Posts:
542
Registered:
May 9, 2002
From:
Northeast USA
Re: Pente variant proposal- ratings point gambling using a doubling cube
Posted:
Jun 30, 2008, 7:20 PM
Hey guys, I apologize in advance that I haven't had time to carefully read this entire thread but I have a couple of comments anyway.
Watsu, I kind of like the idea of a doubling cube but details would certainly need to be ironed out. It looks like you've done that a bit already but I haven't really read it all In general, if a pente match was set up similar to backgammon, where the first one to win 3 games wins the match, for example, then the doubling cube would be interesting, although not perfect since the P1 advantage would mess this up a bit. If it were done to just double the affect of the ratings changes for the current game, that also seems interesting. For example, if the rating formula is examined, there is a constant K = 32 which is the most points that can be gained or lost by an established player. Perhaps the cube would double this to 64? Neat idea, but I can see this being abused with ratings fraud efforts, etc.
As for richard's idea of a weighted rating system, without carefully reading this latest thread but remembering the suggestion several times previously, I will say that the idea is not a terrible one, but at best it is a poor one. Our current rating system has many flaws, but in broad strokes it does a reasonable job of determining relative skill levels of players, especially in the "middle" score ranges. It is unclear whether a weighted rating system would end up making these ratings more "accurate" or not, and it could certainly end up making things worse. It does not solve the core problems at all and is more of a band-aid than a solution and not a very good one at that. There are several reasons for this but I don't have time to elaborate now and I believe I've already done so in other threads.
The correct solution is to play pente in sets, with a split set being a "tie" which does not change either player's rating. This solves the inherant problem of the game of pente and it's corresponding problem with the rating system. Note that I don't say it improves the problem -- it completely solves it. This is so clear that any other suggestion isn't just a difference in opinion, it is actually an incorrect solution deserving of no partial credit with respect to this correct solution. Players often have to become masters of the game before this becomes so clear to them which is why the majority of players don't understand how I can be so adament about it and why I'm truely surprised that a master like richard has not grasped this concept yet when just about all other top players get it and many of them pretty much stop playing because of it.
Posts:
2,241
Registered:
Mar 4, 2007
From:
San Francisco
Age:
45 Home page
Re: Pente variant proposal- ratings point gambling using a doubling cube
Posted:
Jun 30, 2008, 9:58 PM
mmmm,..doubling cube at 64 points can be abused yea,..perhaps limit this option to players of 1700+ or 1900+ ext.. but i dunno.
i actually think that both up2ng and rich are right. a black win does deserve more points. and a tied set should not transfer any points. so both should be done. a won set transfers points in the formula rich suggests. and a tied set transfers none as up2ng suggests. in my opinion this is the correct solution.
and but,..maybe as an addition, if both win in the set as white, then no point transfer,.. but if both win in the set as black then a small bonus to both players of 1-3 points each depending on rank.
whatcha think?
~Z
Scire hostis animum - Intelligere ludum - Nosce te ipsum - Prima moventur conciliat - Nolite errare
Re: Pente variant proposal- ratings point gambling using a doubling cube
Posted:
Jul 1, 2008, 4:21 AM
Hey Up, As I understand the K factor from chess ratings (which may be different from what you mention, but chances are pretty good it's similar) it is a variable which adjusts how quickly or slowly ratings can change.
"It is a number which usually lies between 10 and 40. The higher the k-factor, the faster a rating changes; the lower the k-factor, the slower a rating changes."
"FIDE, the international chess federation, uses the following k-factors:-
25 until a player has completed enough events to have played at least 30 games; 15 for a player rated less than 2400; and 10 once a player has reached 2400 and played 30 games. Thereafter it remains permanently at 10."
I would say allowing a slightly higher maximum than 32 (say 40, for instance) would probably work out okay. In other words, the doubling cube would max out in terms of benefits for the lower rated player at 40 points maximum per game, but would help improve the ratings a higher rated player could receive with 40 still being a maximum value, of course. In practice a significant difference (say 400 rating points) would make it so that the higher rated player wouldn't be likely to come close to the maximum value, so they would end up getting a full double (or multiple doubles if that transpired).
Hope that makes sense... my sense is that anyone who would already be getting a 32 point gain would be unlikely to be in a situation where a double was made and accepted in any case. Raising the bar to 40 would not be (IMO) a significant enough change to merit a rating fraud effect, but would allow a 20 point gain to double into 40 or a ten point gain to be redoubled to 40, etc.
Since I don't have the Pente rating formula memorized or on hand I can't comment much more specifically than that.
Possibly, the K value would have to be lowered after a player reached a certain level- chess uses 2400 as that level, as quoted above. Since we have no 2400 level Pente players, that wouldn't be an immediate concern, but might eventually come into play if the cube works as hoped.
Interestingly, Go has a K range of 116 at rating 100 and 10 at rating 2700 for a much wider spread than chess. Here's a partial example (from this page- http://gemma.ujf.cas.cz/~cieply/GO/gor.html ) for the different K values Go uses-
Posts:
542
Registered:
May 9, 2002
From:
Northeast USA
Re: Pente variant proposal- ratings point gambling using a doubling cube
Posted:
Jul 3, 2008, 10:28 PM
watsu, I couldn't tell if the "k factor" you referred to is really the same as what K is in the ratings formula. I think you have the idea, but you might want to take a look at the formula, which is buried not too deep somewhere in this site's FAQ (which is listed down in the Documentation section). Or, if you are able to find my old thread where I break down the formula in detail that might be your best bet -- of course I forget what the thread is called exactly. But anyway, K is an arbitrary constant which is currently set at 32 points and it's significance is that an established player cannot gain or lose more than 32 points in a game.
zoeyk, unfortunately you have absolutely no idea what's going on. You really should find and study the current rating formulas and find and read my old threads that explain them. What you said here makes absolutely no sense...
> mmmm,..doubling cube at 64 points can be abused > yea,..perhaps limit this option to players of 1700+ > or 1900+ ext.. but i dunno.
Terrible suggestion -- this somehow implies that higher rated players are more ethical and trustworthy which is definitely not the case.
> > i actually think that both up2ng and rich are right.
Richard's solution and my solution are mutually exclusive. Without even going into which solution is "right", they really cannot both exist together -- if you think about it for a bit you should be able to realise that.
> a black win does deserve more points.
Why? (Hint, there is no correct answer -- your conjecture is flawed.)
> and a tied set > should not transfer any points.
I agree.
> so both should be done.
What?? This makes no sense. You cannot award ratings based on a game AND award ratings based on a set. What are you thinking??
> a won set transfers points in the formula rich > suggests.
What!!???? Are you high? Richard's whole idea is based on transferring points for GAMES in a way that evens out the P1 advantage. It makes NO sense to award points for a set in this way -- it's not even possible really.
> and a tied set transfers none as up2ng suggests.
I still agree with this.
> in my opinion this is the correct solution. > > > and but,..maybe as an addition, if both win in the > set as white, then no point transfer,.. but if both > win in the set as black then a small bonus to both > players of 1-3 points each depending on rank. > > whatcha think? > > ~Z
This last part makes so little sense I can't even believe it. I don't have time to point out all the reasons why. zoeyk, please research the issue a little more before commenting further.
in the FAQs for the rating formula we find this: 1
r1 + K * ( w - ( ------------------------ ) )
1 + 10 ^ ((r2-r1)/400))
looks like we are dealing with the same beast, or a close variant of it. Those K value numbers were based on information for Elo system ratings for different games, so I think it would be the same K.
Message was edited by: watsu at Jul 3, 2008 9:36 PM
Retired from TB Pente, but still playing live games & exploring variants like D, poof and boat
Posts:
2,241
Registered:
Mar 4, 2007
From:
San Francisco
Age:
45 Home page
Re: Pente variant proposal- ratings point gambling using a doubling cube
Posted:
Jul 4, 2008, 9:50 PM
well,.. im goina comment any ways. and no i didn't research the formulas further.
again your right that a set tied should be a zero point transfer.
the question i have is,.. for when a set is won,..should that formula be changed? or remain the same? if changed technically i believe it could be changed to equal close to what rich was proposing, although possibly it would be difficult to script. all this would do, is when you win the set you get more points then normal i'll go with that lol.
but perhaps with a tied set zero point system it wouldn't be needed to get extra points for winning with black. since as long as you win your white games and the set ties, your black game loss wont affect anything.
my main concern in the ratings system is the fact the black loss leaves me wide open at any point for a down ward spiral. and that it kinda restricts me from wanting to play rated with any class rank.
by eliminating this weakness,.it can remove boundaries of who i'm willing to play or not play rated threw out the 6 class's of rank. white,grey,green,blue,yellow,red.
to be honest IMO. it doesn't matter what formula you use at all. if the set ties with no point transfer, and your white is invincible, then you can only go up in points.
and in my post about combining you and richs idea, although i see points of logic in both sides, it was more meant in humor, because i know you two would never budge from your stand points on the issue, and to imply compromise was funny to me
but what i find funnier then that is;..we spend all this energy debating a point system, and in the end i have a feeling dweebo has no intention of changing the rating system regardless of our findings. so, may the wheels keep spinning.
~Z
Scire hostis animum - Intelligere ludum - Nosce te ipsum - Prima moventur conciliat - Nolite errare
Posts:
542
Registered:
May 9, 2002
From:
Northeast USA
Re: Pente variant proposal- ratings point gambling using a doubling cube
Posted:
Jul 7, 2008, 6:06 PM
Hey zoeyk, I think we're pretty much on the same page now. If you think about it, the grief you're having about the rating system and what you feel should be "fair" really is not very compatible with richard's solution -- especially if you see the value of playing with set-based ratings. The two ideas just don't overlap at all.
My solution would completely solve exactly the angst you are feeling about losing massive amounts of points as P2. Under my system, you gain points if you win the set, your rating is unchanged if you tie the set and you lose points if you lose the set. There would also not be the issue I think I saw mike bring up a while ago about people intentionally bailing after losing the first half of the set for the following reason: Let's say you are playing against someone rated well below you. Now, similar to what currently happens, let's say you have a chance to gain 4 points but stand to lose 28 points -- there's nothing wrong with this now!! Let's say you lose the first half of the set as P2 (or even P1 for that matter). Well, now you aren't going to gain your 4 points, but it still is very much in your interest to complete the set because if you bail, you will give up another 28 points (due to the familiar Force Resignation dialog) but if you manage to tie the set you net 0 (basically 28 points better than if you had bailed out). But zoeyk, think about this -- if you actually lose a SET against a player ranked well below you, then you deserve to lose that many points -- but you are protected against the "expected" result of losing as P2.
Think about it -- it is the perfect solution for players of all rating levels and would certainly get the top players to play more rated games against players of all levels.
Posts:
2,241
Registered:
Mar 4, 2007
From:
San Francisco
Age:
45 Home page
Re: Pente variant proposal- ratings point gambling using a doubling cube
Posted:
Jul 7, 2008, 11:41 PM
well yes of course. i already realize these things your saying. its been talked about time and time again. and i agreed with it a long time back too.
but the problem....Dweebo doesn't seem to want to change the rating system.
either because he doesn't have time, or because he doesn't believe in it. or because hes waiting for us all to come to a common agreeance, or maybe hes putting it off until the wheel squeaks loud enough before greasing it.
So we are wasting our time to dream about it maybe.
but i will say,..Dweebo,..perhaps consider it,..it might bring some old timers back to play more. it could help the site come alive again.
perhaps Dean you could set up a thread for the purpose of voting on it as a petition type of thing to count how many players support the set based system.
Posts:
542
Registered:
May 9, 2002
From:
Northeast USA
Re: Pente variant proposal- ratings point gambling using a doubling cube
Posted:
Jul 9, 2008, 8:47 PM
zoeyk, you're kind of losing track of what the thread was about. This discussion really doesn't have to do with what Dweebo feels like implementing as a rating system. We're just discussing different ideas that would potentially affect or change the rating system if they were in fact implemented. It started out with an interesting idea watsu had about players actually having an ability to affect ratings during a match -- sort of betting your opponent that you are in a winning position, risking losing more points than normal if you are wrong and end up losing. While describing his idea, he already recognized that it really shouldn't be an option that is included directly into our current Pente ratings because that wouldn't be fair to other players that do not use this feature but perhaps it could be a seperately tracked game type -- then he made a comment that it probably wouldn't make for as "accurate" a rating system as simply playing sets, but might just be a fun seperate option to add anyway. At that point richard chimed in and took exception to watsu's belief that set based play yields the most accurate rating system and made some arguments to support the idea of a weighted system for individual games instead. Of course, seeing this, I just wanted to quickly argue the other side without going into too much detail since I've gone into great detail in other threads about how mathematically the weighted system just doesn't make sense. Interestingly, his answer to your question earlier in this thread basically disproved his own theory and practically makes my point for me to those who will look at it and just think about it for a couple of minutes. But anyway, there's no reason to just choke off the conversation just because "Dweebo doesn't seem to want to change the rating system". First, you don't necessarily know that. And really, it is sort of irrelavent -- people should use these forums to discuss issues related to pente, even if none of it will ever be implemented.
Posts:
542
Registered:
May 9, 2002
From:
Northeast USA
Re: Pente variant proposal- ratings point gambling using a doubling cube
Posted:
Jul 10, 2008, 3:16 AM
Sorry if i offended you zoeyk. I guess I was in an argumentative mood. It seems like we pretty much agree on this stuff anyway, so what were we arguing about!? Oh well.
Posts:
2,241
Registered:
Mar 4, 2007
From:
San Francisco
Age:
45 Home page
Re: Pente variant proposal- ratings point gambling using a doubling cube
Posted:
Jul 10, 2008, 6:04 PM
Tis all good up2. (i was tempted to retort for fun, but held back) well we can always debate about politics or religion to satisfy the thirst for a good argument, hehehe j/k I'll say that out of all the players on this site, your one of the very few that i wouldn't want to get into a thread war with. Not to many can put me in my place with in a couple pages of posts, but your one of them i believe. seeing as how long winded you can be when you get going, (your 5,000 word posts make me nervious) oh and of course its always done with well thought out logic. that was a compliment.
~Z
Scire hostis animum - Intelligere ludum - Nosce te ipsum - Prima moventur conciliat - Nolite errare
In general, if a pente match was set up similar to backgammon, where the first one to win 3 games wins the match, for example, then the doubling cube would be interesting, although not perfect since the P1 advantage would mess this up a bit. If it were done to just double the affect of the ratings changes for the current game, that also seems interesting. For example, if the rating formula is examined, there is a constant K = 32 which is the most points that can be gained or lost by an established player. Perhaps the cube would double this to 64? Neat idea, but I can see this being abused with ratings fraud efforts, etc.