Posts:
542
Registered:
May 9, 2002
From:
Northeast USA
Should Pente "Sets" be the Norm?
Posted:
Sep 30, 2003, 8:46 PM
Greetings to all Pente players:
I believe this might be my first new post, and that's because I've finally found a topic which I believe is important, that I would love to hear many players' view about, and that hasn't already been posted (at least not recently). That topic is ratings.
Specifically, how can we improve upon the current ratings system? I am proposing my solution here.
So, what are "ratings" anyway, and why are they important? Well, I'll tell you. When I first started playing pente online here at DSG, I didn't give a hoot about my rating one bit. I was more interested in playing friendly games and in socializing about topics related or unrelated to pente. For many players, this is fine, and I felt this way for quite some time. Eventually though, I became more interested in the game and became a better player and I watched my rating climb. As I improved, it was nice to have an easy estimate of how much I had improved and how far I needed to go to become a top player.
Ratings are important for the following (and many other) reasons: - It helps you to identify in the "lobby" area, which players you would rather play and which ones would be a "waste of time". Most players enjoy playing other players at or near their own skill level as that gives them the most fun and competitive games. A beginner playing an expert is often not as fun for either the beginner or the expert. - It gives a decent indicator for your own skill level. This means that you know your game needs improvement and also let's you know what level of opponent you should be seeking out. - For competitive minded players, it gives an extra spark of interest that otherwise might die out after "trying the game out". It gets players more "hooked" on playing, which is what we want more of as a pente community. - Ratings give a positive indicator over time of how much you are improving. You can watch your rating climb over time and be excited to know that you are improving, again fostering more interest in the game. - Ratings are great for learning. If you see higher rated folks playing at another table, you can watch how they play and absorb some of those principles into your own game. Even better, here at DSG, if you know of a very highly rated player that you are interested in seeing how they have played many games in the past, you can do that! Just look up the name in the Games History database and watch the geniuses of the game at work. - Lastly, ratings are supposed to be able to distinguish and acknowledge the very top players of the game. They should give the top players something to shoot for (to become #1) and foster interest and innovation at the highest level of the game.
It is this last point that I believe the current rating system is failing miserably and is where I will focus the rest of my attention here.
The basic problem is one that is discussed over and over and over again on the forums and amongst the top players of the game in general: "Player 1 has an insurmountable advantage".
As players become more and more proficient and advanced, this realization becomes more and more clear. As a result, the highest level players generally lose interest in the game and "retire". WE DON'T WANT THAT! It is best for the future of the game that the most skilled players remain actively playing and pushing the envelope.
However, I believe strongly that this "player 1 advantage" problem is compounded 100 fold by the way the rating system is currently handled. You see, in general, the top players take their rating VERY seriously. And every point becomes very hard earned. The current rating system is excellent for showing that a 2000 rated player is better than one rated 1700. 1700 is better than 1400. 1400 is better than 1100. Fine. But is a player rated 1984 better than one rated 1965? That's really hard to say, but I'll tell you one thing. It's hard as hell for that guy rated 1965 to become rated 1985. Why is that? Because that 1984 guy won't play him. In fact, that 1984 guy probably won't play anyone rated below 2000. Or above 1400. WTF? you ask? Well, the reason is simple and can be summed up with one phrase: "Player 1 has an insurmountable advantage".
Ah, it is all becoming more clear now. Lets pick a random example. Let's say a 1750 player challenges a 1950 player to a match. It is assumed by both players according to pente etiquette that each player will get one game as player 1 and one game as player 2, out of "fairness". Now, both 1750 players and 1950 players are both good enough to be considered proficient enough at playing from the Player 1 position to be unbeatable in most cases. Thus, it is expected with reasonable confidence that the 1950 player will win one game and the 1750 player will win one game because of the "player 1 advantage". So the match will be a draw, right?
BUT WAIT!!! The match is not really a draw under the current rating system. The ratings were adjusted twice. Once after the completion of each game. For the exact formula used to adjust ratings here at DSG, I refer you to the DSG main page -> Getting Started -> General FAQ -> How are the ratings calculated?
Ok, so back to our example: Let's say the 1950 player starts out in the Player 2 seat and loses. His rating is adjusted down to 1925.69 and our 1750 player gets his rating adjusted up to 1774.31. Note that this is a large adjustment because the 1750 player is considered the "underdog". Now, let's further assume for a moment that our 1750 player does NOT just take his points and run (which is another problem with the current rating system that I will not discuss any more at this time) and the two players switch seats. Now, as Player 1, our 1950 player wins. His rating is adjusted again, this time from 1925.69 to 1935.13. Our 1750 player goes from 1774.31 to 1764.87.
Yes, you read that correctly.
Yes.
Of course the 1750 player would challenge the 1950 player to a match!
Sadly, Of course the 1950 player would decline.
This is a flaw, and a serious one. The trend is that this 1950 player simply stops playing. And it really isn't his fault. His position is understandable. And while they will claim that they are bored with the game due to the "player 1 advantage", I believe it is that problem COMBINED with the associated ratings flaw described here which drives away the top players. Realize that this concept in practice does not affect beginner, intermediate or even advanced players. As you improve, you will see your rating improve, despite this flaw. It is only when you get near the top of the game that this becomes such a handicap that it cannot be overcome. HOWEVER, we should ALL be concerned about this. Not just because we may find ourselves at the top some day, but because it is for the good of the game that the top players keep playing.
For my last reason of why the current rating system fails, I will go back to my last example of the 1950 guy vs. the 1750 guy (hell, I know they can be gals, that's not the point here). As calculated previously, if the 1950 player plays first from the player 2 seat and loses, then plays from the player 1 seat and wins, his rating becomes 1935.13 and the 1750 player becomes 1764.87. Now, suppose our 1950 player decided to play first as player 1 and wins and then plays as player 2 and loses. This is also a "draw" right?
Um, no.
FURTHERMORE, not only is this not a draw, but it's not even the same draw as the previous example! WTF? you ask? Ok, I'll do the math for you. In this case, our 1950 player ends up with a rating of only 1932.87. Our 1750 player ends up at 1767.13.
Yes, you read that correctly.
Yes.
IT IS ACTUALLY BETTER TO PLAY AS PLAYER 2 FOLLOWED BY PLAYER 1 THAN TO PLAY AS PLAYER 1 FOLLOWED BY PLAYER 2!!!
The difference is small and can sort of be wished away by the fact that occasionally, when you start out as player 2 and lose your opponent will bolt. But, dare I say this is imprecise and two wrongs certainly don't make a right.
THE SOLUTION
My solution is that because of the "player 1 advantage" that is inherent in pente, a match in pente should be played in "sets".
What is a set? A set is a "set of games" consisting of Player A playing in the player 1 seat vs. Player B playing in the player 2 seat AND Player A playing in the player 2 seat vs. Player B playing in the player 1 seat. In other words: Two games of pente make a set, players switch seats between the two games.
It is a concept that closely resembles a tennis match. In tennis, the player who serves has the advantage. When a game is won by either player the other player gets to serve. This advantage alternates back and forth. It is an uncommon yet important event to "break" your opponent's serve -- to win the game despite your opponent being the one to have the serving advantage. But it is not so rare as to ruin the game. And always, the "win by two" principle applies -- someone must break serve at some point to win the match. I believe this parallels closely to pente matches at the top level. A player winning a game as player 2 is an uncommon event, but it does happen.
This concept is already being used in practice at the speed tournaments that are often played at DSG. Each time you play pente against an opponent is a "game". After each game, opponents switch sides. Two games make a "set" (set #1 is game #1 and game #2, set #2 is game #3 and game #4, etc). A match, in a speed tournament, consists of however many sets it takes for one player to win a set, 2 games to 0. No matter how many games are won and lost to achieve this result, this series of games is considered to be ONE match -- the winner has won one match, the loser has lost one match.
I believe strongly that this concept should be incorporated into regular timed real-time games at DSG and that the rating system should be modified to reflect this concept. I propose that when two players agree to play a competitive, rated match at DSG, that this match should be played according to this "set" approach. The basic principle is that you agree to play one set -- one game as player 1 and one game as player 2 against the same opponent. (The series of dialog boxes and buttons that would enforce this principle is beyond the scope here but I do not think it would be unwieldy at all from a development or from a user perspective.)
At the end of the first set, if one player has won both games, he shall be declared the winner of the match, and the ratings shall be adjusted using EXACTLY the same formulas as are currently in use. If the set is split 1 game to 1 (whether both are player 1 wins or player 2 wins), the match is declared a draw and RATINGS ARE NOT ADJUSTED.
Repeat, when a set (or match) ends in a draw: RATINGS ARE NOT ADJUSTED
By writing the above statement, I have just saved the game of pente.
I'm serious. It's that important.
CONCLUSION
The reason why this concept is so important goes back to that last pesky bullet about why ratings are important in general. The fact is, top players are "retiring". Folks that are excited enough about the game to create and maintain a site for us all to play at (THANKS DWEEBO!). Folks that are excited enough about the game to make contacts with the manufacturers, to create a player's association, to organize tournaments, to spread the word, to create newsletters and AI programs and speed leagues. Folks that strive to be the best. These folks are retiring. Top players retire when they lose interest.
This phenomenon is not a new one. For many years people that reach that inner circle of top players have witnessed this attrition problem occur. It has always been cited that the problem is the "player 1 advantage". A flurry of activity has recently occurred where folks are looking for that perfect rule change that will eliminate this player 1 advantage. I have news for you all. It doesn't exist. The simple truth is that in any turn based game, one player always has the advantage. In pente, it is more noticeable since player 1 actually has an extra stone on the board. Tweak the rules a bit and perhaps player 1 will no longer have the advantage, but in this case player 2 will then have the advantage, it can never be exactly equal. I am a big proponent of D-Pente, which incorporates a swap rule within the game. This evens things out significantly but the fact is that player 2 has a slight advantage here. It can never be completely equal.
So why fight it!!!??? Simply play a set. No matter how you slice it, when you play set you have a completely exact, 100% totally and unequivocally even playing field. Both players have exactly 0 advantage to win the match other than their skills.
However, if the rating system is not changed to reflect this, top players will not play. Plain and simple. I believe with a set based rating system, top players will feel refreshed and uninhibited and will flock back to the game in droves. They will seek to play players rated higher than they are since they will score more points if they win and they will not mind playing lower rated players because they will not lose points unless they lose and will win at least a small number of points when they win.
In other words, top players have enough confidence in their game to assume that they will win as player 1 and so the worst that can come of playing a lower rated player is a draw -- which does NOT adversely affect their rating.
Top players will play.
I can understand that many of you will have trouble wrapping your brains around this concept and why it makes so much more sense because of the following argument. I've often heard the reasoning that if that 1750 player can "manage" to split games with the 1950 player then he is basically just as good as the 1950 player and his rating should indeed go up and the better player should be adjusted downward. While this argument sort of coincidentally works out to be true for beginner or intermediate players, it is simply invalid for top players. I don't know how else to explain it. For top players this argument is just plain wrong. And thus top players do not play. To play is to be "penalized". Playing sets just makes more sense.
FEEDBACK
I would LOVE to hear lots of views on this subject. Obviously, my view is that this is not only the right way, but it is critical to be implemented as quickly as possible. In an ideal world Dweebo would throw his entire "To Do List" in the trash and focus immediately and completely on bringing this policy forward. Most of you will disagree. I would love to hear it. Please post your opinions because I think this is an important topic. Especially top players, please express your opinion.
P.S. I apologize for the length of this post. This is more of an article than a forum post really and I would have waited to submit this issue into the WPPF newsletter but I didn't know how long it would be until the next issue goes out so I'm posting it here and I look forward to seeing your responses.
well said. I had always assumed that when I was not able to win BOTH games against a 1750 player that it was because I am not good enouhg to do so, and thus I did not deserve to have a 2000 rating, and thus my 1950 rating was fair.
Perhaps this is the case, but perhaps it is not. I do agree that when a 1750 player and a 1950 player split, it does NOT mean they are of equal skill. I DID get frustrated by frequently getting my rating to the high 1900s (even 2000 once!), only to have it drop 18 points after SPLITTING with an 1800 player. But, since there are so few players over 1800, a player in the 1900s is usually forced to try to eke out a few measly points against 1700 players who are veyr good, especially in the player 1 seat.
ALso, sometimes I have been expected to play several games in a row in the player 2 seat because of my high rating. I don't think that is fair. As up2ng said, the top players take their rating very seriously, myself included. Others may not think that we should, but that's not their choice. If they want to play fast and loose with their own rating, that is understandable, but they should not expect others to do so. This is especially the case since the ratings at dweebo's are used for very important purposes, such as seedings in tournaments.
Message was edited by: speedmitri at Oct 1, 2003 7:17 AM
Re: Should Pente "Sets" be the Norm?
Posted:
Oct 4, 2003, 4:58 PM
Nice job up2ng. I think this is a great idea; one other thing is that perhaps an optional feature of whether to implement ratings based on games or sets would be useful. I think that many times players who are more inclined to play socially would like to have a per game rating; keeping that as an option while also allowing players to choose to have the ratings changed based on sets might be a solution which would please everyone.
Retired from TB Pente, but still playing live games & exploring variants like D, poof and boat
Re: Should Pente "Sets" be the Norm?
Posted:
Oct 6, 2003, 2:47 AM
One other comment I wanted to add after thought is this: even though playing games by sets makes the game fair, it still is affected to a great degree by the player 1 advantage, so in my opinion developments such as D-pente are still necessary to make the game more even. Here is the reason- if two players are relatively equal in strength and know how to use the player 1 advantage to some degree(these could be 1400 players as easily as 1900 players) then it becomes very difficult to determine which of the two players is better because the first player will tend to win. One player might be 30 points higher in skill (arbitrary figure) but the player one advantage will probably be more than enough to offset this, and so the players may trade many games over a period of time without being able to establish which player is better. Therefore, the more we can level the advantage, the more meaningful the ratings will become.
Retired from TB Pente, but still playing live games & exploring variants like D, poof and boat
Posts:
542
Registered:
May 9, 2002
From:
Northeast USA
Re: Should Pente "Sets" be the Norm?
Posted:
Oct 6, 2003, 7:06 PM
Thanks to those that have responded so far. I hope to hear from many players on this issue.
Dmitri made another point about the importance of ratings that I had overlooked which is that the ratings here at DSG are used as a tool to seed players in the big online tournaments and also in some in person tournaments. This makes highly rated players even more shy about playing. I have already pointed out that "splitting games" with a lower rated player currently hurts your rating and that top rated players take these ratings very seriously. But the fact that these ratings are used for tournament seedings, as they should be, means that the top rated players have even more to lose by letting that rating slip since they might be seeded lower in the next tournament, drawing tougher competition in the earlier rounds than otherwise. Remember, we want top players to play. It's good for the game.
This additional importance placed on ratings is significant in the minds of many top players. I have actually witnessed several top players manage to boost their rating to its peak with a streak of wins and then "retire" that username until seedings for the next tournament are fixed. Meanwhile, they create a new username and continue playing with this alternate name. My feeling about this is mixed. There is nothing in the DSG site rules to prevent users from doing this, but you can see how this is unfair to opponents who are now facing the same difficult opponent but with less to gain by winning since this player has not yet brought his new username up to the same high rating as his "retired" username. But at the same time, it's hard to blame them for doing this under the current rating system.
Of course, the other thing that top players do instead is to not play at all -- which is even worse.
Swatsu also makes good points. The first point he makes is that the new proposed system of playing "set-based matches" be optional. I forgot to mention this in my proposal, but my take on this is that the easiest and IMO the best change that could be made is to require ALL rated games to be played based on sets. However, if someone doesn't have time for or isn't comfortable playing a set -- PLAY UNRATED. I propose that unrated games would remain as they are currently -- played as individual games. I can see that there will be some objection to this. But remember, even the very top rated players play unrated games all the time when they are not "in the mood" for a rated game, and for many other reasons. Rated games should be reserved for competitive matches. Otherwise, play unrated. Also, whatever the rating system is, it should be consistant -- otherwise each player would be achieving a rating under different conditions, which reduces the accuracy of the rating.
The second point was that, even in set-based games, the player 1 advantage still exists, which means that there should still be some efforts made to create rule changes to lessen the player 1 advantage. Based on this logic, I have to disagree.
However, just as a general principle, I DO agree that it would be a good thing to reduce the player 1 advantage within an individual game. But, as I have said in the past, I STRONGLY believe that a "rule change" must be considered with EXTREME caution. A change that actually changes an existing rule creates a fundamentally different game, and is no longer pente. In addition, pente was intentionally created with very few rules -- the simplicity is part of it's elegance and grace. Many stone games exist, such as renju, that have so many added rules and constraints to try to eliminate the inherent advantage that the game becomes muddled and confusing and "impure" -- not to mention that new players become more easily intimidated and quit. The current standard, Pro-Pente, which incorporates the "tournament rule" is an example of a very simple rule addition that trades off a very small amount of these drawbacks to make a very large improvement to the game. But every additional rule will be less beneficial and more harmful -- you have been warned.
My last argument on this point is that such a perfect rule addition does not exist. The simple truth is that in any turn based game, and especially in stone games, one player ALWAYS has an inherent advantage. There is NO WAY to make an individual game played have an even playing field. Sorry, but it's true. I only wish someone would go ahead and prove this mathematically so we can all accept it and move on.
But if we want to make it as close to even as it can get, then we should choose D-Pente as the standard. Even in D-Pente, though, there is a small Player 2 advantage (yes, Player 2) but the advantage is very small and is probably dwarfed by human error that results from playing a timed game. But this is a separate topic for another day. Currently, making D-Pente the standard is just wishful thinking. I do hope, however, that it gets added as an additional game choice at DSG VERY soon.
Ok, back to set-based games and why I think that finding rules that reduce the player 1 advantage is a completely unrelated issue. Let me revisit my favorite metaphor of a tennis match. In tennis, players alternate serves (player 1 advantage) from game to game, but to win a set, one player must win by 2. Now, historically, when the No. 1 ranked player in the world plays against the No. 100 ranked player in the world, everybody agrees that the #1 player is a far superior player to the #100 player. Does this mean that he will win every game in the match? Absolutely not! And that rarely happens. The #1 player might win the set 6-4, breaking serve only once. And in rare cases, the #100 player might actually win the set! But this is very rare, as it should be according to their rankings (ratings).
Now suppose for a second that every time someone wins a game, their world rankings are adjusted! #100 serves and wins -- 1-0, he jumps to #95, #1 becomes #6. Now our #1 serves and wins -- 1-1, he eeks back to #5, #95 becomes #96. Now, 2-1 -- #9 and #91. 2-2 -- #8 and #92. ..... Yey, our hero wins the set 6-4! But lo! He is now ranked #27 and his opponent has jumped up to #74! Hmmmmm. Why in the world would the #1 player in the world EVER play!? Clearly this would be rediculous.
I want to also remind everyone here that winning a set is not impossible. In Tournament 4A here at DSG there was only one match that took longer than 3 sets to decide a winner. One! In the entire tournament! In untimed games this might be different, but we are not talking about untimed games -- we are talking about ratings at DSG, where almost all rated games are timed.
Sorry again if I rambled! I hope many more of you will offer your opinion!
Re: Should Pente "Sets" be the Norm?
Posted:
Oct 7, 2003, 8:58 AM
I agree, up, that reducing the player 1 advantage in an individual game is best done with d-pente, and that the set solution works better tahn calculating ratings based on games. My point though, was that for closely matched players to determine which one is better is difficult given the player 1 advantage and ratings by sets don't change that fact. Even though only one match went beyond 3 sets in the last tournament, that most likely means that there were few matches between relatively equally strong players. I was making this point because in your original post you mentioned that ratings by sets eliminate the player 1 advantage. I don't think that we are disagreeing on this, I just wanted to clarify. Tom
Retired from TB Pente, but still playing live games & exploring variants like D, poof and boat
Posts:
55
Registered:
Jan 21, 2002
From:
ohio
Age:
65
Re: Should Pente "Sets" be the Norm?
Posted:
Oct 12, 2003, 11:08 PM
It seems to me that when 2 player of roughly the same rating play, if rating points are going to be an issue,then why not prorate the awards from 1st or 2nd seat. For example....a win by player 1 would be plus only 9 pts instead of 17 and a win by 2nd player be plus 25. Just an idea to more or less keep the GAME simple and let the RATING reflect reality.
Posts:
58
Registered:
Jan 19, 2002
From:
Peoria Arizona
Age:
43 Home page
Re: Should Pente "Sets" be the Norm?
Posted:
Oct 13, 2003, 1:11 AM
One other even more major flaw in the current rateing system is the fact that over time all the rateings will go down due to this one real and bad yet unavoidable fact. The fact is that most new players are not good players, and when they are provisional they usually loose most of the games they play, while they may loose 50 to 100 points the winner may only get 2-10, then after 20 games they become a real player, then they usually are sitting at 1200 or less points. This is sad because they started with 1600, and for that 400 point loss they probably only gave up like 50 points. So this just reduced the number of points in the system, thus reducing the average. So over time, years that is, it becomes harder and harder to get high scores becuase less points exist. I remember with the first 50 people on the top list were all 1600+ rateings, now it is only 22.
I know that players quitting also has great impact on the average...usually in the other direction.
Posts:
54
Registered:
Feb 21, 2003
From:
Hawaii Home page
Re: Should Pente "Sets" be the Norm?
Posted:
Oct 15, 2003, 5:18 PM
The reason there are only 22 players listed with ratings over 1600 is becasue most players are inactive, NOT because their play has deteriorated. Also, the rankings only tell us how we compare to one another. You can't use them to compare players accross time. So if the average rises or falls, it's meaningless.
