Proposal to improve Net Points scoring
30 Mar '17 21:18Russ has recently put forth a proposal to replace the current default scoring method to determine a clan champion, the net points scoring, with a new metric, clan rating.
Russ at the time strongly encouraged discussion on his proposal.
This would lead the community to assume that this is not necessarily the final product.
Accordingly, this has been discussed at length in several threads.
There is support for the proposal along with some discomfort and skepticism.
I had committed to submitting an alternate proposal, that being to improve the net score rather than replace it.
This proposal should not be taken as my personal rejection of Russ's ideas.
But I do feel an alternate solution is worth putting forward.
So here is my offering, starting with a brief summary of the issues.
Followed by some ideas.
So I will say exactly what Russ said.
Feel free to rip into it !! And have fun !!
Primary issues:
1 - problem with the scoring system. Meaning the way net points are awarded.
2 - perceived problems with player rating manipulation commonly called sandbagging.
3 - putting clan goals ahead of personal goals. In other words, players resigning clan games where the outcome has been decided.
4 - collusion amongst amongst several clans to generate points towards one clan.
Secondary issues:
1 - inactive players in clan challenges
2 - accusations of challenge dumping.
3 - others. Feel free to add some.
Proposed ideas:
1 - implement a separate rating for clan games for each player. Best starting point for each player would be the site rating
1A - further to (1) , use the same formula in the site FAQ but modify the K factor by lowering it to 8 from the 32 value in tournament play. This will reduce the rating variations from +/-32 to +/-8 per completed clan game. For 2 equally rated players, a win or a loss will result in a rating change of +/-4 rating points. This will greatly reduce the accusations of rating manipulation.
This coupled with proposal 1 will solve the sandbagging issues.
Once the player clan game rating is decoupled from the site rating, it will not be nearly as volitile as the tournament rating.
2 - Implement a new net point scoring system with 2 components as follows:
2.1 - use the net rating change as a score to contribute to a net score for the challenge. For example, if a player wins his game and his rating goes up 4 points, his clan will be awarded 4 points. His opponent losing the game will result in his clan getting -4 points. Rating changes from draws will factor in as well. They will likely be very small.
Every game will count and the challenge score will be cumulative. So this will address resigning games in decided challenges.
2.2 - the above comes with a flaw as it would discourage larger challenges. Clan leader may prefer 10 one player challenges over 1 ten player challenge. To address this create a bonus point system for challenge size. For example, 0 points for <5 games, x points for 5-10 games, y for > 10 games. Challenge draws award 0 bonus points. Final bonus point values can be worked out. And carefully so as to not just recreate the current scoring system.
This would result in a hybrid scoring system.
In this system, all clan scores are reset to 0 on Jan 1st of each year.
Which is not the case with the clan rating proposal.
3 - all clan games must go at least 5 moves to be set as rated games. This means any games not reaching 5 moves will not update the rating for each player. Additionally the games will not be assigned a W or L to the challenge game score. Abnormally and suspicious short games will not contribute to a challenge result. This will take care of inactive players. Additionally, any players tossing games would at least have to make some moves. Not a perfect solution but a deterrent.
4 - consider a rethink on when the points are rewarded. Specifically 2 ideas:
4.1 award the points on a game by game basis.
4.2 award the clan bonus points once the challenge is clinched.
The proposals above are a departure from the current protocol of awarding the points once the challenge is complete.
Perhaps a radical departure that may be hard to digest. But this would end complaints of resigning games in decided challenges as well as the accusations of year end challenge dumping.
5 - maintain a protocol of human intervention in the event of any improprieties by any clans. This is a requirement irregardless of whichever system this site moves forward with. Data correction may be easier in this system than with the clan rating system. Unless someone has figured out how to correct a clan rating.
What I am proposing is what I believe moonbus (and giannotti) and I were circling in correspondence in the original site ideas thread.
People were expecting individual player ratings for clan games.
I believe that a clan rating can be a useful tool. As information only.
I am acutely aware that the proposals 1-4 do not address collusion.
Which is why proposal 5 has been put forth.
Collusion is not a specific technique.
It will occur in any new set up if certain individuals want to circumvent whatever is in place.
This is my proposal.
Like I said, have fun with it.
Russ at the time strongly encouraged discussion on his proposal.
This would lead the community to assume that this is not necessarily the final product.
Accordingly, this has been discussed at length in several threads.
There is support for the proposal along with some discomfort and skepticism.
I had committed to submitting an alternate proposal, that being to improve the net score rather than replace it.
This proposal should not be taken as my personal rejection of Russ's ideas.
But I do feel an alternate solution is worth putting forward.
So here is my offering, starting with a brief summary of the issues.
Followed by some ideas.
So I will say exactly what Russ said.
Feel free to rip into it !! And have fun !!
Primary issues:
1 - problem with the scoring system. Meaning the way net points are awarded.
2 - perceived problems with player rating manipulation commonly called sandbagging.
3 - putting clan goals ahead of personal goals. In other words, players resigning clan games where the outcome has been decided.
4 - collusion amongst amongst several clans to generate points towards one clan.
Secondary issues:
1 - inactive players in clan challenges
2 - accusations of challenge dumping.
3 - others. Feel free to add some.
Proposed ideas:
1 - implement a separate rating for clan games for each player. Best starting point for each player would be the site rating
1A - further to (1) , use the same formula in the site FAQ but modify the K factor by lowering it to 8 from the 32 value in tournament play. This will reduce the rating variations from +/-32 to +/-8 per completed clan game. For 2 equally rated players, a win or a loss will result in a rating change of +/-4 rating points. This will greatly reduce the accusations of rating manipulation.
This coupled with proposal 1 will solve the sandbagging issues.
Once the player clan game rating is decoupled from the site rating, it will not be nearly as volitile as the tournament rating.
2 - Implement a new net point scoring system with 2 components as follows:
2.1 - use the net rating change as a score to contribute to a net score for the challenge. For example, if a player wins his game and his rating goes up 4 points, his clan will be awarded 4 points. His opponent losing the game will result in his clan getting -4 points. Rating changes from draws will factor in as well. They will likely be very small.
Every game will count and the challenge score will be cumulative. So this will address resigning games in decided challenges.
2.2 - the above comes with a flaw as it would discourage larger challenges. Clan leader may prefer 10 one player challenges over 1 ten player challenge. To address this create a bonus point system for challenge size. For example, 0 points for <5 games, x points for 5-10 games, y for > 10 games. Challenge draws award 0 bonus points. Final bonus point values can be worked out. And carefully so as to not just recreate the current scoring system.
This would result in a hybrid scoring system.
In this system, all clan scores are reset to 0 on Jan 1st of each year.
Which is not the case with the clan rating proposal.
3 - all clan games must go at least 5 moves to be set as rated games. This means any games not reaching 5 moves will not update the rating for each player. Additionally the games will not be assigned a W or L to the challenge game score. Abnormally and suspicious short games will not contribute to a challenge result. This will take care of inactive players. Additionally, any players tossing games would at least have to make some moves. Not a perfect solution but a deterrent.
4 - consider a rethink on when the points are rewarded. Specifically 2 ideas:
4.1 award the points on a game by game basis.
4.2 award the clan bonus points once the challenge is clinched.
The proposals above are a departure from the current protocol of awarding the points once the challenge is complete.
Perhaps a radical departure that may be hard to digest. But this would end complaints of resigning games in decided challenges as well as the accusations of year end challenge dumping.
5 - maintain a protocol of human intervention in the event of any improprieties by any clans. This is a requirement irregardless of whichever system this site moves forward with. Data correction may be easier in this system than with the clan rating system. Unless someone has figured out how to correct a clan rating.
What I am proposing is what I believe moonbus (and giannotti) and I were circling in correspondence in the original site ideas thread.
People were expecting individual player ratings for clan games.
I believe that a clan rating can be a useful tool. As information only.
I am acutely aware that the proposals 1-4 do not address collusion.
Which is why proposal 5 has been put forth.
Collusion is not a specific technique.
It will occur in any new set up if certain individuals want to circumvent whatever is in place.
This is my proposal.
Like I said, have fun with it.
Back to Top