Page 1 of 1

Possible Ranking changes?

PostPosted: Mon Feb 25, 2013 8:40 pm
by Louis
I was wondering if I might offer a suggestion to the way that the ranking points system is worked.

This idea is based on playing a few games of 3 players.

We found that often 2 people would be very close at the end of the game (e.g 64 and 60) while the 3rd player is down on 46 for example.

The issue is that the second placed player is not rewarded at all for being so close to winning. Would it be possible to have a system which involves a bonus to RP based on how close you are to 1st place (and indeed possibly how far ahead of 3rd place)

The bonus need not be huge, say 5RP if you are within 5%, dropping to 2/3RP if you are within 10%. I feel that this would reward strong play over a longer time-frame and reward a player for coming very close to victory.

I would be interested to hear your thoughts!

Re: Possible Ranking changes?

PostPosted: Thu Feb 28, 2013 9:59 am
by ChrisGibbs
Hi Louis, thank you for the suggestion. At the moment, the ranking algorithm is designed to reduce to standard Elo in two of the most common situations: where all players have played an equal number of games and/or when all players have played more than Gexp (current value = 25) games.

At it's core the Elo rating system is a "zero sum" algorithm which means that the net change in RP across all players is zero. Introducing just a bonus to other players would break this condition, and over time could lead to a situation where the majority of players had over 1500 RP. This would then makes it harder to represent a player's skill, as in general 1500 RP should be taken to be the skill level of an average player.

To preserve the zero sum condition, if say the 2nd player was awarded 5 RP for being close to 1st place, then other players would have to be "punished" by a total of 5 RP - this then leads to a tricky situation of working out which players to punish and by how much. This would be especially difficult in a 5 player game where e.g. the final scores were 50 VP, 49 VP, 49 VP (less cash), 48 VP, 48 VP (less cash) or similar - 2nd to 5th places would have to be awarded a bonus and then the 1st player would receive a hefty penalty that could actually result in a reduction in RP.

An alternate system could involve awarding RP on a sliding scale for each player according to the number of VPs amassed during the game, thus making each player work hard for each VP. However, this could change the game substantially in terms of strategy, so I'd be hesistant to switch to that approach for now.

Re: Possible Ranking changes?

PostPosted: Thu Mar 07, 2013 4:47 pm
by Louis
Thanks for replying!

The point you made about representing player skill via the process of a zero sum algorithm may not be the fairest way. This system basically says the person who finished last loses the same ranking regardless of whether they are 1VP or 20VPs behind.

I understand that a change may lead to an uneven playing field but arguably if you want a far representation of skill, those who fight close to the leaders should be considered to be more competent players.

Also, there may be as much chance of people getting to average above 1500 as there is that they average below (if people consistently fall short of the game winner). Also, there is always the danger that people play 1 game, loose and then never play again. Their influence is limited but does reflect on the players who play a lot.

The last part about making players work hard for every VP, isn't that the point of the game? I try to get every VP possible and in the games I have played in, the last few turns involved building as much as possible to squeeze extra VPs out!

I don't have a solution to offer which isn't much use to be honest but its good to debate and look into alternatives!

Re: Possible Ranking changes?

PostPosted: Fri Mar 08, 2013 10:39 am
by ChrisGibbs
Yeah it is an interesting debate - you are right about players having to work hard for VP, but again isn't that part of the fun, to come in 1st place by narrowly defeating everyone else? If my alternate system suggested above of awarding RP based on the VPs was used, then close games would result in virtually no change in RP, in which case what would be the point of playing?

One other point of comparison is the several different ranking systems used for chess (including Elo and Glicko). In all of these, there are no "bonus" points for only narrowly being defeated by an opponent. Granted, chess doesn't have a VP score system, but an analogy could perhaps be created with a "narrow" victory being saying a king, rook and bishop getting a checkmate against a king and a rook.