Latest Updates:
Page Index Toggle Pages: 1 [2] 3 4 ... 7
Topic Tools
Very Hot Topic (More than 25 Replies) Can Fritz estimate my Elo rating? Yes. (Read 31106 times)
ErictheRed
God Member
*****
Offline


USCF National Master

Posts: 2484
Location: USA
Joined: 10/02/05
Re: Can Fritz estimate my Elo rating? Yes.
Reply #86 - 06/11/14 at 14:24:44
Post Tools
Keano wrote on 06/11/14 at 11:07:48:
Not reliable evidence, and not enough to establish probable cause, otherwise the McCanns would be arrested, but thats another story.


Depends on the jurisdiction, the judge, etc.  But it's a poor analogy, I think.

A better is probably plagiarism cases in academia.  How similar to another paper does one paper need to be before it's definitively considered to have been plagiarized?   

To answer my own rhetorical question from a few posts back, I think that if we genuinely only saw one false accusation every 20 years, and there was a just appeals process, that would be a very small price to pay to protect the game.  I am presuming, as verbiage suggested in the paper, that a pattern of engine-matching moves over a course of multiple games/events will be needed for a conviction, so that a single well-played game or a single piece of home preparation wouldn't be a problem.  I highly doubt that a well-prepared miniature is going to result in a cheating accusation.

There was absolutely no reason for Ivanov to have been allowed to continue to play for as long as he did, for instance.  That was just blatant and shameful, could have easily been stopped, and should be stopped in the future.  When someone's moves correlate something like 98% of the time with an engine's top choice, over multiple tournaments, he needs to be investigated and, possibly, banned for cheating even before the mechanism he used to cheat with is found.

Think of the French players in the 2010 Olympiad.  They had nothing on their person and no "proper evidence" would ever have been found by searching them.  More has to be done to discourage cases like this.
  
Back to top
 
IP Logged
 
Keano
God Member
*****
Offline


Money doesn't talk, it
swears.

Posts: 2891
Location: Toulouse
Joined: 05/25/05
Gender: Male
Re: Can Fritz estimate my Elo rating? Yes.
Reply #85 - 06/11/14 at 11:07:48
Post Tools
ErictheRed wrote on 06/10/14 at 18:57:25:
"Sniffer dogs finding a scent" is of course evidence, enough to establish probable cause and depending on the jurisdiction can result in a legal search without requiring a warrant.

Statistical evidence is evidence; the question is how strong of evidence it constitutes.


Not reliable evidence, and not enough to establish probable cause, otherwise the McCanns would be arrested, but thats another story.
  
Back to top
 
IP Logged
 
RdC
God Member
*****
Offline


I Love ChessPublishing!

Posts: 865
Joined: 05/17/08
Re: Can Fritz estimate my Elo rating? Yes.
Reply #84 - 06/10/14 at 23:10:02
Post Tools
ReneDescartes wrote on 06/10/14 at 21:16:47:
However, it is not clear to me that anywhere near that many chess games will be played in the near future if we exclude blitz. 


There are now over 100,000 players with International ratings
http://ratings.fide.com/card.phtml?event=24514748

So if each of them play an average of only 10 games a year, that's a million without trying. It varies by country, but perhaps only around 10% of "serious" players have International ratings. If that's a valid statistic, one in a million measured by games could be as frequent as once a month. It is after all FIDE and its marketing partner Agon, who claim astronomic numbers for the number of chess players in the world.

Chess servers have been busting people for unauthorised computer use for years. In the anonymity of your home, who knows what you are consulting. But OTB play by its very nature is conducted in public.
  
Back to top
 
IP Logged
 
ReneDescartes
God Member
*****
Offline


Qu'est-ce donc que je
suis? Une chose qui pense.

Posts: 1098
Joined: 05/17/10
Gender: Male
Re: Can Fritz estimate my Elo rating? Yes.
Reply #83 - 06/10/14 at 21:16:47
Post Tools
As a matter of fact, if one million chess games are played, there is a 63.2% chance that at least one of them will hit the one-in-a-million z-score at random. You don't have to be Regan to make that calculation: 1-[(1-.000001)^1000000]=.632. However, it is not clear to me that anywhere near that many chess games will be played in the near future if we exclude blitz.  I might be comfortable with a higher z-score. But this is not a criminal court. We are not putting anyone in jail, and we have to protect the game, which would become unplayable if cheating were widely regarded as prevalent.

Regan casually estimates the rating of perfect play. The notion of a "perfect rating" at first seems strange. But (supposing that chess is a draw) if an engine could draw a complete tablebase 2/3 of the time, it would win 1/3 of the points and have a rating 100-200 points lower than the tablebase. If that same engine beat the world champion a certain percentage of the time (or beat an engine that beat an engine that beat the world champion), we could calculate the rating of perfect play relative to a human player accurately. I would conjecture that we are still extremely far away an engine that could draw a tablebase regularly--especially if the tablebase were administered by a program that selected among equivalent moves in such a way as to maximize, for example, the horizon effect!

I do not believe, however, that we have any clue how a perfect tablebase would perform against current engines. Even a good model of the currently observed decrease in errors or increases in draws as ratings go up could not be extrapolated, because the instrument we are measuring with (another non-tablebase engine) is flawed to an unknown degree.
  
Back to top
 
IP Logged
 
ErictheRed
God Member
*****
Offline


USCF National Master

Posts: 2484
Location: USA
Joined: 10/02/05
Re: Can Fritz estimate my Elo rating? Yes.
Reply #82 - 06/10/14 at 18:57:25
Post Tools
"Sniffer dogs finding a scent" is of course evidence, enough to establish probable cause and depending on the jurisdiction can result in a legal search without requiring a warrant.

Statistical evidence is evidence; the question is how strong of evidence it constitutes.
  
Back to top
 
IP Logged
 
Keano
God Member
*****
Offline


Money doesn't talk, it
swears.

Posts: 2891
Location: Toulouse
Joined: 05/25/05
Gender: Male
Re: Can Fritz estimate my Elo rating? Yes.
Reply #81 - 06/10/14 at 13:53:34
Post Tools
Statistical analysis can never be used as evidence imo.
It is equivalent to sniffer dogs finding a scent - humans must then find proper evidence.

In the end the only way is to search the player for electronic devices, this is how Ivanov was caught and how future cheaters will be caught.
  
Back to top
 
IP Logged
 
RdC
God Member
*****
Offline


I Love ChessPublishing!

Posts: 865
Joined: 05/17/08
Re: Can Fritz estimate my Elo rating? Yes.
Reply #80 - 06/09/14 at 22:15:32
Post Tools
ErictheRed wrote on 06/09/14 at 20:21:48:
If one innocent person is accused every 20 years and a just appeals committee handles the case (or preferably a panel of experts would look over the evidence before an automatic ban), is that acceptable? 


I would say no it isn't. I think they've underestimated the amount of chess actually played. Even if you just take the number of rated players and multiply by the number of rated games they play, I'd reckon you get well over a million games a year. So one in a million chances are nothing. If you run with some of the more extreme estimates that FIDE sometimes come out with for the number of chess players in the world, a million is "hardly any".

You might be able to detect the influence of a computer in a game. That of itself is not evidence of cheating as the whole game could have been prepared beforehand. A key point is that you shouldn't be able to cheat in over the board chess, as in principle you are always in sight of your opponent, the arbiters and spectators. Setting aside those who come tooled up with spy like gadgets, the menace is where players can slip away out of sight and communicate with an engine or a helper in private.
  
Back to top
 
IP Logged
 
ErictheRed
God Member
*****
Offline


USCF National Master

Posts: 2484
Location: USA
Joined: 10/02/05
Re: Can Fritz estimate my Elo rating? Yes.
Reply #79 - 06/09/14 at 20:21:48
Post Tools
Smyslov_Fan wrote on 06/09/14 at 18:44:54:
If you read the proposal closely, the ACC states that if the certainty of engine use exceeds a z-score of 4.75, (1/1million) that alone will be enough to conclude that an engine was used. (page 23 of the PDF, Annex C section 2 point D.)

http://www.fide.com/images/stories/NEWS_2014/FIDE_news/ACC_draft_proposal.pdf

So yes, if the numbers are extreme enough, they alone will be enough to convict someone.


I don't think that it says that explicitly, and I think that panic is misrepresenting the proposal.  I'm not a lawyer, though.

Quote:
The ACC does not simply use either the standardly-recognized “5% threshold” or “1% threshold” for significance of p-values, but rather demands more stringent thresholds depending on the absence or presence of other evidence, the size and nature of the tournament, and the circumstances of the complaint.  The following guidelines are recommended:

A. A z-score under 2.00, commonly regarded as failure to pass the 5% threshold, may be considered a finding that statistical evidence does not support a complaint.
B. A z-score of 2.75 or greater, representing a 0.3% threshold, may constitute strong supporting evidence in the presence of physical or observational evidence.
C. Higher thresholds may be deemed needed for further stages of a FIDE-level judicial process.
D. For statistics to be considered as sole evidence for judgment, a z-score of at least 4.75 (p = .0001% or 1-in-a-million threshold) is needed, from one event or as a combined z-score from several events in close succession.

For comparison, the scientific standard for declaring new components of Nature such as the Higgs particle or gravity waves to exist is z >= 5.00. Based on the volume of recorded chess games, even if full tests were done on every game by every player, a z >= 4.75 would be observed in normal play only once every 20 years, and z >= 5.00 once in 60 years.

When a full test is conducted in response to a formal complaint, the results shall be included in the report on the complaint. A full test performed at the CA’s discretion when there is no allegation is private. Test results may also warrant overt measures taken by arbiters onsite, such as increased watch, searches, changes in game locale or environment, subject to considerations in other parts of this document.



There is also a proposal for an appeals process.  If one innocent person is accused every 20 years and a just appeals committee handles the case (or preferably a panel of experts would look over the evidence before an automatic ban), is that acceptable? 
  
Back to top
 
IP Logged
 
Smyslov_Fan
YaBB Moderator
Correspondence fan
*****
Offline


Progress depends on the
unreasonable man. ~GBS

Posts: 6902
Joined: 06/15/05
Re: Can Fritz estimate my Elo rating? Yes.
Reply #78 - 06/09/14 at 18:44:54
Post Tools
ErictheRed wrote on 06/09/14 at 17:47:34:
RdC wrote on 06/09/14 at 16:53:10:
If a statistical method were ever used entirely on its own to ban a player, I think there would be serious legal challenges and the "confidential" statistical method would have to be exposed to external scrutiny. That would show whether it satisfied the obvious tests of not labelling someone from before the computer era as having used a computer and being able to correctly identify games played wholly or partly by engines.



I don't think that anyone is talking about that.  The statistical analysis will be a piece of evidence, and likely no-one will ever have their games scrutinized unless they exhibit suspicious behavior to begin with.


If you read the proposal closely, the ACC states that if the certainty of engine use exceeds a z-score of 4.75, (1/1million) that alone will be enough to conclude that an engine was used. (page 23 of the PDF, Annex C section 2 point D.)

http://www.fide.com/images/stories/NEWS_2014/FIDE_news/ACC_draft_proposal.pdf

So yes, if the numbers are extreme enough, they alone will be enough to convict someone.
  
Back to top
 
IP Logged
 
RdC
God Member
*****
Offline


I Love ChessPublishing!

Posts: 865
Joined: 05/17/08
Re: Can Fritz estimate my Elo rating? Yes.
Reply #77 - 06/09/14 at 18:38:36
Post Tools
ErictheRed wrote on 06/09/14 at 17:47:34:
I don't think that anyone is talking about that.  The statistical analysis will be a piece of evidence, and likely no-one will ever have their games scrutinized unless they exhibit suspicious behavior to begin with.


The problem is that this is not what the Committee are proposing.

The draft proposals can be read at
http://www.fide.com/images/stories/NEWS_2014/FIDE_news/ACC_draft_proposal.pdf

On page 2, they propose an anti cheating committee which wants powers to
ii) perform sample checks on players and tournaments both on-site and remotely

and on page 3
The Committee recommends the implementation of a FIDE Internet-based Game Screening Tool for pre-scanning games and identifying potential instances of cheating, together with the adoption of a full-testing procedure in cases of complaints.

On page 5, you can read

With a view to creating a sufficient unbiased database of games and to make statistical analysis even more accurate, all games played after 1.1.2012 are subject to potential screening
by the ACC.

  
Back to top
 
IP Logged
 
barnaby
Senior Member
****
Offline


The night is dark and
full of terrors.

Posts: 345
Joined: 01/09/12
Gender: Female
Re: Can Fritz estimate my Elo rating? Yes.
Reply #76 - 06/09/14 at 18:17:28
Post Tools
MartinC wrote on 06/08/14 at 07:53:40:
Why only 250? You're talking about storing 100+ mobile phones, many with replacement value of 5/600 pounds.....
(And a mass of fairly crucial personal data for a lot of people too.).



fwiw: i just checked with the insurance company we use to write policy on tournament here (we usually have from 100-500 people).

This phone policy can be had for three days for a rider of $107.00 US currency.

Erie Insurance Co.

  
Back to top
 
IP Logged
 
ErictheRed
God Member
*****
Offline


USCF National Master

Posts: 2484
Location: USA
Joined: 10/02/05
Re: Can Fritz estimate my Elo rating? Yes.
Reply #75 - 06/09/14 at 17:47:34
Post Tools
RdC wrote on 06/09/14 at 16:53:10:
If a statistical method were ever used entirely on its own to ban a player, I think there would be serious legal challenges and the "confidential" statistical method would have to be exposed to external scrutiny. That would show whether it satisfied the obvious tests of not labelling someone from before the computer era as having used a computer and being able to correctly identify games played wholly or partly by engines.



I don't think that anyone is talking about that.  The statistical analysis will be a piece of evidence, and likely no-one will ever have their games scrutinized unless they exhibit suspicious behavior to begin with.
  
Back to top
 
IP Logged
 
RdC
God Member
*****
Offline


I Love ChessPublishing!

Posts: 865
Joined: 05/17/08
Re: Can Fritz estimate my Elo rating? Yes.
Reply #74 - 06/09/14 at 16:53:10
Post Tools
Smyslov_Fan wrote on 06/09/14 at 16:32:44:
There are going to be logistical problems with scanning in the game scores, and other issues. I think the issue of where the cell phone is kept is a very minor one compared to that one.


As a matter of routine, many tournaments use duplicate score-sheets and have someone inputting them to create a pgn file of the games. If nothing else, it helps tournament publicity if the games are published on the web and in TWIC. So that doesn't introduce anything new apart from a need to find a volunteer willing to do the inputting.

The real problem is that someone remote to tournaments is asking for the power to label players as cheats based on no more than an analysis of the moves played.

If a statistical method were ever used entirely on its own to ban a player, I think there would be serious legal challenges and the "confidential" statistical method would have to be exposed to external scrutiny. That would show whether it satisfied the obvious tests of not labelling someone from before the computer era as having used a computer and being able to correctly identify games played wholly or partly by engines.

Given that ICCF permits computer use, or at least doesn't outlaw it, scrutiny of recent correspondence games would throw up additional insights.
  
Back to top
 
IP Logged
 
Smyslov_Fan
YaBB Moderator
Correspondence fan
*****
Offline


Progress depends on the
unreasonable man. ~GBS

Posts: 6902
Joined: 06/15/05
Re: Can Fritz estimate my Elo rating? Yes.
Reply #73 - 06/09/14 at 16:32:44
Post Tools
And that's where the groundbreaking news that FIDE will start to use statistical testing to uncover computer cheats comes in.

I'm flabbergasted that such a conservative organization as FIDE would go from virtually no method of catching cheats to creating an Anti-Cheating Committee with the power to analyse every game played. This is a huge step. I think it's in the right direction, but it really will change how tournaments are run.

There are going to be logistical problems with scanning in the game scores, and other issues. I think the issue of where the cell phone is kept is a very minor one compared to that one.
  
Back to top
 
IP Logged
 
ErictheRed
God Member
*****
Offline


USCF National Master

Posts: 2484
Location: USA
Joined: 10/02/05
Re: Can Fritz estimate my Elo rating? Yes.
Reply #72 - 06/09/14 at 14:43:00
Post Tools
MartinC wrote on 06/09/14 at 14:29:12:
Definitely getting too obsessed about them Smiley Anyone who's premeditating cheating in a halfway competent manner will use something that's much harder to detect than staring at their smartphone!

Worries about the threat of casual cheating perhaps. The other thing of course is that smartwatches/ wearables are coming/even arrived. So do we ban watches next?


Well, a determined cheat will find a way.  And even a non-determined cheat can have a smart phone on their person and not turn it into the organizers (are we going to search every person)?  Or leave it with their friend who they meet in the bathroom, or just hide it in the bathroom a la the handgun in The Godfather...
  
Back to top
 
IP Logged
 
Page Index Toggle Pages: 1 [2] 3 4 ... 7
Topic Tools
Bookmarks: del.icio.us Digg Facebook Google Google+ Linked in reddit StumbleUpon Twitter Yahoo