are added together to determine the total score. The tiebreaker is the combined score for the six projected division winners. The league champs and world series winner will be used as information only.
LOW SCORE WINS!!
July 22/20:42 in response to Slag's reply- Sure, no problem. In the meantime, sometime tomorrow I will make a couple of copies of the Top3 page and plug Bob's picks in them along with two or three others, and then score them our way and his way and see what happens.
I also have another consideration to define. Using his method the Standings 1) wouldn't change much from day-to-day and 2) the player scoring would be concentrated between about 100 and 60 points and the tie breaker would become increasingly important.
July 22/21:29 fm Nelson
First, let be thank you for your interest in the game. In the 50 years, Gerry, my brother and I have run the game you are the first to seriously question our scoring system. But that is good. We always felt that player responses are to infrequent, thus your comments are most welcomed. Lets me try to address your comments issue by issue.
I believe you are saying if we are scoring by how many games you miss a teams position you should predict how many games the team will be out of first rather than just their position. That would certainly be a viable way to predict and more precise. However, entering and computation of the stats would be much more time consuming. In addition, we would get many fewer predictions as few if any of the sports writers predict in that manner. Also, many people would drop out as it takes more time than they are willing to spend on the prediction.
The 2nd comment was the logic of a person picking Cleveland 1st leading the predicts. That one is easy. As the standings are currently, it makes little difference who you picked #1 as 190 out of 194 had Cleveland or Detroit 1st and only 15 predictors had another team predicted 2nd besides Cleveland and Detroit. Regardless of the scoring system almost everyone would 'flunk' their predicts in that division. However, note in the results to date that 4 of those 15 picking Chicago or Minnesota 1st or 2nd are in 2nd thru 5th. Those four also predicted the other divisions well to place that high but not as good as Dominick in 1st. The other 18 of the top 22 predictors currently have the same score in the ALC. As a result that division is not important at this time in your overall standings.
As for what makes predictions of those predictors on the first page better than the last page, Gerry has that well covered in his 'keys' section of his web site. Simply put, predictors on the first page have more keys correct than those on the 4th page.
You are correct in your comment than in theory it is possible to predict one team poorly while predicting everything else in all other close divisions correctly and be behind someone that just did the opposite. But in actuality that seldom happens as 4 or 5 of 6 divisions seldom are not close from top to bottom. Also keep in mind it is not only what you predict but what the other have predicted also. For example if you were one of 5 players predicting the Angels 1st you would pick up 18 games on almost everyone (9 games up as of today) and that would be a great pick. But if 189 of 194 pick the Angels 1st it doesn't make much difference because everyone has the same team. The key to a good prediction is having most of the keys correct including a couple that most don't have (like Wash last).
I think the key thing to
remember is that in our opinion it is a much better prediction
if you missed a teams place by one game rather than 10 games as
you were a lot closer to being correct. Thus following that assumption,
the low totals on all 30 teams is the best prediction.
Your comments are welcomed. Thanks again for your interest and your prediction.
July 23/11:21 Some New Numbers
By now you are aware that I have shared your comments with the Slagle brothers. I always do that when ever siginificant comments of suggestions are submitted. You see, Bob, this is Nelson Slagle's game (he is (was) the 16 yr old HS Jr.) and I am simply the current caretaker of the game. I have always valued their judgement and would never make a change to the game without consulting with them first.
I have taken this opportunity to add a pair of temporary pages to the site. When you go to the Top3 page you will find links to Temp1 and Temp2 pages. These are temporary pages and will go away by the end of the week. Here is the link to the homepage:
From there go to the Top3 page where you will find a pair of new links. These pages represent copies of the Top3 page for which I have retained the top two players and added two new players: Patty Elmore and Bob Schaffer. Temp1 illustrates the current scoring based on last Monday's MLB Standings while Temp2 illustrates your suggested scoring method. (no longer listed on the site)
I have used Patty Elmore because she is the only player that has nailed the White Sox and Twins in AL Central and currently has a huge advantage over the field in that division.
Here is a recap of these four players from the two new pages:
Some observations of
The most obvious difference in the two methods is how much that the margin between 1st place and 180th has been reduced (90 pts to 13 pts). The reason for this is because in the suggested method there are absolute limits on both the high (122) and low (46) scores, and since no one will ever come close to these limits then I would guess that most player scores would be concentrated between 80 and 90 pts. FYI, my current score based on the new method would be 91 pts while the score for the current last place person (Henson) would be 82 pts. This would bring the tie breaker into play and the game would become much to dependent on the TB. This condition, alone, would lead me to dismiss any additional consideration for the suggested method, but it gets worse.
Also the new system would shift the significance of certain divisions, putting a greater value on the NL Central (6 teams- 30/12) and less value on the AL West (4 teams- 12/4). As an example, currently the top two players have a big advantage in the AL West where Dominick has nailed it and Dave has just missed. They currently have a 26 and 25 pt advantage over the field while in the suggested method that advantage virtually disappears.
And finally, the last nail: As I have mentioned before I have no idea how to mechanize such a game. I have had to manually compute every value on the two temp pages. This function alone builds in an additional risk of errors, lots of errors .... Hold the presses!! It can be mechanized, even within the current AW spreadsheet..........
After writing yesterday's message, I thought about it for few minutes and realized that the existing spreadsheet would easily handle your suggested scoring method so this morning I decided to update the MLB Standings and run the data both ways. It's no big deal.
A pair of Excel spreadsheets are attached. Please take a look a look at both files and compare them. If you have a problem with them please let me know and we'll try something else.
I'm not sure if the tie breaker works but I think that it does. It now represents the current position for each player's top division teams, ie: 3 pts for Boston; 0 pts for Cleveland, etc. My score is correct but I haven't looked any deeper.
As expected the range of scores is quite limited:
High score: 98
Low score: 75
I thought that the low score would be much lower. If we look past the top and bottom three scores, there are 192 players crammed into just sixteen (94 to 79) scores and an endless series of ties. The current tie breaker just doesn't cut it. This point alone, makes this optional scoring unexceptable for me.
Yes, the player standings are much different as we have a different leader and a much different player at the bottom, but this would probably be the case with any major shift in scoring methods.
One more obervation: The suggested scoring method takes out much of the day-to-day and even week-to-week player movement as teams don't change positions that often. It would also take the Keys page out of play.
Attach: PlyrStds_Jul 24th.xls, PlyrStds_Jul 24th/Opt.xls
Wow, I really didn't expect this much action so fast. I am sort of a little stunned (in a good way). One thing I did want to say regarding the last email you sent was that at least from what I can gather that your reactions are tending to resist, based on what are really quantitative factors it seems. And my initial query into this whole scoring matter was more to answer the question of what does it all mean? Meaning that it is very much a qualitative matter to me.
I saw the game objective you re-sent me the other day and essentially it states that the game objective is to pick the correct order of finish of all 6 divisions for the MLB season. It goes on to say that the game will be scored by such and such a method. My initial objection was not that the scoring was not clearly stated in the game rules, but that I wasn't quite sure what the scoring system was in fact measuring. But in order to answer that question I guess we (or someone) needs to answer the question of what is considered an optimal prediction ...one in which every team is out of order but cumulatively by fewer games out since it's a very tight division (like the ALE), or one in which 1 team is out of order but that team is so far out or in front that the number of cumulative games back exceeds the other case where every team is out of order. I hope that was not too confusing. But it's an essential question in my mind for what the contest is trying to achieve.
As I may have mentioned to you, this question occurred to me last season but I didn't take it very far. However, it seemed a bit odd to me recently seeing how Dominick was sitting in 1st place even as he had called for the Tribe to win the AL Central and the Rays to finish in last in the AL East if I recall correctly. Now I'm sure a lot of people picked something similar to that, but out of 190 something entries some picks had to be a little closer to reality, or so I thought. Well, I haven't looked through the new spreadsheets (isn't Excel wonderful? NO!!), but I intend to do so and will come back to you with some comments.
Again, I must say that I am really pleased that you took the time and effort to look into this further. And let me say also that I came up with the scoring method I suggested in a few minutes and so there is no way I'd be willing to say that it's a superior method. But I do expect that it would yield some different results. And based on what you've said it would require an active TB system. So going back to what I wrote in the opening paragraph, if I may have stumbled onto a concept that scores the game more accurately but needs further refinement, wouldn't that be better in the end?
Thanks again and I look forward to discussing this further.
July 24/14:43 Fm John
Hi Bob. We appreciate your response, as you can perhaps tell by our lengthy responses!
In developing a scoring system, it of course has to be "quantitative". The scoring system calculates (or perhaps "measures" to use your word) the sum total of differences in games between actual and predicted results. Or, to use your word "measured", the current scoring system is the best system that we know to measure how good a prediction is.
Somehow a penalty needs to be associated with the difference between actual and predicted results. The essential difference between the current and proposed scoring systems is that we believe that a predictor should be penalized more for predicting a team that is many games away from their predicted position than a predictor who picks a team only a few games away from their predicted position. For example, in the NLC example that you initially raised, when you wrote Milw was ahead of St L by one game and St L was ahead of Cinci by 8 games. If predicting using the proposed scoring system, transposing the Milw and St L teams (one game diff) is scored the same as transposing St L and Cinci (8 game diff). That would be unfair. St L is clearly outperforming Cinci this yr, but Milw might tie St L with the bounce of a ball in one game.
Regarding Dominick, there are no picks that were "closer to reality", at least under the current scoring system. Prediction errors are only bad if few other predictors made the error. 185 people picked Cle first or second out of 194 predictors. Relative to other predictors, that wouldn't hurt Dominick much under either scoring system. Only 1 person picked Cle last or next to last, and that predictor had 29 teams to predict and apparently wasn't as good predicting those 29 other teams. Hence, it should not be surprising at all that Dominick is in first place even though he picked Cle first. If that was his most grievance error, he made an amazing predict because virtually every other predictor predicted Cle far too high (based on current standings). I haven't looked at Dominick's picks, but I'm sure that he has prediction errors that hurt him on a relative basis more than this error.
I do understand your main point, and good point, that missing one team badly can override an otherwise excellent prediction. But like Nelson said, that rarely happens. Somehow a penalty needs to be associated with the difference between actual and predicted results, and we believe "games back" does it best. And if a predictor misses one team very bad that very few other predictors miss, then on a relative basis that predictor should face a severe penalty - a penalty that arguably should override a number of other good predictions where the differences in games back was minimal. Minimal differences in games can be due to luck, and a good scoring system should minimize luck. If a predictor picked the Angels last (nobody did), they should pay the price for a huge prediction mistake that nobody else made.
There is one way that comes to mind that essentially represents a scoring system in between the current and proposed scoring systems and that addresses your primary concern Bob. Instead of taking the difference in games, you could take the square root of the difference in games! I am not suggesting this be done at all. Imagine explaining this to the predicting pool who are not mathematically inclined. However, it would address Bob's basic point which is that a predictor could miss one team that is, for example, light years in last place, and that one mistake could override an otherwise perfect prediction. Square root (SR) examples: 1 is SR of 1, 2 is SR of 4, 3 is SR of 9, 4 is SR of 16. Hence, one bad miss doesn't blow you out of the water as much as the current system can. However, it's too complex. Imagine manually checking to see if the spreadsheet had no errors - yuk!
Nelson, congrats on developing what I think is the best scoring system that I have seen as a skinny, 16 yr old junior in high school!
Bob, thanks again your response!
You may well predict differently with a different scoring system.
I think that may be a key to the game. I may have predicted differently, though who can really say?
The objective of the game is really to have the fewest cumulative games out of the order predicted - that's essentially it. It's not to have the order closest to the correct finish, as long as games back is part of the scoring.
Again, I haven't had a chance to analyze the spreadsheet with the proposed? scoring system, but I wonder if it yields results that appear to reward those contestants who are closer to the actual results. That was my only intention.
Thanks for your time and interest in my comments.
Regarding the issue you raised on the relative significance the NL Central would take at the expense of the AL West, I think those scoring numbers would be justified because you could correctly argue that it's more difficult to predict the order of finish in a 6 team race than it is in a 4 team race... no? It may even be mathematically proven that the point differentials in those divisions are appropriate, though I haven't given it the time that would deserve.
The issue all of this would address I hope would be to even out the "distortions" caused by teams with either enormous leads or teams that have completely tanked. For example, if I had predicted the Padres to finish in 4th and the Giants to finish last, but it ends up that the Giants finish 4th and the Pads finish 15 games behind them, under the current method assuming I had the rest of the division in correct order of finish, that would cost me 30 points (the 15 GB both ways, I think). I look at that prediction and say, "I got that pretty damn close." But the game views it differently. By the game's standards, that's a pretty egregious error of judgment.
All I know is that at the beginning of the year, I try to predict the correct order of finish as best as I can. I don't give any regard to how far back a team may finish in or out of each place, as I'm sure most others don't. I don't think it's relevant, or should be relevant to the objective. What I think mitigates some of the effect of these "outliers" is that they affect everyone equally and so their effect is somewhat offset all around the board. Again, I'd like to look at the results of the proposed scoring method to see if it represents any better a reward system.
July 24/16:54 ....There is one advantage of Bob's system that hasn't been mentioned though. Since big swings can occur with the result of a single game causing a change in the standings, it would be far easier to make a quick move up in the standings. The game would more often go down to the last few days. So it would be more exciting from that standpoint for the predictors, but I don't think as fair.
Good points John. But I doubt more that a handful of predictors follow their predicts at all.
July 24/:59 (first message to Pete Palmer- ggh)
As you might have gathered from my recent message the method that we use to score our game has been called into question and another method has been suggested. I have compared this new scoring to ours and it is quite unsatisfactory.
Today I offered to compare our two methods to the one that you have now used for over 50 years. Yours appeared, to me at least, to be close to what he was looking for but he doesn't seem to get it, not yet at least.
So that I might be better prepared to field his next series of questions, I wonder if you or Steve have ever given any thought to our very different methods of scoring: difference in team position squared vs games out of predicted position. If you and Steve have talked about it or if have any thoughts on the scoring, would you mind sharing those thoughts with me.
No urgency. I'm going to close this discussion down in another day or so anyway but I thought that I would at least ask first.
Also hope that you are enjoying the current season. Not at all what I had expected.
7/25 9:59 (edt)
Geez, can't we settle on a method where I come out at least in the top half!? Just kidding of course.
Attached is a simple illustration of the effects of scoring the different methods which you've probably noticed already. Take a look at these 2 hypothetical divisions and simply answer the question of which was the better prediction. As you can see, the NL West prediction was nearly spot on with the exception of the last 2 teams being flipped. But the NL East prediction is a mess. Not one team was picked in the correct order of finish. But as you can see, the two scoring methods yield very different results for these 2 predictions. The current method yields a score that's 25% worse in the NL West than the NL East prediction which appears to be the far worse prediction. The proposed method however rewards the NL West prediction for its accuracy and appropriately penalizes the NL East prediction for its inaccuracy.
As far as the relative point weighting in the NL Central vs. the AL West, one could argue that it makes sense since it's much harder to pick the correct order of finish in a 6 team field than it is in a 4 team field. Therefore, the extra points available in the NL Central are
"earned" in a sense.
In any case, this is fun to look at but I don't think I can come up with a better illustration than the one attached to present my case. If you still feel your method represents the spirit of the contest in a better way, then this has little place left to go.
All the best-
nb: please let me know if you have any trouble with the attachment.
|NL West||NL West||W-L||GB||Scoring||Scoring|
No problem with your attachment, Bob, and I was already aware of how the different scoring methods yield strikingly different results. I certaily hope that you are wrong about San Diego although they do appear to have quit the season already. I will sum up my arguments and conclusions later in the day and then (hopefully) we will all move on.
I have spent most of the morning wasting my time pulling all of our collective correspondence into a single, time phased document. It is now twelve pages long- this will make it thirteen- and will probably continue to grow a little more.
Then I'll try to sum it all and list our reasons for retaining the game has it has been for so long.
If either of you have any closing thoughts, now's the time...
Nothing to add. It's all been thoroughly explained to Bob. If I were to respond to him it would have been brief, but nothing really new. Basically, as I see it, he uses one argument, and that argument is debatable. The main argument for the current system probably isn't debatable (ie a huge diff in games back should count against a predictor more than a small diff). Also not debatable is the tradition and consistency of the game. And like Jim said, it is your (you being Gerry and Nelson) game so you can run it as you see fit. Bob can choose to participate or not participate. The scoring system certainly can't change for the current year, although it almost seems like that is what Bob is arguing for! I agree with what you and Nelson said earlier, it is time to move on.
I agree with John. There is no perfect scoring system. We are allowed to disagree on what is best (however the game operator gets to decide). I think it is hard to argue that missing a team's place by 1 game should count the same as missing a team's place by 10 games.
July 25/18:44 Summing Up
It is time to sum up our arguments, make a decision and move on.
First of all I would like thank you, Bob, for your comments and suggestions. Too often, most of the game's participants make no input to the game and as a result we rarely get even a glimpse of our fellow predictors reaction to the game. We can only hope that they enjoy the challenge enough to keep coming back every year. While we don't often make changes to the game or the way that it is played, we are always open to new ideas because we know that there is always the potential to make the game a little better. It is after all, just a game.
Background: Bob has made the observation that the the current scoring method applies a big penalty when there are significant gaps between teams in the MLB standings, while failing to take into account when certain teams may be only one position out of their predicted order of finish. Bob then suggested a revised scoring method based entirely on division standings on a team-by-team basis without regard for games behind method currently used.
Actions: First a pair of new Top3 pages were prepared and posted on the site for all to see. These new Top3 pages retained the current top 2 players and then added two additional players to be used to demonstrate both the current scoring method and the suggested new method. Later, when it was realized the the existing spreadsheet would accommodate the new method another spreadsheet was prepared to demonstrate the two methods for the entire 2008 player field. When the suggested scoring method failed to provide satisfactory results, and alternate method, Pete Palmer's squares method, was offered as an alternative and a third spreadsheet was created.
Results: The new method when applied to the entire player field gave a very focused range of results with 97% of the total players sharing a very limited range of scores. This would bring the tie breaker into play which also yielded a limited range of results. It would make the final results almost entirely dependent on the tie breaker.....
Tradition: This game is now in it's 49th year and, with a few minor exceptions, it has remained essentially the same since it was first conceived in 1955. It was in the Slagle family for many years before I ever became a part of the game. Their opinion still matters and I value it very much. The suggested change in game scoring has been rejected by both Nelson and John, and now Jim, because the the "Games Behind" portion of the scoring is an essential part of the game and the new method fails to take that into account.
Recommendations: The suggested new scoring method is rejected for the following reasons:
1. The suggested method is much less satisfactory than our existing method because:
-- It fails to take in to account the "games behind" factor which is so much a part of the game and it takes the "Keys" out of the game.
-- It is a rigid scoring system that yields a fixed an unacceptably narrow range of results and would be almost entirely dependent upon a tie breaker to determine the final results.
-- Using the suggested system, MLB division standings would yield little change in the day-to-day and week-to-week player standings as teams simply do not change positions in the standings that frequently.
2. The suggested method represents a dramatic departure form our existing game and would represent an essentially new game:
-- Results yielded from the new scoring method would not relate in any way to our existing data.
-- Moving to an essentially new game can not be adequately explained to those players that have become experienced in the way that the game has been played for so many years.
3. The suggested new score method has been unanimously rejected by the creators on this game. That in itself would give me cause to reject if there were not so many other good reasons.
Once again, Bob, I would like to thank you very much for your show of interest. It is deeply appreciated.
July 26/13:27 (edt)
Thanks again for all the attention my comments received. I feel like this somehow turned into a contest over the last week and my entry (which was really sort of an off the cuff sort of thing) didn't make the final cut. As the week progressed, I attempted to boil this thing down to a simple question of what represents a better prediction and I illustrated that question with the examples of the NL East and NL West divisions. By your answer below, I can only assume that you guys believe that the NL East was the better prediction in the example given. I simply don't agree and I just wonder how the rest of the entrants would view that basic question.
As far as the problem of the scoring distribution in the method I suggested, I'm sure a little tinkering with it could yield a more disparate distribution. I don't really think that's the issue. You probably could widen the distribution by simply assigning a wider range of point values to each place in or out of order (i.e. instead of 4,3,2,1,0... 16,12,8,4,0 or something such as that). My point was about the general concept of trying to match the scoring to what should be considering a good vs. a bad prediction. I realize the tumult involved with changing the game's scoring system and I wasn't really seeking that sort of thing; only to get a better understanding of the current system, and also to show other ways of looking at the scoring.
Again all the best,
Bob, Bottom line: Any system that is based on a rigid set of fixed values and does not account for "Games Out" part of the scoring ain't gonna to cut it. The "Keys" are the key to the game and I'll try to make that clear in any future game invitations. Also, if you want to modify the system in any way, here are two things to keep in mind:
1. Start with zero (0)- Low score wins!!
2. Keep it simple (KISS it)!!!!!! or you're going to lose me.
Thanks again for your suggestion. I'd rather receive criticism about the game than nothing at all.
Bob, in your example of the NLE and NLW, which is a better predict is debatable. Based on "games back", the NLE is a better predict. Based on placement of teams, the NLW is a better predict.
What is not debatable is that a predictor should be penalized more for predicting a team that is many games away from their predicted position than a predictor who picks a team only a few games away from their predicted position. For example, Washington is in last place by 11 games and Houston is in last place by 0.5 games as I write. Your scoring method makes absolutely no distinction between picking Washington next to last and picking Houston next to last. Clearly the predictor who didn't pick Washington last made a much worse prediction based on results than the person who didn't pick Houston last.
Further, it is not debatable that a major change in the scoring system would alter the tradition, consistency and understanding of the game. It would take a clearly better scoring system for us to alter the game as such, but we believe the suggested scoring system is clearly worse.
You are focusing on one isolated example which is debatable and continually ignoring the obvious faults of the proposed system.
July 29/8:18 (edt) (Pete Palmer checks it)
I haven't taken part in your discussion about rankings. I never tried to analyze your method, but I was confident it was valid, since you have more experience than me. Mine is very simple but has some shortcomings, in that it doesn't matter how many games you miss by and also it gives no extra credit for picking the leaders as opposed to any other position. However, it does have an actual technical name, the Spearman Rank Coefficient. If you manipulate the numbers, you can get a value between -1 and 1.
For a 4-team division, the coefficient equals 1 - squares/10, so a sum of squares of 0 gives a +1 value and a sum of squares of 20 gives -1, for 6-teams it is 1 - squares/35.
Steve (Mann) and I never discussed it. He just accepted what I was doing. Actually we use two methods, one is squares of the differences in position and the other is squares of the differences in wins. The wins method we always thought of being more meaningful because it was more detailed. If one guy had the teams in the same order, but the first place gap was bigger, then his wins difference would be smaller if the team actually won by a big margin. In order to be the champion between us, you have to win both methods, otherwise we consider it a draw. We couldn't use that to compare with others, though, because most predictors we were up against didn't give wins, just position. Our main goal was to have our consensus picks be better than the consensus of the various magazines we were comparing against.
I am happy with your method and see no reason to change.
Thanks for checking in. You give me too much credit, but thank you for explaining you your method. If I can figure out a simple algorithm- but probably not- I'll edit a copy of my spreadsheet and see how it works out.
Also, a friend and fellow SABRite has requested to the 15 page discussion so it will be added to the site later today, including your response. If you or Steve are interested, you can follow the discussion there.