Posts: 112
Threads: 6
Joined: Oct 2011
Alan Jelfs Wrote:Andrew McHarg Wrote:andyburnett Wrote:it would be revealing to know the probability that a 2200 player could mirror Houdini 2's (rated about 3100) first choice over 95% of the time? I'd guess we're in the realms of lottery numbers (10 million to 1).
Much, much more unlikely than that. :U
But wouldn't you have to measure that against the probability of a 2600 player mirroring Houdini's moves?
Yeah, pretty much. You'd probably have to do something with the average strength of his moves to check if there is a difference between the mean strength of his old moves and new ones. You can gauge some kind of statistical evidence from that, as is done in normal cases, then look towards human evidence. Using an expert in computers and cheating (eg. Lilov) seems sensible. Looking at behavior during the time of accusation helps as well, for example, in the Feller case they worked out the captain was passing moves via a code of standing near boards.
Looks pretty guilty to me.
Posts: 678
Threads: 29
Joined: Apr 2017
Reputation:
3
Alan Jelfs Wrote:Andrew McHarg Wrote:andyburnett Wrote:it would be revealing to know the probability that a 2200 player could mirror Houdini 2's (rated about 3100) first choice over 95% of the time? I'd guess we're in the realms of lottery numbers (10 million to 1).
Much, much more unlikely than that. :U
But wouldn't you have to measure that against the probability of a 2600 player mirroring Houdini's moves?
If you look through historic games played by all the GMs in the history of Chess graded over 2600 you'll not find a single one whose moves as closely resemble Houdini's as Ivanov's moves did for that many consecutive games (except perhaps GMs who were cheating ;P ). I seem to recall reading somewhere on Chess.com that Bobby Fischer's moves were first choice top modern engine moves around 60% of the time, for instance. One of the things about most engines is that they often have different styles to top humans. Computers will go down apparently uncomfortable lines more than most top GMs, because computers are much surer of their calculations. Ivanov's games tick all the boxes of an engine. So, rather than Ivanov seeing the brilliance of the moves he played (which is as near to impossible as to make it practically impossible), for him to be innocent he'd have had to make the moves largely by chance. It could be said that some of the moves he made were so outwith his ability to calculate them that they are essentially - to him - random moves. To do that for a whole tournament the way he did is about as likely as winning the Euromillions jackpot a hundred times in a row.
Posts: 200
Threads: 6
Joined: Aug 2011
Reputation:
0
Here is a somewhat more convincing look at a statistical analysis of the moves of the player in question.
<!-- m --><a class="postlink" href="http://www.cse.buffalo.edu/~regan/chess/fidelity/ACPcover-and-report.pdf">http://www.cse.buffalo.edu/~regan/chess ... report.pdf</a><!-- m -->
<!-- m --><a class="postlink" href="http://rjlipton.wordpress.com/2013/01/13/the-crown-game-affair/">http://rjlipton.wordpress.com/2013/01/1 ... me-affair/</a><!-- m -->
It is good to see that Alan Tate's performance was vindicated, his opponents playing more computer moves.
It saddens me however to realise that, while I recognise many of the terms used, I have forgotten almost completely the statistics that I studied some 40 years ago. Is that Bayes rules perhaps?
Posts: 1,003
Threads: 101
Joined: Mar 2012
Reputation:
1
These Bulgarian 2100's are pretty strong! Top untitled player in the Masters was another youngish Bulgarian, rated 2126, playing at 2452 level for the tournament.
In the Challengers A played in the morning of the first week his stats were
Rating international 2126
Ratingperformance 2065
By the 2nd week, again the morning competition.
Rating international 2126
Ratingperformance 2536
Obviously warmed up by the 2nd week! By my reckoning he walked off with at least £3000.
Discuss
Posts: 1,003
Threads: 101
Joined: Mar 2012
Reputation:
1
Am I the only one who finds this stuff fascinating?
From Chris Rice on the ECForum
"Round 5 Yavor Todorov (2069) 1/2 - 1/2 Borislav Ivanov (2342). So a little tot up so far. With a starting rank of 34th, Ivanov now lies in 97th place, 21.9 Elo points down with a rating performance of 1981. He hasn't played anyone stronger than 2069 yet is only on +1. Can this really be the same guy that claimed he beat Houdini and Rybka by 10-0 scores?"
<!-- m --><a class="postlink" href="http://www.ecforum.org.uk/viewtopic.php?f=31&t=5085&start=135">http://www.ecforum.org.uk/viewtopic.php ... &start=135</a><!-- m -->
Posts: 247
Threads: 6
Joined: Dec 2011
Reputation:
0
andyburnett Wrote:These Bulgarian 2100's are pretty strong! Top untitled player in the Masters was another youngish Bulgarian, rated 2126, playing at 2452 level for the tournament.
In the Challengers A played in the morning of the first week his stats were
Rating international 2126
Ratingperformance 2065
By the 2nd week, again the morning competition.
Rating international 2126
Ratingperformance 2536
Obviously warmed up by the 2nd week! By my reckoning he walked off with at least £3000.
Discuss
To be fair, it is possible. Some games you can just lob pieces, and a bad start gives you a bad mental frame. If he starts week 2 with a big win, confidence can grow. Personally speaking, I don't like the lynch mobs who try and seek out cheats based on people having good performances. The odds of these sorts of results are small, but then again, someone does win the lottery most weeks.
In the case of the now famous Mr Ivanov, I was pretty sure he was cheating, but not enough to condemn the man. The phrase "beyond reasonable doubt" springs to mind, and sometimes people can just streak huge runs in agreement with a machine. For example, I had a league game recently where my 1800 opponent played a near perfect game, and obviously there is nothing dodgy there, it just occasionally happens that people can play like that. The clincher for me though about Ivanov was when he claimed he had beaten Rybka and Houdini 10-0 each. That is ridiculous, and it is weird that a 2000+ can even say that and hope it would be believed. The easy way to disprove that is to give him a challenge game against Houdini, and I would be tempted to bet my life savings that he would get wiped out 10-0.
Posts: 144
Threads: 32
Joined: Aug 2011
"The easy way to disprove that is to give him a challenge game against Houdini,
and I would be tempted to bet my life savings that he would get wiped out 10-0."
I'm not too sure.
Say the challenge was 10 one hour games v Houdini.
You get Houdini to play itself at a one hour setting for say 30 games.
Then memorise the games where it won. I reckon I could store 10 complete games at the very least.
Use the same commercial package and produce a couple of these.
Or even more in fitting with the whole scheme.
Write your own or tamper with Houdini. (this lad is a computer programmer)
Save it to disc, make the said disc to look the Houdini disc with a disc printing kit
and place in it's Houdini box.
Go around to someone's house Invite a load sceptics around,
load up the fake Houdini and beat it once with White and
once with Black. (this doctored Houdini only plays two games...two brilliant games.)
Switch off the computer, look at the startled crowd and say: "how easy was that."
You now have a crowd of witnesses, also some spark will run a match up and no doubt
found it does not match up with the Houdini percentages thus bringing into doubt the
whole match up system that is hanging him.
Posts: 112
Threads: 6
Joined: Oct 2011
Adam Bremner Wrote:In the case of the now famous Mr Ivanov, I was pretty sure he was cheating, but not enough to condemn the man. The phrase "beyond reasonable doubt" springs to mind, and sometimes people can just streak huge runs in agreement with a machine. For example, I had a league game recently where my 1800 opponent played a near perfect game, and obviously there is nothing dodgy there, it just occasionally happens that people can play like that.
I remember that game, your opponent did play pretty well. Two things though,
1- the position was relatively simple, making moves easier to find and
2- perfect isn't a good word there: you mean the moves the computer thinks are the best. Again, the position was simple and the computer can't show it's full strength. I'd like to think if I played a "perfect" game against most people, I'd win...
Posts: 247
Threads: 6
Joined: Dec 2011
Reputation:
0
Geoff,
Just get some reputable organisation to set the match up, and randomize the opening play in some way. Then you will get 10-0. Also, if you are running the copy of a different machine, it will yield different results when you start changing the cores and memory about, so you don't need to worry about memorizing games.
Calum,
Yeah, it is meant more as an example to illustrate a point more than anything. People can streak the same as a machine, so in itself I would find it hard to condemn someone based on just that. There are other factors to consider.
Posts: 144
Threads: 32
Joined: Aug 2011
Hi Adam.
I was thinking about this at work.
The flaw in memorising the games where Houdini beat Houdini is that you would
has to memorise the exact time the winning Houdini played it's move.
If you played say 18.Ne4 after one minutes 'thought ' when infact the original Houdini
took say 4½ minutes to play the same move. You will get a different reply from the other machine.
Then you are own your own and in trouble.
Plan 2. Is the way to go.
|