Join Bridge Winners
All comments by Nicolas Hammond
You are ignoring the author of this comment. Click to temporarily show the comment.
http://www.bridge.dk/2131/Klub1/SuitCombinations.pdf
Page 20. #38
Low to 9; then low to Q.
Feb. 8
You are ignoring the author of this comment. Click to temporarily show the comment.
No. One of the steps was to find if the bots were cheating. No evidence. No surprise either.

Second was to compare the quality of the play with humans. The bots are getting into the top 500.

What you are proposing is actually a little hard. Program a bot to cheat. If they cheat all the time, it is trivial to detect. So… they only have to cheat some of the time. This is actually a lot harder problem to solve. What's the difference between a bot that only cheats some of the time and a bot that had a software update and plays better. Not much. You could add a random factor - we cheat on x% of boards, and keep changing x until detectable. My reference point for cheating pairs is the known cheating pairs. All of who were already in the top 200 of Bridge pairs. Add cheating to their skill level was detectable. If you improved a bot from the top 500 to the top 400, it is difficult/impossible to prove cheating and not a good software update. But put them in the top 10 and it is probable to find. Cheating bots; and cheating humans will be different. There are other tests I use which can detect certain types of cheating.
Feb. 7
You are ignoring the author of this comment. Click to temporarily show the comment.
Already done this. Chapter 27 of the book.

Ran it against data kindly provided by Al Levy from the World Championships.

Data shows that the bots are getting better (implicitly obvious, but the cheating detection software doesn't know this but was able to work out from the play). Data shows that as of 2017 the Robots are getting close to breaking the top 500 pairs if they were to play against humans.

I am able to compare the bots ability with the average ability of different ACBL MP ranges.
Feb. 7
You are ignoring the author of this comment. Click to temporarily show the comment.
No details. Person(s) involved sucked it up. Moved on. Doesn't complain. Doesn't want to get involved in these types of discussions.

Case 2 was complicated for ACBL and all involved. At least one of the two teams had > 4 players. Not all players necessarily knew or were involved. Some willing to plead guilty, take a light punishment and move on irrespective if they were ‘guilty’. Others not so. Any plea-deals are typically not public (I have no specifics).

The ‘victims’ would be the other teams (excluding these two) who were in the event. I believe the event was rescored with 0-0 VPs for each team, so the other teams were not victims. Some of the players on the two teams inadvertently became victims as well.
Feb. 7
You are ignoring the author of this comment. Click to temporarily show the comment.
That we should play together. We are as bad as each other on defense.

In all seriousness…. what I look at next, for a pair with a wide divergence as described, is the individual player ability within the partnership. Who declares better etc. Often in these cases there are some Law 40 issues. Typical with pro/client. The pro is declaring a lot. The client little. The numbers will show this. Then, if I were ACBL, stick a camera on pair. The most common thing to watch for is 1NT openings; pro has a wide range, client is 16 HCP, 4333/4432 hands only. Er…. I'll stop there…. the ACBL NDA may kick in.

Next time I pull up the various data sets, I'll post where you are w.r.t. other players with similar number of boards and we can see. Where do you think you really are? PM a friend; I'll post the results on you and your regular partners, then you can post what you PM-ed a friend and see how close we are.

In the process of a big software update on the cheating detection code; which will make it a lot easier to pull up information a lot quicker on different data. Previously with the very large data sets involved, I generated the data to an Excel spreadsheet for subsequent processing.
Feb. 7
You are ignoring the author of this comment. Click to temporarily show the comment.
I don't think the legal counsel in case 1 would like to be called “massive”. ('massive legal counsel').

Bonus points for anyone who knows the connection between cases 1 and 2. (No need to PM me).

There were some serious personal repercussions, though little known, for some of those in case 2; for example, inability to represent the US in world events for a period of time. I know of at least one player affected. Fines are not typical in the US.
Feb. 7
You are ignoring the author of this comment. Click to temporarily show the comment.
TLDR (Too long, don't read) for those unfamiliar with the acronym.

I hadn't checked this page for a couple of days; so didn't realize Boye had posted. I happed to talk to Boye yesterday without knowing that he had posted here.

“Most of the top pros know who they are”. Turns out he does go the right bars! We had the same set of names :-)

It is frustrating for Boye: he plays at that level; I don't. It is frustrating for me, because when I do supply the name of a pair, little seems to be done. For one pair that I supplied to both ACBL and WBF, the video quality in each case was too poor for any analysis. It is a pair that Boye was aware of through the “rumor mill”. There was no need to enlighten him; he already knew.

I'm very fortunate: I have a $1.5M piece of Bridge software from ACBLscore+. I've continued to work on that software, added a lot of cheating detection software, done a lot of research into the statistics and mathematics of Bridge. I have filed for a patent on the work. For those interested further, I wrote a book on Detecting Cheating in Bridge, shameless plug: http://www.detectingcheatinginbridge.com/

What I haven't done, so far, is charge anyone for any of the work I've done. I now have a commercial product that can detect cheating. Or, to be more precise, it can provide details on the likelihood that a pair is cheating.

Two years ago I made a proposal to the WBF. To detect cheating at the elite level requires data and analysis from tournaments from around the world. My cheating detection software currently has data from EBL, WBF, ACBL, BBO, TVP (The Vugraph Project). Sometimes it requires the combination of data from all these sources to provide sufficient data for analysis on a pair. I know, with a reasonable confidence level, the pairs that have statistically questionable data from ACBL tournaments (regionals/sectionals) and have provided some data to ACBL.

Like Boye, I've been in some of the ACBL Anti Cheating Committee (ACC) meetings. I signed a NDA. The ACC works best when there can be open discussion about any and all pairs. The discussions should remain confidential. I haven't disclosed anything from those meetings. Occasionally I get asked by ACC members about data on some pairs outside of the ACC meetings; those aren't covered by any ACBL NDA and I usually provide information. The ACC members who have asked me (including Boye) know what I provide and what I don't.

The ACC would be better split into two: one focused on play at the top level (think NABC+), and one focused on cheating/incidents at the club/sectional/regional tournaments level. The ACC has some non-residents of ACBLland (e.g. Boye + others), who care little/know nothing about some B level players at a Sectional. There are also players on the ACC who do not compete at the elite level and offer little value to discussion on top pairs. Yes, there is a small intersection (I went to the same college as Venn!), but it is a small intersection. Separating the committee into two sub-committees would help. The NABC+ committee can use information from other tournaments around the world. The regional/sectional committee would be typically focused on the lower level players.

One of the biggest advantages to the cheating detection software is the ability to show a pair is probably not cheating. Most people may not realize this. For example, there would be discussion in the ACC about a pair and a suspicious lead. “They led this when that was clearly called for”. My software is able to show the numerous times a pair did make an unusual lead and it did not work out. The one lucky time was reported, the numerous unlucky times are never reported. Having this knowledge significantly reduces the effort in deciding which pairs to look at. It is the most beneficial tool, IMHO, for the ACC.

A better example might be the Blue Team. The perception is that everyone on those teams cheated. My software says no. I was chatting to Bob Hamman that not everyone on those teams cheated, and he said, “oh, you mean a/b, c/d ..”. The results matched. This doesn't mean that a pair is not cheating; it just means that the results that they are achieving are typical for normal players at that level. It is only when cheating improves your game beyond statistical limits that it can be detected.

The last ACC meeting I attended was March 2019. I was never a member of the ACC, simply an invited guest. “Fox in chicken house” is what I wrote to the person next to me, who I know well through the cheating detection work, when I saw who was in the room. I haven't been to an ACC meeting since. I have yet to see who the 2020 appointees are.

This doesn't mean I stopped working on any anti-cheating efforts.

Since March, I have provided ACBL details on some ACBL players even though I no longer attend the ACC meetings. I provided some after the summer ACBL and also gave ACBL a copy of my book. My reports are not, “I think x is cheating”, they have been either detailed reports backed up with full documentation and cross-referenced with original source material showing the statistical likelihood of a pair doing something; or have been references to videos supported by documentation that a pair is doing something.

What happens next is up to ACBL. I only know of the success of anything I submit by checking the ACBL discipline list: http://web2.acbl.org/discipline/CurrentlyUnderDisciplineList.pdf. I have had some successes.

I am free to discuss anything that I generate from the work I do, and data I have; I am not bound by an ACBL NDA. I generate data from publicly available sources. The data I have from non-publicly available sources is only shared with the organizations that provide it. I won't mention who they are, but some may be surprising. What happens in public discussion groups (like this), and what happens in private are two different things.

“but either he undersupplies of overstates his names sharing”

I'm not going to be baited. I've worked with Boye since 2015 and shared a lot of information with him, and it has been a two-way communication. The conversations are private. He has knowledge from playing against pairs, I have knowledge from the data. When we chatted yesterday I rattled off the list of at least three pairs that I have provided ACBL since March. Remember this is information from my data so I am free to share. I don't provide the information to the entire ACBL ACC; that's not how the process works. Only Boye will know if the information has been shared with him in the ACC meetings; I'm not going to ask, the ACC discussions are private.

And, before anyone asks, I was not paid for any of the information I provided to ACBL; nor did I ask to be paid.

What Boye and others would like from me are the list of pairs pre 2015 that were cheating. To be more precise, there is statistical likelihood that they “must” be cheating. These are >1 pairs that were better than, say, BZ or FN in defense and played a large number of boards. Boye (+ others) have seen the charts from the data, so they know these pairs exist. The problem is, for all these pairs, there is not enough video data to find a code. Believe me, I've probably watched more pre-2015 Bridge videos than anyone, and some of the Bridge organizations have been kind enough to share their private video to help me with the search. I'm not going to name these pairs. Some of them still play together. I do statistics on the tournaments they play in. All these pairs had significant and noticeable drops in their statistical performance after 2015. Back to the level of typical expert players. In other words, they have probably stopped doing what they did pre-2015. Naming these pairs won't help anything. I may have to wait for them to pre-decease me before publishing, though all my data is from public sources.

My background is computer security. Back in the day (20+ years ago!), I would be at a site, and the CTO (Chief Technical Officer, though they were known as something different back then) would complain that their Internet speed was slow. It was trivial to show the CTO how to install software to monitor the company's outbound Internet connections. Usually porn and gambling sites. My recommendation was always the same: At week 1, publish the fact that sites were being monitored and a list of sites would be made public at the end of the week. We would publish the list. There would have been a drop, but only a small drop, people didn't believe it would happen. The following week the company would announce that they had installed better software and the list of people going to the sites would be listed. Of course, we had that information all along, but didn't release. At the end of week 2, there was almost no activity on porn or gambling sites. Ironically it was usually the CEO that was the most guilty party. This solution fixed the problem. No-one, except the CTO, was aware of who had been doing what and everyone walked away happy.

What Boye did in 2015 was to change the behavior of top bridge players in a similar way.

What needs to happen next is for the behavior of the non-elite bridge players to change. The greyer areas that Boye refers to. And I know, from the last NABC, when Boye/I talked that this has been a focus area for him for the top players. For us in the lower levels, we are all aware of the players that “hitch”, lead obvious singletons v. doubletons that partner seems to get right. I'm not saying that these are collusive cheaters, just by playing together for so long that they are attuned to how partner reacts. These are top pairs, pro/client, and amateur/amateur. For example, I have ranked lists of how good various ACBL partnerships are on defense and declaring. If you are in the bottom 10% of declaring and in the top 5% of defense for partnerships with a similar number of boards played, there is some suspicion. The rankings I use are calculated from public data, and though perfectly publishable according to my lawyer, it is easier for me not to publish. One of the ways of improving changes in the greyer area would be to start publishing some names. I may do this; or for the first week, may just publish sufficient information to show that I can and have been generating this data, but without the names. Then start to publish the names, but only use data post initial publication.

I have been honest with both Boye, ACBL, ACC and WBF. I have commercial software that I would like for them to license. In my mind the best way is for the WBF to license and then sub-license it out to all the NBOs. This way all NBOs would have access to not only their local data; but also data on top pairs from multiple tournaments. It is not in my commercial interests to give all the information away for free, including who is likely to be cheating, and all the methods that are used.

Before anyone would license the software, they need to know it works. This is a difficulty problem. Giving source code to mathematicians doesn't work. There is no single magic formula, though I have a couple that are close. For example, supposing I said the % of contracts defeated on defense is the formula. Great. But how to you prove this detects cheating? You can't without applying the known data sets. My answer to anyone that wants verification is to start with a clean database, extract the data from a known site, and see what names are shown. If further verification is needed, do the same but ask the organization to change player names and remove some of the data. My software will still work. Another way is for me use the software, find an as-yet-unknown cheating pair, then find their code. As anyone who has looked at videos knows, this is a very complicated and time-consuming project and, unfortunately, not something I can outsource to volunteers easily because the names would be quickly known. Alas, there just isn't enough video data from pre-2015 to find a code.

Post 2015, statistics are the only way to show that cheating is happening. Any pair that deliberately cheats will use a non-breakable code that varies each session. Never get caught from video.

There are still some active collusively cheating pairs at the elite level - Boye knows who some of them are - but there are far fewer than there were pre 2015. Unfortunately, the videos from 2019 were not of sufficient quality to detect cheating, i.e. even assuming that they were using a simple enough code to crack, the camera wasn't pointed at the table, one player was not visible, camera out of focus etc. etc.
Feb. 7
You are ignoring the author of this comment. Click to temporarily show the comment.
You are ignoring the author of this comment. Click to temporarily show the comment.
Does this affect your opinion on using win/loss or VPs for deciding the second place team in three way matches?
Feb. 5
You are ignoring the author of this comment. Click to temporarily show the comment.
“The ACBL managed to piss away 2+ million dollars on what should have been a simply software update”

It was a complete re-write of the software from Pascal on DOS (ported to Windows) to a modern programming language running on a SQL database that had to run on modern Windows and Mac. No original line of code in the new software. This is not what you call a simple software update!

Well under $100K would be the answer. Much cheaper than you may think because a lot of the ground work has already been done.

Richard is correct that getting the dataset into a reasonable shape is a large part of the work. The current datasets have far too many errors in them for release.

I have the majority of the work done for the cheating detection software. My rankings help identify who is cheating.

The datasets contain very high level data - who sat in which chair, contract, result. It is difficult to extrapolate from that who was responsible for good/bad boards. Only very high level data analysis can be done.
Feb. 4
You are ignoring the author of this comment. Click to temporarily show the comment.
@Robert: “The malefactors have mostly escaped punishment because the ”signals“ (if that's what they were) had to be decoded and proven illicit with specific meanings to the satisfaction of the CAS or a similar civilian court.”

The Laws of Bridge are here: http://www.worldbridge.org/wp-content/uploads/2017/03/2017LawsofDuplicateBridge-paginated.pdf

73.B.2 states “The gravest possible offence is for a partnership to exchange information through prearranged methods of communication other than those sanctioned by these Laws.”

I believe the laws were changed some time ago (1963?, 1975? - some may be able to provide the specifics) so that it was only necessary to show that the partnerships were exchanging information NOT that it was necessary to decode the signals as well. Before then the Law was worded in such a way (as I understand it) that you basically needed to decode the signal.

This is a common misunderstanding of the Laws.

Unfortunately this misunderstanding made it into the WBF report on the German Doctors. Which was then used in the CAS hearing on Fantoni/Nunes. A little misunderstanding by some has been very expensive to the Bridge world. The WBF has yet to issue a clarification on the German Doctors report. This is much overdue.

The Laws do not require decoding of exchange of information to show that information was exchanged. It certainly helps, but is not required.
Feb. 4
You are ignoring the author of this comment. Click to temporarily show the comment.
@Nigel: ACBL has all the source code from the ACBLscore+ project. There is a bilateral royalty-free agreement allowing either of us to use any code in whatever derivations we want. I've used it in other projects for non-bridge related projects. Only stipulation for ACBLscore+ would be that the copyright for the code would have to be acknowledged. Something like having “Portions of the code copyright by Hammond Software 2011-2014” in a legal disclaimer somewhere. ACBL negotiated away Copyright during contract discussions in return for $$$. Post ACBLscore+ contract, ACBL wanted to have that disclaimer removed, own the copyright, and remove nearly all rights my company had to the software, including them owning any derivates works my company might create for other industries in perpetuity. I said no. ACBL outside lawyers then them they could not use the software without the Copyright. The story has been repeatedly told.

I've taken the ACBLscore+ code, renamed it Bridgescore+, added a lot of cheating detection code.
Feb. 4
You are ignoring the author of this comment. Click to temporarily show the comment.
Let's complete Robb's fable. Pure fiction of course. Hofmeister has some prior with the tiddly-winks League which is not well-known. On two separate, previous occasions the tiddly-winkers reached out and enquired about some of Hofmeister software with the potential offer of future work. In both cases, the tiddly-wink League took the ideas and information and then used them elsewhere leaving Hofmeister high and dry. Third time around, Hofmeister is a little less forgiving. Happen to share all work, but not going to provide all the gory details so that the tiddly-wink league steals all the ideas, unless there is a contract contingent on mathematicians verifying the software works.
Feb. 4
You are ignoring the author of this comment. Click to temporarily show the comment.
See http://www.detectingcheatinginbridge.com/selections.html Chapter 45.1, “Costs of Going After Cheats”
Feb. 4
You are ignoring the author of this comment. Click to temporarily show the comment.
I don't know the first cracking example you spoke of, because I think you said it was published on a Polish site. There were lots of people posting thoughts about BZ in 2015. It was difficult to investigate all the claims.

I have met one of the two people behind the second cracking. but he/she credited the other with the majority of the work. I know that one of them occasionally follows the various BW posts. All I know about the second crack is what he/she has told me. Great story. Hope they tell it at some point. It was a lot of effort.

I have no idea the poker skills of either of the two.

The YouTube video was posted by from a one-off account. The cracker(s) wished to remain anonymous.

I will suspect from your post that you know who they are. I hope at some point that they are able to come forward so they can be publicly thanked for their effort.
Feb. 4
You are ignoring the author of this comment. Click to temporarily show the comment.
http://www.bridge.dk/2131/Klub1/SuitCombinations.pdf
Page 29. #56. (Or search for the 7th and last instance of “A J 8 x”)
It says play 10 then low to J.
Feb. 3
You are ignoring the author of this comment. Click to temporarily show the comment.
“I have shared the names with ACBL and WBF.”

I should have written, “shared some of the names”.
Feb. 3
You are ignoring the author of this comment. Click to temporarily show the comment.
@Nedju: Vastly different post 2015.
See http://www.detectingcheatinginbridge.com/selections.html
Take a look at the Chapter 13 selection. Look at the chart. Look at the data pre 2015 and the post 2015. Look at the pre-2015 data with and without known cheating pairs. The high level game is a lot cleaner after 2015 than it was before.
Then take a look at the Chapter 36 snippet.
The facts in Chapter 36 were correct when the book was written; pre-2015 data isn't going to change. As there is more data post-2015, some pairs appear to have reverted.
Feb. 3
You are ignoring the author of this comment. Click to temporarily show the comment.
“Have you shared your evidence with the ACBL Recorder? ”

I have shared lots of information with ACBL. What I have shared isn't for public knowledge.

“The Anti-Cheating Commission?”

Same. Though there is a really funny back story with them, I'm sure I'll tell it at some point. Maybe second edition of the book. Make your guess about would be really ironic :-)

“Do you think the current system is working?

No.

How many pairs have been caught by the authorities since BZ?

”To imply that “most of the top pros know who they are” is very disturbing to me". Not implying. Stating.
Feb. 2
You are ignoring the author of this comment. Click to temporarily show the comment.
“Nic, ” - I prefer Nicolas.

“statistically there are still some cheating pairs at the top level.”

Yes. I also know the areas they are cheating in.

The last set of videos I was shown, which were for a specific pair I named before the event(s), were not good for cheating detection. Those placing the camera(s) don't know where to place them to detect cheating. If you cheat, and know where the cameras are, it is very easy - trivial in fact - to mask signals from the camera view. If you know anything about magic, or camera placement, you will know what to look for. (Disclaimer: as a teenager I was on the Paul Daniels Magic Show in the UK, full day of rehearsals so I learned a lot about camera placement!)

Shameless plug: the statistics details are in my book: http://www.detectingcheatinginbridge.com/ There are snippets from the book on the web site.

I continue to maintain the statistics. I know which pairs stopped being as good after 2015, though perhaps they just aged and lost their skill level and 2015 had nothing to do with it. Without videos on pairs before 2015 it is hard/impossible to prove anything. However, with statistics, it was clear there were more pairs before 2015. You can see in the charts in the book.

I also look at statistics post 2015. It is clear that there remain cheating pairs at the top level. I have shared the names with ACBL and WBF.

“Most of the top pros know who they are. is quite the statement.”. Nah. Chat to them. They may not tell you, but they have told me. It helps to have the statistics confirmed with personal comments.
Feb. 2
.

Bottom Home Top