Path: csiph.com!x330-a1.tempe.blueboxinc.net!usenet.pasdenom.info!goblin2!goblin.stu.neva.ru!newsfeed1.swip.net!uio.no!ntnu.no!not-for-mail From: Bent C Dalager Newsgroups: comp.lang.java.programmer Subject: Re: Basic prisoner's dilemma? Date: Mon, 18 Apr 2011 13:38:23 +0000 (UTC) Organization: Norwegian university of science and technology Lines: 31 Message-ID: References: NNTP-Posting-Host: decibel.pvv.ntnu.no X-Trace: orkan.itea.ntnu.no 1303133903 20628 129.241.210.179 (18 Apr 2011 13:38:23 GMT) X-Complaints-To: usenet@ntnu.no NNTP-Posting-Date: Mon, 18 Apr 2011 13:38:23 +0000 (UTC) User-Agent: slrn/0.9.8.1pl1 (Debian) Xref: x330-a1.tempe.blueboxinc.net comp.lang.java.programmer:3108 On 2011-04-18, theglazeb wrote: > And perhaps try to avoid the "death spiral" from occurring when two > players use the tit for tat strategy? If both players use tit for tat then the death spiral doesn't ever occur because both players start out by cooperating, and both players continue cooperating. The death spiral only occurs when one player deviates from tit for tat for whatever reason and defects, while the other player continues a straight tit for tat strategy, and the first player then also starts adopting either tit for tat or always defect (which is effectively the same thing in this case). The obvious alternative to tit for tat in this situation is to sometimes cooperate in spite of tit for tat dictating defect: either by keeping track of the play history and doing periodic cooperates; or alternatively by randomly cooperating (say, with 5% chance anytime you would normally defect) if you don't want to keep track of state. If memory serves, I think the general conclusion is that tit for tat is always superior to trying to be clever, unless you have some information about your opponent's algorithm (personality) that you can exploit to your advantage. Cheers, Bent D -- Bent Dalager - bcd@pvv.org - http://www.pvv.org/~bcd powered by emacs