Avoiding lottery effects in tournament rounds

classic Classic list List threaded Threaded
1 message Options
Reply | Threaded
Open this post in threaded view
|

Avoiding lottery effects in tournament rounds

grampajohn
Administrator
Dear colleagues

Some of you have noticed that some of the game parameters do not seem to be as variable as usual in the 2015 final round that's running right now. That is an accurate perception. We (Wolf and I mostly) had been discussing the issue of lottery effects for a while; in fact, a workshop paper by Wolf back in 2004 led to a major re-design of the SCM game when he showed that the outcomes of the previous year's games could be predicted by the outcome of a couple of lotteries early in each game. Given the way tournament scoring works in Power TAC, it is also possible for the overall outcome to be strongly influenced by "the luck of the draw" on values like bank interest, distribution charges, game length, and a few other parameters that are normally highly variable from game to game.

The agents running in the current tournament appear to be much better than what we have seen in past years, and the number of games we can run in a 4-5 day tournament with 11 participants is so small that statistical significance could easily be wiped out by lottery effects. So, given that teams are not allowed to change their agents once the round starts, we decided to narrow the ranges of a few parameters fairly dramatically in order to put us on solid ground when it comes time to announce winners in Istanbul. We purposely decided not to announce the decision until after the start of the round because we did not want agent teams to take advantage of the reduced game variability.

Once the tournament is over, we would welcome an extended discussion on this topic -- is it a good idea to reduce variability in tournament rounds? Which parameters should be constrained, and by how much? Should such decisions be announced in advance? Although I will not be at the workshop in Istanbul, I would hope that this topic will get some serious discussion and that someone will summarize the discussion on this list for the benefit of those of us who cannot be there.

Best of luck to all the teams. I will be following the results with considerable interest.

John