Prizes

Home > Diplomacy > Cogitations > Prizes

Cogitations on AI in the Game of Diplomacy: Prizes

Here is a brain dump of full details my current Cogitations about prizes I may offer. (Later Cogitations override earlier ones if contradictory.)

Contents

• 2008-07-24

2008-07-24

I would award specified cash prizes for the top bot in each of a specified set of DEMO environments, normally realms, according to the following rules:

1. Free automatic entry of any bot announced in DipAi that I can easily get to run satisfactorily and test in DEMO.
2. Evaluation shall be with respect to all other players when playing in the environment, either actually or estimated when the ranking, at least, is unlikely to be wrong.
3. The single champion in each specified prize environment shall be determined from the latest state of DEMO, exactly at the end of each calendar month, provided that the difference is significant enough to be unlikely to change if more evidence is accumulated. (A tie-break rule shall be used if necessary, normally by running further experiments until there is a difference.)
4. For each prize environment where one of my bots was champion I would add a sum of money to an accumulating prize for that environment.
5. For each prize environment where one of my bots was not champion I would pay any accumulated prize for that environment to the owner of the champion bot.
6. I may withdraw any offer of a prize without notice, explanation or any liability.
7. My decision on all aspects of winners and prizes shall be final.

Thus the prize in a given environment builds up as long as I have the champion bot there. To encourage competition, the more difficult it proves to beat the more I offer. I would probably mainly concentrate on the one environment where my best tuned bot performs best compared with the best of the rest (as determined by DEMO). But in the event that this were too difficult for current bots, simpler challenges might be defined to maintain interest. Or it may be of interest to encourage development in different areas, for instance ability to play well in non-Standard variants, or lightening play.

There is no cost to me unless and until I have such a bot, nor would I be adding undue (financial) encouragement to other developers to steam too far ahead of me. But if and when I do have such a bot I will want to encourage competition, and would be willing to pay something for that. Adding piecemeal to a jackpot would make it seem less daunting when paying a large prize. Prize payouts comparable to my existing Diplomacy hobby costs seems about right – maybe about £30 ($60) p.m. or £360 ($720) p.a. (for a high-end CPU/RAM upgrade of two PCs, say, every 3 years, plus 24/7 electricity). This would not multiply my costs too much, while multiplying the value of what I am spending on, by improving my enjoyment of the hobby. And such a bot may win me prizes from other members. And not being beaten, despite offering a prize to do so, increases kudos!