Number of turns?

[mgodfrey]

The introductory material suggests we test our bots over 1000 turns, while the View Game page says that only the first 200 turns will count. Which is the right number of turns per tournament game? Will bots be able to parse their input to determine how many turns are left in the game?

[Evgenius]

[gkisel]

I opened an issue related to this in the issue tracker the other day - http://code.google.com/p/ai-contest/issues/detail?id=74

[DanielVonFange]

I'd recommend the number of turns officially be fixed at 200.

[lloydgoodall]

I really hope in the final torny the limit is above 200.

With my current bot, when fighting the DualBot i lose in 200 turns but i win if i let the battle finish, at about 270 turns.

So even tho "most" are settled in 200 turns they defiantly are not settled everything.

So, please make the limit bigger, at least for the torny at the end. if not the current rankings..

Thanks, Lloyd

[Janzert]

I agree that I would like to see a longer maximum game length. I wonder though if it would help alleviate the short game length problems if a winner is only declared at the cut off if one side has both the most ships and the highest total growth rate?

Janzert

[lloydgoodall]

hmm, maybe you could make it a draw, if the winner is very clear or something. then when u get down to finals let it go into "over time" more like soccer or other sports?

[Vector_Matt]

I am also in favor of longer matches. Perhaps the max limit of 1000 as in the starter package? (or maybe 500 if 1000 is too many)

Or perhaps a 200 cutoff, only if one side has more ships AND a higher growth.


I'm more used to Galcon Fusion than plain Galcon, but comebacks do happen with moderate frequency, and I think we should allow for that in the matches.

[RebelXT]

another vote for longer games (if not 500, maybe 300 is a good compromise?). End game optimization techniques have to kick in way too early if game has to end at 200 moves.

[xmarkd400x]

I also vote for longer games. Most of my games are not completed by the 200 turn mark, although there is usually a clear favorite. Why not let the games run their course? Is there a computing bottleneck on the backend?