07 Jan 17
"2] in a clan challenge make every point a prisoner, say 10 v 10, winner gets 20 points loser gets 0 then bonus points for every game won, so even the losing clan can get 9 points, every game will mean some thing, again this will cut down sandbagging."
This is in effect a bulk-wins ranking system. It favors large clans which play short timeouts. This not a fair measure of success in the present field of solo-clans at one end of the spectrum and large clans at the other end of the spectrum, with some players members of multiple clans.
07 Jan 17
2 edits
Originally posted by mghrn55ELO is not the salient point. The salient point is this: if the clan's ranking is pegged to the clan members' net ratings changes (using whatever formula RHP is in fact using, or some tweaked variation thereof), then there is no longer any advantage to be gained by throwing games, because every thrown game lowers not only the individual player's rating but also automatically his clan's rank. This undercuts sandbagging -- not by preventing it, but by rendering it ineffective as a means of manipulating clan standings.
I may have missed some posts in here.
I am a little fuzzy I'm some areas.
Maybe I missed some posts in there.
1-Can you explain to how the ELO system works as you propose it ?
2-explain the proposal of a clan rating. Is this to mean that the clan rating floats as opposed to the player ratings ?
I need to be clarified.
Some of these ideas can work. ...[text shortened]... ings move under the ELO system.
My current frame of reference is the rating formula in the FAQ.
To work, clan members' ratings must be isolated from tournament ratings (to prevent people from manipulating their clan ratings by dumping tournament games). Ratings would have to be tracked in order to restore a previous standing in case disciplinary action were indicated.
The advantage of pegging clan rank to clan members' collective net ratings change is that it is neutral with respect to clan size, long or short time limits, whether a clan has high-rated players or low-rated players, and therefore levels the playing field. In short, what is assessed is whether a clan's members' ratings (collectively) go up or down, not the total number of wins.
Pegging clan ranking to clan members' collective net ratings change also undercuts collusion. This is how it works: if Clan A gives Clan B easy wins (for example by resigning games which might have been won or drawn had they been played out, or by timing out without a move), then Clan B will reap some unearned ratings increases, true -- but only the first time. This cannot be prevented by the rating system alone, which is why an adjudication committee should be convened to oversee such cases. However, if Clan B thinks they can keep on reaping unearned ratings increases by repeatedly bashing Clan A, then the returns get smaller and smaller because every time the feeder clan throws games and a challenge, their own clan ranking goes down and the increase for the beneficiary is reduced. At some point, the feeder clan's ratings would drop below the 200-point differential and no further challenges would be allowed. Whereas clans which play straight (against evenly rated clans) would pull away from the clan trying to benefit from a feeder clan. A system in which clan rank were pegged to members' ratings would be self-regulating over the long-haul because the best way to raise a clan's rank is by playing good chess against evenly ranked (or slightly higher ranked) clans; whereas repeatedly bashing a clan which is falling farther and farther behind (i.e., collusion) is a loser's strategy.
For those with less patience, an arbitrary limit of 2 same-clan-same-player challenges could be implemented to reduce the effect of collusion to a negligible factor.
Several one-off collusions by multiple feeder clans benefiting a single recipient clan would still have to be monitored and if necessary remediated. Dead players and no-move timeouts would still have to be monitored. Pegging clan rank to members' ratings alone will not prevent or correct these types of fraud. An adjudication committee could be charged with overseeing such cases.
Redressing the score from 2016 and punitive action for behavior widely seen to have been unsportsmanlike (whether or not specifically mentioned in RHP ToS) is a separate issue. I believe that is the issue which most rankles.
Originally posted by moonbusThanks for that.
ELO is not the salient point. The salient point is this: if the clan's ranking is pegged to the clan members' net ratings changes (using whatever formula RHP is in fact using, or some tweaked variation thereof), then there is no longer any advantage to be gained by throwing games, because every thrown game lowers not only the individual player's rating but a ...[text shortened]... entioned in RHP ToS) is a separate issue. I believe that is the issue which most rankles.
I also read some of your original posts.
You refer to a clan ranking. You mean average clan rating, correct ?
One concern here.
Clan member turnover.
Some members leave, others join.
For example, a 2000 rated player leaves a clan and is replaced by a 1000 rated player. An extreme case.
That will significantly alter the clan's average rating.
How do we measure the true ranking or rating of a clan ?
At the very minimum, we are talking about completely changing he nature of the beast.
And holes in the new system will soon appear.
Originally posted by moonbusSo ELO is not the promised land after all
ELO is not the salient point. The salient point is this: if the clan's ranking is pegged to the clan members' net ratings changes (using whatever formula RHP is in fact using, or some tweaked variation thereof), then there is no longer any advantage to be gained by throwing games, because every thrown game lowers not only the individual player's rating but a ...[text shortened]... entioned in RHP ToS) is a separate issue. I believe that is the issue which most rankles.
It still needs a committee to run it effectually and is not self regulating as promised
My worry is that as the person who single handedly destroyed the clan system last year knows more about how this system works and could do the same thing again
We need to come up with something better
It appears there are 4 things that are to be considered
1 Collusion
2 Sandbagging
3 Dead players
4 Points awarded for a win
1 The way around collusion would be that a clan cannot challenge the same clan in a 90 day period
2 The way to stop sandbagging would be the player who resigns there would be no rating change
I would like to see that one back dated to the 1st of January
3 The way to stop dead players would be to able participate in a challenge player must have moved in the last 14 days
4 Points should be awarded on the size of the challenge
So that a 10 man challenge would be 10 points for the win and 1 point for everygame won
If challenge finishes 11 - 9 winning clan would get 10 + 11 = 21 points losing clan 9
If a 5 man challenge ended 6 - 4 winning clan would get 5 + 6 = 11 points losing clan 4
Draws would be just games won
07 Jan 17
Originally posted by padger1: In my case the 90 day exclusion would mean I would hardly play any games. (I'm a one man clan band because nobody has ever asked to join). Playing clan games gives me much more control of my game load than tournaments. Important.
It appears there are 4 things that are to be considered
1 Collusion
2 Sandbagging
3 Dead players
4 Points awarded for a win
1 The way around collusion would be that a clan cannot challenge the same clan in a 90 day period
2 The way to stop sandbagging would be the player who resigns there would be no rating change
I would like to see that one back ...[text shortened]... ded 6 - 4 winning clan would get 5 + 6 = 11 points losing clan 4
Draws would be just games won
2: Sandbagging would also have to take into account rating manipulation by playing private games against lower ranked non subscribers and resigning early.
I've found this discussion very enlightening. Behaviour that's been going on that's escaped my attention entirely. Except for McTayto's protests I'd be oblivious to all.
Great site, great work and I look forward to any improvements. Just don't forget the little guys.
P
07 Jan 17
Originally posted by padgerPadger wake up and smell the roses
It appears there are 4 things that are to be considered
1 Collusion
2 Sandbagging
3 Dead players
4 Points awarded for a win
1 The way around collusion would be that a clan cannot challenge the same clan in a 90 day period
2 The way to stop sandbagging would be the player who resigns there would be no rating change
I would like to see that one back ...[text shortened]... ded 6 - 4 winning clan would get 5 + 6 = 11 points losing clan 4
Draws would be just games won
1. 90 days between challenges will mean virtually no clan games in progress
2. No rating change with a resignation means by resigning a lost game you save your rating
3. If I return from a fortnights holiday then I will have no pending clan games as I haven't moved in 14 days
4. Some merit here which surprises me with your other solutions being crap
07 Jan 17
2 edits
Originally posted by mghrn55No, it's not average rating that matters. That would simply mean that a clan with many high-rated players would be defacto winners. It's rating change (ratings going up or down) that matters. It's dastardly tricky to explain, but an example would make it clear, I think.
Thanks for that.
I also read some of your original posts.
You refer to a clan ranking. You mean average clan rating, correct ?
One concern here.
Clan member turnover.
Some members leave, others join.
For example, a 2000 rated player leaves a clan and is replaced by a 1000 rated player. An extreme case.
That will significantly alter the clan's ave ...[text shortened]... out completely changing he nature of the beast.
And holes in the new system will soon appear.
Clan A has two players, A1 rated 2000 and A2 rated 1900 at the beginning of the season. They win some, they lose some, they draw some. At the end of the season player A1's rating is 1950, A2's rating is 2000. The net rating change for each player is thus -50 for A1 and +100 for A2. The net rating change for the clan is therefore +50. THAT is the clan's ranking.
Clan B has four players: B1 is initially rated 1100, B2 is rated 1200, B3 is rated 1300, B4 is rated 1400. They win some, they lose some, they draw some. At the end of the season their ratings are as follows: B1 is rated 1250 (+150 net rating change), B2 is rated 1275 (+75 net rating change), B3 is rated 1200 (-100 net rating change), B4 is rated 1375 (net rating change -25). Now tally up the net changes to get the collective net rating for the whole clan: 150+75-100-25 = +100. THAT is Clan B's ranking.
Thus Clan B has scored better than Clan A, irrespective of individual players' high or low ratings or long or short time limits or clan size or players who leave or enter the clan in mid-season.
This would work only if initial ratings were accurate, and this requires that clan ratings be isolated from tournament and general ratings at RHP to undercut people entering tournaments and dumping games there in order to enter a clan with wildly depreciated ratings and then play at strength for the clan leading to a wildly inflated positive rating change. This would have to be monitored closely (e.g., by an adjudication committee).
It would be computationally complex to implement and maybe this is important to Russ.
An alternative metric to net ratings change would be win ratio (percent wins). This would be computationally easier (for Russ) to implement, and also neutral with respect to clan size, high or low rated players, time limits, and players who enter or leave a clan in mid-season.
Yes, it would fundamentally change the system. This is why some people here (myself included) have suggested that several metrics be used and that the title "best clan" be dropped and replaced by multiple alternative 'best' categories such as "most improved ratings clan," "highest win ratio clan," "most daring attacking clan," etc. I honestly don't see any way to award any meaningful title to a single clan in such a diverse field where there are solo-clans, large clans, clans with high-rated players and clans with low-rated players, clans which play long and short time limits, and clans with no fixed membership (players rotating in and out and with multiple allegiances) etc.
Bulk wins (points awarded for wins) unfairly favours large clans which play many games at short time limits; that would be a fair metric only if all clans played the same number of games over the course of the season.
Think of it like this: compare the number of medals won at the Olympics by the USA compared with Switzerland. The USA routinely hauls in scores of medals at every Olympic games; Switzerland can be proud if it brings home 3. Does that mean the USA is the best team there? Not necessarily, especially if you consider that the Swiss team might have pasted the Americans by winning gold, silver, and bronze in some specific event (lets say, track cycling) in which both participated. Give small countries (and small clans) credit where credit is due; simply looking at total medals is a very biased view of the qualities on display.
07 Jan 17
1 edit
Originally posted by padgerPoint four is equivalent to "bulk wins," which favours large clans which play many games at short time limits. Not a level playing field.
It appears there are 4 things that are to be considered
1 Collusion
2 Sandbagging
3 Dead players
4 Points awarded for a win
1 The way around collusion would be that a clan cannot challenge the same clan in a 90 day period
2 The way to stop sandbagging would be the player who resigns there would be no rating change
I would like to see that one back ...[text shortened]... ded 6 - 4 winning clan would get 5 + 6 = 11 points losing clan 4
Draws would be just games won
EDIT: Point two "The way to stop sandbagging would be the player who resigns there would be no rating change," makes no sense. If resigning a game does not drop a player's rating, what would? Only checkmate??
Originally posted by Mctayto1 There are 180 clans on this site I do not see this as a problem
Padger wake up and smell the roses
1. 90 days between challenges will mean virtually no clan games in progress
2. No rating change with a resignation means by resigning a lost game you save your rating
3. If I return from a fortnights holiday then I will have no pending clan games as I haven't moved in 14 days
4. Some merit here which surprises me with your other solutions being crap
2 Exactly .It would stop your cheating ways dead in your tracks
3 you will still have all the games you had before you went on holiday .Its just that you could not have any new clan games until you were back for 14 days
4 lovely
07 Jan 17
Originally posted by The PostmanThanks for joining the thread. It is good to hear from someone other than 'the usual suspects.'
1: In my case the 90 day exclusion would mean I would hardly play any games. (I'm a one man clan band because nobody has ever asked to join). Playing clan games gives me much more control of my game load than tournaments. Important.
2: Sandbagging would also have to take into account rating manipulation by playing private games against lower ranked no ...[text shortened]... t site, great work and I look forward to any improvements. Just don't forget the little guys.
P
Point 2: This is why it is essential to isolate clan individuals' ratings from private games and tournament games.
Originally posted by moonbusThis would encourage clans to grow not stay as a one man band
Point four is equivalent to "bulk wins," which favours large clans which play many games at short time limits. Not a level playing field.
EDIT: Point two "The way to stop sandbagging would be the player who resigns there would be no rating change," makes no sense. If resigning a game does not drop a player's rating, what would? Only checkmate??
Originally posted by padgerThat is the most complete solution so far proffered.
It appears there are 4 things that are to be considered
1 Collusion
2 Sandbagging
3 Dead players
4 Points awarded for a win
1 The way around collusion would be that a clan cannot challenge the same clan in a 90 day period
2 The way to stop sandbagging would be the player who resigns there would be no rating change
I would like to see that one back ...[text shortened]... ded 6 - 4 winning clan would get 5 + 6 = 11 points losing clan 4
Draws would be just games won
Add to this the committee to evaluate unethical play and that should fix the system.
Well done Padger !!