Why does it seem that the higher rated players
Posted:
Oct 26, 2005, 4:44 PM
don't want to play lower rated players ? Just because someone doesn't have a high rating doesn't mean that they wouldn't at least be a challenge to you (hopefully ) I don't have a high rating ... yet... but I do think that I am an okay player and can be a challenge most of the time, and I am not talking just pente, I love all versions including Keryo and go-moku. But yet it seems that no one ever wants to play. I think this is a great site, just wish for games against the better players, so that I might improve my game playing.
Posts:
542
Registered:
May 9, 2002
From:
Northeast USA
Re: Why does it seem that the higher rated players
Posted:
Oct 29, 2005, 12:43 AM
peanut, this is a good question. Unfortunately, the reason for this has more to do with the fact that the rating system is slightly flawed rather than people thinking that lower rated players would be boring to play against.
This problem is talked about at length in previous forums but the correct solution continues to be ignored. So unfortunately, people sometimes avoid playing games rather than jumping at every chance to play, which in turn, causes other players to lose interest and helps keep the lobby relatively empty all the time, which perpetuates the problem even more. In the best case, if we always had a full lobby, you really should be looking to play most of your games with players only slightly better than you (within 100 points). But for now, when you enter the lobby and there is a pretty good sized gap between the ratings of the few people who are there, you will continue to see this problem until it is addressed more seriously.
To summarize, the problem arises from the fact that pente inherently possesses a "player 1 advantage". This means, if both players play perfectly, player 1 will always win. This is true of all variations of pente (except d-pente, which actually has a "player 2 advantage").
This fact alone is not a problem. However, the rating system currently is set up to adjust ratings based only on the difference in the players' skill level (assumed from their current ratings). The other significant factor, who happens to be sitting in what seat, is not addressed in the formula. It is like trying to solve a problem possessing two variables with an equation containing only one variable -- it is inaccurate.
The only way to solve this problem is to remove one variable's effect on who wins a match from the mix completely. It can be proven that all current variations of 19x19 pente (and all other possibilities dreamed up so far) possess this P1 advantage so overwhelmingly that it is a significant factor in predicting who will win a single game of pente. No amount of rule changes can eliminate this factor for a single game.
The solution: Pente should be played in SETS, which means, playing two games in a row, one as player 1 and the other as player 2. This completely eliminates the P1 advantage from who will win a MATCH (a match generally means either a single set, which may end in a draw, or an unlimited number of sets played until one player wins 2 games of a single set). The key, however, is that if a match ends in a draw, the ratings of both players should remain UNCHANGED. Once these simple changes are made, there will never be a legitimate reason for a higher rated player to refuse a challenge from a lower rated player.
If players do not wish to play a set, they can always play unrated.
peanut, please do not let this behavior discourage you from coming to DSG and playing lots of fun and interesting games! It is a great site, which I'm sure you will enjoy as you play and continue to improve your game.
Posts:
141
Registered:
Feb 29, 2004
From:
the land of smelly cheese
Age:
38
Re: Why does it seem that the higher rated players
Posted:
Oct 29, 2005, 11:34 AM
i have to say that i agree strongly with up2 - i wish the game was always played in sets, and that a set resulting in a draw meant no change in rating. This would definitely seem more 'right' in terms of the true value of ratings.
Re: Why does it seem that the higher rated players
Posted:
Oct 29, 2005, 7:30 PM
unfortunately, while I agree with up that changing the rating system to reflect results of sets alone would result in more higher rated players being willing to play games against lower rated players, my feeling is that it really doesn't completely address the entire issue. Here's the unaddressed part in my mind: if ratings were based on sets alone and 3 different players were playing against each other (call them 1500, 1700, and 1900), player 1900 would seldom lose a set to player 1700 and almost never lose a set to player 1500 no matter how many sets were played. Player 1700 would seldom lose to player 1500. Player 1500 would lose almost every set s/he played. Granted that this is oversimplification. The problem to my mind is that players 1500 and 1700 receive no credit for improvement for their playing ability unless/until they start to win sets, regardless of the fact that when starting out may have a result of drawing 3/5 sets and after a while may be drawing 9/10 sets. This only translates into a slowing of the loss of points, not a gaining of them. The only way I see for a new rating system to work would be for it to start from scratch on a set based premise (everyone becomes a 1600 player until proven otherwise in set based play), because otherwise, the addition of a revised ratings calculation will do more to cement the current levels in place than to give a more accurate picture, IMO. Now the question becomes- how many established players are going to want to let their hard earned ratings under the current system be replaced by a brand new 1600 rating in a new system? I don't know. Also, how many players will want to play 10 games against an opponent who has a rating of 200 points above theirs, win 5 of those games and leave the table with an unchanged rating? Again, I don't know. I like the fact that the change talked about here would potentially keep the best players playing against opponents of all levels, but I dislike the thought that it might make pente unappealing to the majority of players at the site because they are able to see little change to their rating (except downward).
Retired from TB Pente, but still playing live games & exploring variants like D, poof and boat
Re: Why does it seem that the higher rated players
Posted:
Oct 30, 2005, 3:21 AM
I would feel much more comfy with a weighted rating system. i.e. winning as player 2 having a higher weighted value over winning as player 1, Under the same system losing as player 2 would be far less costly and an easier pill to swallow.
It always burns me when a 2000+ player can win 10 games straight vs a 1700 player and maybe gain 20 points, but then lose one player 2 game and BAMB! 30 points go down the tube. A 2000+ player losing a game as player 2 vs a 1700 is not that significant, and dropping something like 10 points would be fair in that situation. On the other hand a 2000+ player losing a game to a 1600 player as player 1 is rather significant, and would be fair in costing those 30 points.
I also think that if a 1200 player happens up and wins as player 1 vs a 1900 player it should gain the 1200 player more than it costs the 1900 player for losing as player 2. Something like a 30 point gain for the 1200 player and a 10 point loss for the 1900 player.
Perhaps a 10 point max lose limit for losing as player 2 should be set while allowing a low rated player to gain as much as 30 points from winning as player 1 vs a high rated player.
Re: Why does it seem that the higher rated players
Posted:
Oct 30, 2005, 5:58 AM
I think that the weighted system has a lot to do with it. If a higher rated player looses to a lower rated player, the higher player looses about 30 points. If a set is played then and the higher rated player beats the lower rated player, he may gain about 10 points. So the set does not necessarily cut it. There is a problem with the weighted system.
Re: Why does it seem that the higher rated players
Posted:
Oct 30, 2005, 6:18 AM
I actually find the whole question ironic in the sense that I spend a LOT of time on the site and continually open a room to play pente. I've left rooms open for hours at a time and unless someone with a high rating comes on, I never get a player. Yet, there are other rooms full of people with 'low' rankings who are watching others with low rankings and chatting for hours on end - not playing at all. Or, more often, low rated players are so busy playing the computer, they never meet a human opponent with the ability to think creatively and wonder why they're getting their *** kicked when they could beat ai3 so easily! lol! further, the only thing i typically here from low ranked players is 'I wish i could play that good' (as they sit and watch, not daring to play) and 'no thanks, you're too good for me. so for what it's worth peanut, I dont see your problem at all and in fact, see precisely the opposite to be true. so im not asked again, let me tell you how i learned to play pente the way I do now (for better or worse). I played richard and jayhawk for months on end and took tens of dozens of *** kickings. I rarely asked them what they were thinking, one cannot really explain that. I watched and lost and watched and lost and lost to infinity. and then I won.
Re: Why does it seem that the higher rated players
Posted:
Oct 30, 2005, 6:24 AM
and another thing, as far as playing those who are just slightly better than you as up mentioned - what do you propose richard should do? who should he be playing?
Huh punk? lol
I say thats bullshit - play people who beat you mercilessly and repeatedly and who dont make you think you're good by letting you win. a little patience, a few hundred games and the ability to not let your pride get the best of you and you'll be kicking anyone's *** in no time.
Posts:
542
Registered:
May 9, 2002
From:
Northeast USA
Re: Why does it seem that the higher rated players
Posted:
Oct 30, 2005, 7:09 AM
Well, I guess I've steered this discussion off onto a tangent, but I'm glad to see the ideas flowing again about this topic. And hey, 1 out of 4 people actually agree with me so far! Woohoo! Actually, I think Laurie agrees with me too, she just doesn't realize it yet...
Laurie, if you like math and are interested in how ratings are calculated here, it is described in the FAQ for this site (find that FAQ link on the main page in the "documentation" section). In the FAQ, click on General FAQ to see some entries about ratings. But, to make a long story short, your argument actually supports what I'm trying to say. When two players of different rating levels play a set under the CURRENT system, the higher rated player may indeed lose 30 points for losing the P2 game, then win back only 2 points for winning the P1 game, for a net loss of 28 points, even though this match overall was a draw. My point is that ratings should not even be recalculated at all after a half-a-set -- instead, the system should wait until both games of a set are played, then use the current formula to calculate the result for a set win or a set loss. For a set draw, there should simply be no change in either player's rating. Incidentally, Brainking already gets it right. They provide an option to play a match consisting of a single set, and if the set ends in a draw, both ratings remain unchanged.
Watsu, I can understand your concerns about possible negative "side-effects" of what happens to ratings over time with set based play. Obviously, it's not a good idea to do something that negatively impacts the playing experience for a majority of players. However, my opinion is that your logic of what would happen just doesn't hold up. The difference in ratings would already reflect just how often the underdog is expected to win a set. A small difference in rating means this might happen relatively frequently -- a large difference in ratings means this would almost never happen. (After all, players should be playing against other players with similar ratings if possible -- not just because of ratings, but because that's the most fun for all and is the best way to learn.) And the underdog is not being penalized for this. He loses only a small amount of points for dropping a set and would win a large amount for winning the set. This is essentially the same as what happens right now, except that it would have the added bonus of actually being accurate. Meaning, the P1 factor has been completely eliminated from the equation (well, of course, it's not part of the current equation at all even though it's a significant factor in who will win, which is the whole problem), thus allowing each adjustment in ratings to be based completely on the difference in skill level as reflected in the pre-match ratings. I believe that the math shows that there's no reason to be concerned. However, just in case there is still concern, I actually really like your idea of "restarting" everyone's rating to 1600 and continue forward with set-based play and set-based ratings. Of course, people are not going to want to give up their current rating. The answer might be (you know, because Dweebo clearly needs more work to do ) to create a new GAME TYPE called PENTE SET or SET BASED PENTE or whatever, that can actually be selected seperately from PENTE in the game room and thus would have a whole seperate rating system. One headache is that draws would have to be incorporated into the database and into the win/loss/draw records for the players. And of course, this could essentially DOUBLE the number of game types eventually! But this might be just the right compromise to get this thing started. I can't wait!
Richard, you are a really smart guy so I'm kind of surprised at what you came up with for a solution. I wish I had the energy to give a strong mathematical argument for why the weighted system does not work at all, but I guess I feel like it should just sort of be obvious. Instead of eliminating the P1 advantage variable from the mix, you are just trying to choose some numeric representation for it. How is this magic number to be chosen? For one thing, when you have a 1200 against a 1000, the P1 advantage does not mean as much as it does when you have a 2000 against an 1800. It does in terms of the game itself, but practically speaking the players are not skilled enough yet for it to make very much difference. And as for one player losing less points than the other player gains, that just doesn't make sense at all. It might be useful for you to check out the FAQ I talked about above and look at how the ratings are calculated. I have a feeling that you just haven't really concerned yourself with or have spent very much time thinking about this topic yet because even if you don't like my solution, I'm sure you could come up with a better solution than the one you've presented here.
Dweebo, I didn't mean to say that this topic has been ignored. I know you have a long priority list of things to think about and implement. I know that you have thought about this topic before as well. But, I urge you to consider watsu's suggestion of adding a new game type with "fresh" ratings to give set-based play a trial run which can always be scrapped if it doesn't work out.
And remember, if a player is not in the mood to play a set, they can always play unrated.
Posts:
542
Registered:
May 9, 2002
From:
Northeast USA
Re: Why does it seem that the higher rated players
Posted:
Oct 30, 2005, 7:13 AM
> and another thing, as far as playing those who are > just slightly better than you as up mentioned - what > do you propose richard should do? who should he be > playing? > > > Huh punk? lol
Re: Why does it seem that the higher rated players
Posted:
Oct 30, 2005, 5:49 PM
As a reply to Tragedy He is very correct when it comes to learning this game. I truly learned this game from playing Dufus/bugnone/evilone/Mike321s brother and his many other names Hundreds of games mostly practice. Not caring about the ratings but really learning the game. He would play alot of his new moves until either I would beat them or go home with my tail between my legs. Everyone needs a Mentor?? Someone that will help you with your game? But you have to want to get better by analyizing all the great players. Virag, Gary Barnes/progambler, Dmitriking, Scott Justice,( Alexander Nosovosky learned the game from Stepanov (misspelled sorry) Student was another pupil of his), I no most of you have never heard of these players but these players are the top 10 players of all time and they have all got mentors so find one. Work at your game by playing alot of better players unrated (learn from your mistakes) and then you are ready to step up on your ratings. I bet there isnt a player that wouldnt play you unrated If asked nicely so you could learn the game from them. Even Mike321 that never wants to play unrated played his brother 1000s of games to learn and practice all of there different moves. Everyone has someone that is better than them early that they can learn from. Unrated is not a dirty name its just practice. Theres not a Pro team on the planet that doesnt practice and analyze along time before they play that game on Sunday. Stop whining about top players not playing you rated until you have stepped into the years of analyzing, practice,touneys, and the many other sites they have all played in. Everyone starts someplace so find a MENTOR.
Re: Why does it seem that the higher rated players
Posted:
Oct 30, 2005, 7:42 PM
Up2ng>>I wish I had the energy to give a strong mathematical argument for why the weighted system does not work at all, but I guess I feel like it should just sort of be obvious. Instead of eliminating the P1 advantage variable from the mix, you are just trying to choose some numeric representation for it. How is this magic number to be chosen?
R3>> My suggestion was not definitive in nature, just subjectively suggestive. If u want definitive then one can be drawn up with a little research comparing equally matched players in their win\lose averages as P2 and as P1. I haven’t done this but it can be done with a db query defined based on games played vs an opponent’s rating being within say a 60 point zone tolerance. You may cut out say players under 1400 to remove the “I’m just learning“ factor. With the overall set of queried games you are just a fart away from determining what the over all win\lose % factors are for P1 and for P2. Those %s can be applied to determining a base weight factor for P1 and for P2, with the weighted adjustments being a numeric derivative, that when factored in, levels out the % differences to Zero. That weight factor can then be bell-curved based on ratings differentials. The overall effect would be to accurize ratings to reflect the real differences between winning or losing as P1 as apposed to winning or losing as P2.
Up2ng>>Richard, you are a really smart guy so I'm kind of surprised at what you came up with for a solution.
R3 says: Kiss ma buttakeeessssss! I am a freaken genius!
Posts:
45
Registered:
Feb 1, 2005
From:
Philadelphia
Age:
48
Re: Why does it seem that the higher rated players
Posted:
Oct 30, 2005, 9:33 PM
I think people sometimes get a little obsessed with ratings, and not just at DSG. I used to be a regular in the Hearts rooms at Yahoo where people routinely punt you if they feel your rating isn't worthy of them.
I agree, though, that playing sets for ratings is at least a partial solution. On occasion, I've taken seat 2 to begin play, lost the opening game, and then had my opponent run off without giving me a rematch. That's a little annoying. In those cases, the abandoned second game of the set could go down as a forfeit for the guy who abandons, and then the ratings could return to where they were prior to the game.
I also like the idea of weighting games played as p1 & p2 differently. It just makes sense since a win as p2 is a greater accomplishment whereas a win as p1 is kind of like holding serve.
)
) to create a new GAME TYPE called PENTE SET or SET BASED PENTE or whatever, that can actually be selected seperately from PENTE in the game room and thus would have a whole seperate rating system. One headache is that draws would have to be incorporated into the database and into the win/loss/draw records for the players. And of course, this could essentially DOUBLE the number of game types eventually! But this might be just the right compromise to get this thing started. I can't wait!