Sunday 31 May 2015

Pairing and rating multiplayer games

I spent part of today at the Canberra Bordgames Titles, both to test (and fix) the tournament management software, and also to get a feel for how these events run. I quite enjoyed the experience and in part it reminded a little of my early chess playing days when events were a little more informal (and signs like "Consider your opponents. Please shower" were more common).
For some of the multiplayer events a fairly simple pairing algorithm was used. The winners of each (4 player) game earned 4 match points (MP), second was worth 3, third 2 and last was 1. Victory Points (VP) were  Then the players were sorted by total MP (and total VP as the next ranking value)  and the top 4 were placed on Board 1, the next 4 on board 2 etc. To avoid byes if the field was not a multiple of 4, the bottom boards were 3 player games (so as to give everyone a game).
While this rough and ready pairing system is generally OK, there were of course a couple of questions about whether there was a better system. For example, in a close event, it seems better to be on Board 2 (with a higher chance of picking up 4 points) than on Board 1 in the final round (NB The top 4 finishers then qualified for a final playoff). Also there was no restriction on playing the same opponent(s), so in some tournaments the players at the top (and the bottom) remained fairly constant (and familiar).
Related to this is the question of ratings. For this event ratings were not used, meaning everyone started in a random order. If this event is repeatable, then there is the possibility that the results from this tournament could be used to produce some sort of ranking for the next. I wondered if something akin to the ELO system could be used to produce such a list. My guess is yes (and I have already thought about such a method), but what the k factors and Score Expected (Se) tables should be is something I haven't established as yet.
The broader question is whether such a ranking system would help or hinder such events. I think the majority of people who played over the weekend did so because they simply enjoyed playing the games (Settlers, Dominion, TTR, Carcassone and Rummikub) and not because of any ranking/rating glory. While there is no doubt that having a rating system has contributed to the increased popularity of chess, I am not sure that such a system is for everything.

No comments: