EDGAR - Everyone Deserves a Game Above Reproach
|
ACBL Computer Based
Cheating Detection EDGAR - Everyone Deserves a Game Above Reproach |
|
Updated April 27, 2025
Key
Links EDGAR
INSTITUTE FOR BRIDGE ARBITRATION
Online Discipline Timeline
|
Online ACBL bridge has undergone major change. The ACBL
employs
Computer Based Cheating Detection. What follows outlines what the
process is, how the detection system works, and what arbitration means.
This information intends to supplement official ACBL publications so
members impacted by this process can know what to expect. The ACBL
remains the final authority in all disciplinary matters.
About 80,000 ACBL members have played bridge online since March 2020. Since there is no statue of limitations for cheating (here, illicit exchange of information), the ACBL will review everyone who has played online. To accomplish this Herculean task they have adopted Computer Based Detection and are using a system called EDGAR. The first group of players were scanned in January 2024. By March 2025, almost 38,000 members have been scanned, some more than once. About 3.25% of these have been accused of cheating online. Of the 784 convictions, 12 have completed arbitration and none have resulted in a reversal or acquittal. EDGAR is tuned to identify players who are clearly cheating by comparative analysis of patterns of behavior. How does Edgar work? That's what the links below reveal. Keep in mind that hundreds of hands showing patterns of behavior are necessary to convict a player. A one-off play can never convict. The system also effectively separates the novice or error-prone player from the cheater, and even unexpected winners from the cheaters. Each link has excerpts from the referenced article so that you can scan before you read. Additional Information: ACBL Policies and Governing Documents ACBL Bylaws - All members are required to follow the Code of Disciplinary Regulations and agree to arbitration with the Institute for Bridge Arbitration for ethical disputes. (see Article III.3.4 and 3.5 and Article XII). ACBL Codification - See Chapter I - Membership Operating Guidelines F. ACBL Code of Disciplinary Regulations and Disciplinary Administration. ACBL Code of Disciplinary Regulations -
Institute for Bridge Arbitration - Website ACBL Members Under Discipline ACBL Members Resigned with Pending Charges |
|
What is EDGAR and How Does It Work? |
|
Introducing EDGAR April 8, 2022 by A. J.
Stephani, Brian Platnick, and Franco Baseggio
"Instead of identifying single hands that look suspicious in a
vacuum, EDGAR will be able to identify whether funny-looking
bridge decisions are representative of a larger pattern or
constitute a one-off aberration. Inspired by the work of Nicolas
Hammond, Kit Woolsey, and others in the field, EDGAR will
finally show us the forest through the trees."
|
|
EDGAR Bad Landings Nov 7, 2022
by Brian Platnick and Franco Baseggio
"The core components of the system are “detectors”, each focusing on
one aspect of the game. All detectors, whether the focus is
bidding, opening leads, or defense, are constructed similarly: a
player or partnership has an opportunity to do something that looks
incriminating (a “hit”), absolving (a “miss”), or neutral.
Depending on the detector, hit/miss might be nothing more than a
good or bad decision/guess. One hit or miss only moves the needle a
little bit, so the results are aggregated across many plays and many
detectors before reaching a conclusion with confidence. How many
deals depends on how blatant the cheating."
Discusses the
meaning of "Bad Landings" as one exculpatory variable.
|
|
EDGAR the Tell-Tale Heart Nov 17, 2022
by Brian Platnick and Franco Baseggio
"LaterDefLeads (LDL) counts the double-dummy errors by each
defender, but only for tricks 2-12 and only when a defender leads to
the trick. It excludes the opening lead, and any cards defenders
play when following suit or discarding. Score = % of trick 2-12
leads by defenders which are double-dummy errors."
|
|
The Kit and the Algorithm Dec 14, 2022
by Brian Platnick and Franco Baseggio
"Edgar’s main opening lead (OL) detector, called “KOM”, ... All opening lead cheating detection approaches evaluate the lead actually made against a benchmark. Double-dummy accuracy simply evaluates whether the lead gives up any tricks at double dummy. But, perhaps the only DD accurate lead is quite unreasonable. Or, perhaps a very incriminating lead doesn't matter on the actual layout. ... KOM adds a twist in that some hits and misses are more meaningful than others. In this, it follows a more traditional investigation approach, which also typically distinguishes the strength of evidence in addition to classifying it as incriminating or absolving. For KOM, the more “swingy” a lead is, and the more it out- or under-performs plausible alternatives, the more points won or lost. ...
For example, KOM has separate detectors for Opening Leads vs. suits
and Opening Leads vs. NT.
(1) Weird Leads is a separate detector for the kinds of leads that experienced players almost never make, such as leading low from AQxx or KQJx against a suit contract. (2) Trump Leads are also excluded from KOM. Evaluating the attractiveness of a trump lead is more complex than other suits (or at least we haven’t solved it well enough yet). We still note their frequency and evaluate them in other ways. (3) Confusing Auctions. Edgar records explanations of calls, but players don’t include them consistently and don’t intend to undertake the sort of difficult natural language processing that would be required to understand them." |
|
EDGAR Tis the Win and Nothing More Jan
11, 2023 by Brian Platnick and Franco Baseggio “EDGAR: Tis the win and nothing more?” will look at data from winning pairs to see if EDGAR will mistakenly flag them as cheaters. Can EDGAR tell if the winners of an online pair event cheated their way to victory? Almost certainly not. Statistically based methods like EDGAR require much more than 2 sessions of data to determine the likelihood a pair cheated. It would be useful if there were a pair who won dozens of online regional events. Since we couldn’t find such a pair, we created one of our own - “Frankenstein”, the pair who always wins." "As you can see from the 2 tables above, whether or not they win, cheaters look like cheaters, and non-cheaters look like non-cheaters. One thing cheaters and non-cheaters have in common – they have better results when they make fewer errors as declarer. Similarly, in their winning sessions, their opponents make more errors than they do in their losing sessions." |
|
How Does EDGAR Work Dec
16, 2023 by Kit Woolsey "In real life, a pair isn't going to be faced with the identical problem 20 times in a row. But the pair will be faced with relatively similar problems. This is what EDGAR looks for: the run-of-the-mill decisions which an honest pair is going to get wrong some of the time, not just the "smoking gun" ace underleads which hit partner's king-doubleton. If a pair gets an unusually low number of them wrong, that is an indication that the pair may be cheating." |
|
EDGAR A Succession of Weird Leads June 29, 2024
by Brian Platnick, Franco Baseggio, and Andrew Yeckel
"we present an analysis that focuses on how EDGAR responds to weaker players who frequently make unusual leads. Will EDGAR mistake their inexperience for cheating? Our answer is no. ... EDGAR has a detector which tracks weird leads. A weird lead is deemed successful (a “hit”) when partner has a touching honor, and unsuccessful (a “miss”) when partner doesn’t. In millions of hands analyzed by EDGAR, the miss rate for weird leads hovers around 52%. Notably this rate is independent of playing ability, a point we return to later. A 52% miss rate with these underleads is a pretty bad result when you consider that the average error rate on opening lead is less than 25% for most players. It’s no wonder the typical expert almost never makes these leads. Yet these weird opening leads are commonplace among weaker players. In the hands analyzed by EDGAR, the weakest players make a weird lead about 1 in 40 times, while for the strongest players it is more like 1 in 1700. ... If I am an inexperienced player, should I be concerned that EDGAR will flag me as a cheater simply because I frequently make weird opening leads and get lucky sometimes? The short answer is emphatically no. When these leads are made innocently, their self-destructive nature will quickly exonerate the player. ...cheaters make weird leads much more often than non-cheaters, and they miss much less frequently doing it. The biggest improvement is found among experts, but even the most inexperienced players achieve a sizable reduction in miss rate. " |
|
The View from EDGAR Sep 12, 2024
by The Edgar Association Official
"Each EDGAR report is classified as RED, YELLOW, or GREEN. RED = overwhelming evidence of cheating. We look at every RED report before forwarding to the ACBL. If there is an anomaly, we will downgrade to YELLOW or GREEN." |
|
End of the Year Perspective from the IBA Dec
22, 2024 by Mark Friedlander
Overall, the arbitrators have done an excellent job. They have been judicial, unbiased, perceptive, detailed and (except for one instance) very prompt and timely. "1. To me, the most important lesson is that no charged party has yet been acquitted. 2. The charged parties have come up with a host of reasons to try to explain their odd but successful actions: miss-clicks, luck, a superior bidding system, superior play, etc. None of these explanations survived the arbitrators’ detailed examination of the hands played. 3. Interestingly, EDGAR (the computer algorithm that detects online cheating) has not been used as evidence in the arbitrations. It is used to help the Recorder’s office focus on players whose performance indicates that they were likely cheating, but then the Recorder painstakingly examines the deals played and assembles by hand dozens of such deals that imply that cheating has occurred. The ACBL’s evidence in the arbitrations consists of the Recorder’s personal hand analysis, not EDGAR statistics" 4. Because the ACBL’s “charging letter” to an accused player does cite EDGAR statistics, many of the charged parties have focused their defenses on attacking EDGAR’s credibility, even though its credibility isn’t an issue in the arbitrations for the reason above. 5. Overall, the arbitrators have done an excellent job. They have been judicial, unbiased, perceptive, detailed and (except for one instance) very prompt and timely. 6. At least in my opinion, not all cheating is the same. There appears to be a “spectrum” of cheating. At the most reprehensible end is the experienced pair that is fully aware of what it is doing and is cheating in order to beat other pairs and win masterpoints by illegally sharing information. At the other end of the spectrum is the confused, often elderly player who is un familiar with the online environment and to whom online play isn’t fully “real,” more of a “fantasy” form of bridge. 7. There is currently a controversy over what has been called “split discipline.” ACBL policy and rules require that all players who have been found to be cheating be given the same penalties, which have different gradations based mostly on masterpoint levels. The rules do not allow the arbitration panel to differentiate between the purposeful, intentional cheaters and those who were confused or careless and therefore not fully aware that what they were doing is cheating. 8. The $900 filing fee does not appear to be a problem. There has been only one case so far where the filing party claimed to be unable to afford it, and the IBA was willing to reduce the fee substantially for that party. 9. One weakness of the present approach is that charged parties are incentivized to postpone and drag out the arbitration process as long as possible because they are allowed to continue playing while their case is under consideration. 10. Although some charged parties who have been found guilty have protested and denied the correctness of the decision, threatening appeal and other court action, no one has yet seriously attempted to challenge the arbitration award. |
|
INSTITUTE FOR BRIDGE
ARBITRATION |
|
Institute for Bridge Arbitration September 27, 2021
by Mark Friedlander "Our concept is that arbitrations with bridge expert arbitrators certified by IBA would substitute for the disciplinary hearings that national bridge organizations typically hold as an initial adjudication of cheating allegations." |
|
Institute
for Bridge Arbitration Up and Running February 10, 2023 by
Mark Friedlander "I am very happy to announce that the Institute for Bridge Arbitration is fully operational and has been since approximately February 1. We have 43 certified bridge arbitrators, approximately one-third of whom are also certified to “chair” a bridge arbitration. The arbitrators include several world-class players, a significant number of national champions or frequent contenders, as well as many highly skilled and experienced players of somewhat lesser talent. Although most of the arbitrators are from North America, there are also several – all top players – from Europe and elsewhere. " |
|
Is the ACBL Trying to Discourage Arbitration of Cheating Charges
October 11, 2024 by Ray Yuenger "At 11 pages, this article covers a number of topics that I've tried to arrange in digestible pieces. Here's an index. (I'd provide hyperlinks if I could.) P. 2 The provisions requiring arbitration Pp. 3-4 Who is subject to fast track discipline; what discipline is imposed unless arbitration is elected; the deadline for responding Pp. 5-8 What discourages the arbitration option; the shock of being charged P. 6 What can be learned from the concise EDGAR report and the spreadsheet P. 7 What can be learned from the EDGAR report on hotspots P. 8 The costs of arbitration; the threat of increased discipline Pp. 9-10 The slower track to arbitration; notices of intent to initiate arbitration and offers of negotiated resolution P. 10 Deterrents to arbitration on the slower track P. 11 Final thoughts Note: This article describes the contents of the information provided to the accused player. Ray Yuenger is an ACBL member and active volunteer and does not represent official ACBL views. |
|
Fast-Track Discipline (Version 2.0) November 4,
2024 by Ray Yuenger " ... the American Contract Bridge League ("ACBL") has expanded the scope of its fast-track discipline ("FTD") to apply to all ACBL members, not just the 88% of members who have fewer than 2,500 masterpoints. This expansion has been accomplished through major changes to the Code of Disciplinary Regulations ("CDR") effective on November 1, 2024, ..." |
|
What Evidence is the ACBL Presenting to the Institute for Bridge
Arbitration? December 14, 2024 by Ray Yuenger " ... there are normally two stages of evidence presentation in a contested arbitration of cheating allegations. The ACBL first provides prehearing disclosure of all the evidence on which it may rely at the arbitration hearing. Second, at the hearing, in order not to prolong the hearing the ACBL presents a selected subset of boards from its prehearing disclosure, often organized to illustrate common patterns of misconduct. ... Following is a (crudely constructed) table of hearing reports listing where the ACBL's prehearing disclosures and hearing presentations can be seen. " |
|
The ACBL's Guidelines for Retroactive Disciplinary Changes
December 20, 2024 by Ray Yuenger "On November 4, I published an article that explained how the ACBL has revised and expanded its fast-track discipline ("FTD") system effective November 1, 2024. ... ... the ACBL's guidelines for applying the CDR changes. These were adopted, according to the draft minutes, as an action item during a special Board meeting two months ago on October 17, 2024, and concurrently with the CDR changes." |
|
How Often Have Accused Cheaters Won in Arbitration
March 17, 2025 by Ray Younger "...we now have 15 awards by the Institute for Bridge Arbitration ("IBA") that likely have resulted from ACBL charges arising from computer-based cheating detection (CBCD"; also know as EDGAR). In this article I'm going to try to classify those awards as wins and losses for the charged parties, though some may be debatable. ,,, If a charged party has competent representation capable of informing a panel of relevant CDR provisions and prior IBA awards, there may be a decent chance of improving on the ACBL's original discipline proposal, particularly if it was based on FTD Version 1.1. While we have seen awards based on discipline levels in FTD Version 2.0, we haven't yet seen awards resulting from charges under Version 2.0, so I have no basis to be as optimistic in encouraging future arbitrations." Wins so far are limited to reduced punishments, and not acquittals. |
|
The ACBL's Official Warning February 1, 2025 by Ray Yuenger ..."This article will examine its contents as well as the specified circumstances under which Official Warnings issue. ... There are several qualifications making an ACBL member eligible for an Official Warning: (1) no more than 500 masterpoints, (2) no Platinum points, (3) no record of either prior discipline for an Ethical Violation or an Official Warning, (4) no evidence of collusive cheating with more than one partner, and (5) completion of "the Ethics Certificate Program and Fair Play Pledge (EC/FPP) within 30 days." (§§ 502(B), 501(A)(1).) In addition, "Official Warnings are intended for less experienced players who do not hold, nor have held, substantial positions of influence or visibility within the ACBL. ACBL Management, with concurrence from A&C, will determine whether a player's role or involvement within the ACBL disqualifies them from receiving an Official Warning." (§ 502(A)(2); "A&C" means the ACBL's Appeals and Charges Committee.) " |
|
The ACBL's New "Notice of Fast Track DIscipline" Form April
27, 2025 by Ray Yuenger ..."the FTD Notice form has been a work in progress. Comparing Notices from late May and late August 2024, there were some changes in the later form, although both forms were still two single-spaced typed pages. The August form helpfully added advice to charged parties (a) that the burden would be on the ACBL to prove its charges if arbitration was initiated and (b) that the 30 days in which to elect arbitration was not the entire time allowed them to prepare their cases for arbitration." ... |
ACBL Online Discipline Timeline
|
Bridge Arbitration adopted in the bylaws |
|
|
Collusive Cheating Defined in the Code of Disciplinary Regulations (CDR) |
|
|
Ethical Violations updated and parsed into degrees of offense.
Negotiated Resolution. |
|
|
Appendix O – Conditions of Contest for Online Play |
|
|
Collusive Cheating Online definition added to the CDR. |
|
|
Institute for Bridge Arbitration Introduced |
|
|
ACBL Codifies the Institute for Bridge Arbitration as sole arbiter for
disputes involving ethics or behaviors. |
|
|
First Degree Ethical Violations Disqualify Hall of Fame and Grand Life
Master Titles |
|
|
EDGAR Introduced |
|
|
EDGAR approved for use |
|
|
Jan 31, 2024 |
First EDGAR runs – Testing the tool |
|
Fast Track Discipline for U2500 Players implemented
Computer Based Cheating Detection implemented |
|
|
Feb 14, 2024 |
EDGAR in Production |
|
Aug 7, 2024 |
EDGAR Paused |
|
Nov 1, 2024 |
Fast Track Discipline (2.0) – Applies to all members with exceptions |
The Appeals & Charges Committee Chair publishes Online Cheating
Reports every General Meeting, and those are made available on this website.