Most of these are taken from my answers to real world questions or posts on various Bridge social media sites.
I am currently working, or have worked with, about ten different National Bridge Organizations (NBOs).
First: do NOT send me an email of who you think is cheating!
My first step is to get data from the NBO, store it, process it. For this I need to know either the BBO Host name, e.g. "NBO_1", "NBO_2" or the common titles used by the NBO, e.g. "NBO Monday Pairs".
After I have processed the data I will ask you to send between 5-10 BBO handles of players from your NBO. This is to avoid confirmation bias. I will reply with a short, private, email that describes the data that I have on each of those players. If a player is suspicious, I will let you know.
Assuming that the suspected player/pair is identified, we have avoided, partially, the first step of confirmation bias.
I will then provide information on how best to proceed.
I strongly prefer using anonymized data. I will create a private URL of the player/pair under suspicion, and also create similar anonymized data on a player that is not under suspicion. Let's say P1 is the player under suspicion and I have data with P1 playing with Q1. You also gave me P2 who is not under suspicion who has played with Q2. Let'e say P1/Q1 have played 350 boards and P2/Q2 have played 400 boards. I would split the data into approximately 100 board "chunks". The data would be anonymized. There will be about 7-8 chunks of data. These chunks would be investigated using experts of your choice.
For some cases like this, particularly if it is someone that has worked on one of the cases, I will create the data and put on a private URL for them to investigate further.
Real world answer:
There are 17 unusual leads. I have not looked at them. I don't think this pair are cheating, but look at the unusual leads and see.
For the example hand you sent, if collusive, surely they bid and not defend.
When I picked pairs for the "junior detectives" to look at, I use additional data to decide if the pair should be reviewed.
My two questions for reviewers are "Are they cheating?" and "Would ACBL convict?". I have data that helps me answer the second question (I don't share this).
For your pair, my data says the answer to the second question is "No". Doesn't mean they are not cheating - just that if they are, it is unlikely that this would lead to a conviction.
"False positives" are difficult. How do you prove cheating? The current threshold is to ask experts to review the hands. Sometimes they are not in agreement.
I have some tests that are better than others. I generally look at the totality of the tests, not just a single one.
If you cheat, you improve your game. Suppose you go from a Flight C player when defending, but with a certain Flight C partner you defend at a Flight A level over a large number of boards. However, your raw stats DECWER/DDOLAR/DEFWER will not show you are cheating because you are not even at the Expert level. Other tests indicate.
I have some "false positives" the other way. Expert analysis stated they were not cheating. I said they were. Further investigation, including a non-coerced confession, showed that I was right. I always am careful with "false positive" and "false negative" because it depends on the perspective of what is being asked. Are they cheating? Are they honest?
In almost very case, further manual analysis is needed.
There are a rare few cases where the number of boards (say, 8000+ boards) and the raw statistics (defending better than Lotan Fisher playing with a non-expert) very strongly indicate cheating.
Or 2000+ leads with a DDOLAR over 88%.
Detecting cheating is not a black/white issue. More shades of gray. The more you cheat, the obvious it is that you are cheating.
My two questions are:
1. Are they cheating?
2. Would an NBO convict?
For some pairs the answer to the first question is yes, the answer to the second question is no.
These are not my answers. But those of the "Junior Detectives". In other words, if they are cheating, then the belief is that the NBO would need more boards to prove.
We are really talking about the level of satisfaction needed to find someone guilty.
An NBO should have a much higher level of satisfaction that someone is cheating than, say, an online invitational tournament.
Ultimately it is the Bridge logic that helps decide if someone is cheating. The statistics can certainly help to identify.
=4*(SUM(LEN(AK2)-LEN(SUBSTITUTE(AK2,"A","")))/LEN("A")) + 3*(SUM(LEN(AK2)-LEN(SUBSTITUTE(AK2,"K","")))/LEN("K"))+2*(SUM(LEN(AK2)-LEN(SUBSTITUTE(AK2,"Q","")))/LEN("Q")) + 1*(SUM(LEN(AK2)-LEN(SUBSTITUTE(AK2,"J","")))/LEN("J"))
Replace the value AK2 with the value of the cell containing the hand record.