Author: Robert Hyatt
Date: 20:34:42 12/13/03
Go up one level in this thread
On December 13, 2003 at 13:48:29, Omid David Tabibi wrote: >On December 13, 2003 at 13:23:14, Robert Hyatt wrote: > >>On December 13, 2003 at 12:38:27, Omid David Tabibi wrote: >> >>>On December 13, 2003 at 12:18:43, Sune Fischer wrote: >>> >>>>>A CCT style tournament can never turn into an official event, since you cannot >>>>>prevent any kind of cheating. I can run 5 engines on different computers, see >>>>>which analysis I like, and then force my engine to play that move by feeding the >>>>>move via a file it checks once a second. How are you going to prevent that? By >>>>>looking at the analysis I output?! I can force my engine to print a spurious PV >>>>>starting with the move I want it to play... >>>> >>>>You can do exactly the same in WCCC. >>>> >>>>Here is one way, you do a ssh to some remote 'super' computer not monitored by >>>>any officials. >>>>Then you get your cheating partner sitting at home to use multiple engines to >>>>decide on the move. >>>>He then feeds the best move into your proxy which then makes up some bogus node >>>>count and PV and sends it back. >>>> >>>>You can even let it show analysis on the fly, it just has to have the right fail >>>>high at the time when the move is sent. It's a nobrainer to fake. >>> >>>It's all true. But the only person running remote in Graz was Vincent. And we >>>know enough about him and what he does to be confident that he is running on >>>what he says he is running. I ensure you that if an unknown person shows up >>>running on some unknown remote machine, everyone will question that and will ask >>>for verification. >> >>See my other post. I won't run remote. I _will_ cheat. Because you simply >>can't stop it without the expense I mentioned. Want to provide a wide-band >>frequency scanner to detect RF? Won't catch a burst transmission. Nor an >>IR transmission. Nor a transmission into the room over building power (do >>a web search for electrical power line modem). Etc... > > >See my other post. It is far easier to cheat online (needs about half an hour of >coding to enable a full range of cheatings), than to use RF or IR. No it isn't. With the rule to kibitz analysis, it is much more complex to make your program play a strange move, and keep kibitzes consistent. And then there is the timing as this is all done in real-time with lots of people watching. Kibitz A and play B and someone will notice. Move at an odd time point and someone will notice. It is _not_ that easy. It can be done, yes, but cheating can be done, period. It's certainly happened at ICCA events, for example. > > > > >> >>> >>> >>>> >>>>>The only reason why CCT tournaments are popular is that the stakes are not high. >>>> >>>>You may be on to something here, we are all so scared when the stakes are high. >>>>:-) >>>> >>>>Nah, I think the answer is more straight forward: it's because it's so >>>>accessible to everyone while the ICGA event is so unaccessible to most. >>>> >>>>>Most programmers join CCT only to test their engine against others (speaking for >>>>>myself, I will enter a totally experimental and untested version of Falcon). >>>> >>>>I will enter whatever version I think is the strongest, probably it will be >>>>experiment and somewhat untested, all my versions are. >>>>I would have done the same had I been at the WCCC. >>> >>>I wouldn't. In Graz I used a version which was tested enough (tens of long time >>>control matches against top programs), and was proven to be stable. During the >>>whole tournament I didn't encounter even one bug in Falcon. I learned quite a >>>lot about its weaknesses, but no programming bugs. >>> >>> >>>> >>>>I assume you also won't be joining with something you know for certain to be >>>>weaker 'just for the testing'. >>> >>>Untested means I don't know. It could be stronger or weaker. >>> >> >>I _always_ enter with what I believe is the strongest version. "believe" >>!= absolutely certain. But I do that at WCCC/WMCCC events as well. I >>treat them no differently from that perspective. >> >> >>> >>>> >>>>>But >>>>>when you give an official title to the winner, expect many (if not most) >>>>>participants to cheat in various degrees starting from "move now" to playing all >>>>>the moves as dictated by the operator. >>>> >>>>So you expect most people at WCCC to be cheaters given the chance? >>> >>>The stakes in WCCC are so high that you cannot rule out that option. If you >>>don't do any drug tests in 100 meter sprint in the olympics, how many contenders >>>will use drugs in your opinion? >> >>Not me. I can't speak for others. But I notice they still get away with >>it. Testers are human. They can be bought just as easily as the competitors >>can. >> >>> >>> >>>> >>>>And you think showing up in person made cheating impossible for everyone there? >>> >>>Makes it very hard to cheat, and almost impossible to fake a whole game. >>> >>> >>>> >>>>>The physical presence of the programmers (or operators) is inevitable for an >>>>>official event, especially one giving the title "World Computer Chess Champion" >>>>>to the winner. >>>> >>>>It doesn't prevent cheating at all, nothing can prevent cheating 100% unless you >>>>want to release codes etc... >>> >>>In a physical WCCC there is no easy way to cheat, apart from plagiarism which is >>>a different story. When playing against an engine in WCCC, I know that the move >>>it just played was produced by itself, not some external entity (machine or >>>human). Could you say the same in CCT? >> >>Again, see my previous post. I could use any program I choose and you would >>_never_ know or prove it. You might leave scratching your head "how did he >>do that?" but that's it.. >> >> >>> >>> >>> >>>> >>>>-S.
This page took 0 seconds to execute
Last modified: Thu, 15 Apr 21 08:11:13 -0700
Current Computer Chess Club Forums at Talkchess. This site by Sean Mintz.