Lead analsysis

This set of instructions is for those that have been sent a URL with a list of link to leads to evaluate for a possible cheating case.

Watch the following videos:
https://youtu.be/x4Sq1ElcvQc
https://youtu.be/_eEQh8XXRh4

The user "KIT" is in these videos is Kit Woolsey. There is no indication he is cheating; he is used as a real world example and I thank him for allowing his leads to be used.

Update

February 22, 2021 - The lead data now includes three columns, "BBO1", "BBO2", "BBO4". These represent the data that only the person on lead can see "BBO1". For collusive cheating the data for the opening lead (OL) and their partner is shown in "BBO2". For self-kibitzers, the data for all pairs is shown in "BBO4". If you are doing an analysis, pull up the data in BBO1 first. Using only the information from the auction, select the lead. Do not look at the actual lead. For collusive cheating cases, then look at "BBO2", select your lead, do not look at the actual lead. For self-kibitzing cases only, look at "BBO4", select your lead, do not look at the actual lead. After you have selected your lead, or possible leads, then look at the lead that was selected. Compare to the lead you selected after BBO1 and BBO2. The suggestion is to do, say, 100 leads looking only at BBO1, then 100 leads looking at only BBO2 and hope that you have forgotten the lead that you would have selected.

Analysis

If are reading this because you have been sent a link to a player or pair's lead data, then

  1. Please keep metrics on how long it takes you. This is important for future work by others.
  2. Find out if this is suspected as collusive cheating or as self-kibitzing. This affects how you look at hands.
  3. Please watch the videos before you start!
  4. Please remember that the values/description in the spreadsheet you downloaded:
    ARE SUGGESTIONS ONLY TO HELP SPEED UP THE WORK! THEY ARE NOT DEFINITIVE. AND MUST BE CHANGED IF YOU DISAGREE.
  5. A recommendation from others is that you review, say 50 boards, before making any evaluations. This will help you understand their bidding methods, carding methods etc. Then start from the beginning. If you start to edit immediately, you will probably find that you have to go back and redo the work and it will take longer than first watching 50 boards without doing anything. I know this is frustrating; take notes in pencil if you need to, and then go back and write up definitive notes in the Excel spreadsheet.
  6. Almost certainly for some of the boards, particularly ones that the computer assigned a "1" or "5" will need to be changed. This is the difference between double dummy and single dummy leads.
  7. It is OK to occasionally underlead an ace; there are some known "guidelines" for when to do it. The computer does not know this. Every underlead of an ace is marked "1" or "5" It is up to the human reviewer to assign the correct value. Keep every underlead of an ace as either a "1" or "5". This is important for review in committee; they will want to review ALL underleads of an ace. Some pairs may play Rosenkranz doubles making an underlead of an ace more common than other pairs.
  8. If you notice something obvious, that you believe that the computer algorithm incorrectly assigned a value, let me know and I will update the algorithm. The algorithm is constantly being improved.
  9. Some things like quality of suit for two level overcalls may be partnership understanding, so underleading an ace after a weak two can be understood if the partnership standard is one of the top two honours for a week two. Therefore if you see some, record which board, so you can then later review for different style/quality of suit.
  10. Some pairs may play Rosenkranz doubles making an underlead of an ace far more common than other pairs.
  11. For some pairs, you may want to create a column which is "underlead of ace" so that you can later reference just those. There are so many of them (not all work, some are clearly anti-cheating hands), that they need to be referenced for separate review by committee, i.e. the committee can just go to those hands.
  12. If you get to about 50-100 boards and think they are not cheating, stop.
  13. If you think something suspicious may happen, follow the card play for a few tricks. No need to do with every board. Some hands look like something weird may happen. Should only take a few seconds. If it does, write it up in a separate column. More material for the disciplinary committee, the easier their job is.

The assignment of lead values should be done as follows:

  1. Bring up the deal. As soon as the deal is displayed, click on the name field of the person on lead. This will hide the other three hands.
  2. Review the auction.
  3. Decide which card you would lead. If this card is led, then this will be scored as a "3". It will be rare to change this value. If however, you have two possible leads, for example AK32 AK32 432 32 on lead against a 1NT all pass auction, then you might lead the 2 of spades or the 2 of hearts. If there are two choices, you should probably assign a "2" or "4" after you have looked at the other hands.
  4. Bear in mind the pair's lead agreements. For example, if they lead A from AK, or K from KA, or lead third/fifth, or low from a doubleton, or coded 9s and 10s. This is the reason that you should watch several boards before starting to record values.
  5. Decide which other possible cards you might lead or would consider reasonable. These will be scored as "2", "3" or "4".
  6. Any other lead MUST be scored as either a "1" or a "5". The next person reviewing may only look at "1" and "5". It does not matter how the lead worked out, you MUST score it is "1" or "5" if this is an unusual lead. NEVER mark an unusual lead as a "2", "3" or "4".
  7. Click on the name of the person who is on opening lead again. This will redisplay all four hands.
  8. If this is a collusive case, try and look at only two hands. The person on lead and their partner. Assume they are sitting on a couch next to each other and CANNOT see the other two hands. Decide what card you would lead only look at those two hands. If the lead is DIFFERENT than the lead you would have selected, but is a card you considered or you think a reasonable person might choose, you will choose either "2" or "4". Look at the lead made. If the lead made is the lead you would have selected when you only saw one hand, this is a "3". If the lead is one that you would have selected when you saw both hands, this is a "4" (possibly a "5"). If the lead is one that you would have considered when only looking at one hand, but is NOT one that you would have selected when you saw both hands, this is a "2" (never a "1").
  9. If this is a self-kibitzing case, look at all four hands and decide what lead you would want. If the final lead is one that you would have considered, score it a "2" or "4". If the final lead is not one that you would have considered, score it a "1" or "5".

Your goal is to highlight the unusual leads, so that a subsequent reviewer can immediately go to those. Always leave an unusual lead, one that you would not select or believe a normal player would select, as a "1" or "5".

What is important is not how you grade, but how an investigation will look at your data. Let's say you have marked ten boards as a "1" and fifteen boards as a "5". They will look at those 25 boards. They will then grade those 25 boards, based on their criteria, as to a cheating or anti-cheating. If you have marked all of the original "1" as a "3", then the investigation committee never get to see the times when an underlead of an ace did not work out. This is very important to them and the accused pair.

The assignment values are:

  1. An anti-cheating lead. This is not a lead that you would pick. It is a lead which does not work out.
  2. Values of 2,3,4 are used for a card that you may have chosen to lead. If it is not a lead you would have selected, it should always be scored as a 1 or 5. A value of 2 is used for a lead that is not optimal if you could see partner's hand (collusive) or all four hands (self-kibitzing). For example, there is a better lead that could have been made that would not have raised suspicions.
  3. This is a lead that you would have made, or is an obvious lead. For example, leading partner's suit or implied suit, or the unbid suit. This is not a hand that is indicative of any cheating.
  4. This is a lead that you would have considered, but is not your main lead. It is a lead that you would have selected if able to see partner's hand (collusive) or all four hands (self-kibitzing).
  5. This is an unusual lead that worked out.

Any lead of an unsupported ace in a suit contract should be considered a "1" or "5", or should be marked in some what so that an investigation committee can review at a later time. Sometimes a lead of an unsupported ace is perfectly acceptable. However, if you lead an unsupported ace and partner is void or has a singleton, this is suspicious.

In a suit contract, any lead where partner has a void.

In a suit contract, any lead where partner has a singleton and you have the ace of trumps.

Any lead from Kx is unusual.

Failing to lead partner's bid or implied suit is potentially suspicious.

Some hands may not have an obvious lead. You may have several choices. These should be scored as a "2", "3" or "4" based on partner's hand.

The two questions when you have finished the report are:

  1. Do you think this pair is cheating?
  2. If you do, do you think a disciplinary hearing would also convict?

Sometimes the answers are "Yes" and "No"; in this case it is best to stop. If you do not think a disciplinary hearing (which will put the benefit of doubt on an accused pair) will convict, you probably need to wait until there is more data.

Collusive cheating lead analsysis

If a pair is accused of collusive cheating on the opening lead, the investigation should proceed as follows:

Initial investigation

You want to look at a subset of the boards to determine if there is likely cause:

  1. Prepare a set of boards, say 100 boards where the pair is on opening lead. If necessary, anonymize the data. Only provide the hand that is on opening lead, no other hands. Give the data to an expert for review. The data should include the bidding, with explanations, but not the lead made nor any other data.
  2. The expert should record their preferred lead. The expert should also record other possible leads that they considered.
  3. The expert should turn in their work to a reviewer.
  4. The reviewer should compare the expert's leads to the actual leads made at the table. Expect there to be differences. Leads can be quite subjective; some players prefer a passive style, some active. If the reviewer determines that there are substantive differences, proceed to the next step. The decision to proceed is subjective, no guidelines are given here. Experience with knowing the expected difference in data is useful. If you are not sure, contact me with the data.
  5. The same expert should now be provided the same data but with two hands shown, the opening leader and the partner of the opening lead. The expert, without consultation to the previous data, should record the expected opening lead, the opening leads they would consider. In addition, the expert should look at both hands and decide the best lead given that the expert can see both hands. The expert should now record the lead they would make given both hands, with the understanding that the pair is cheating and does not want to get caught.
  6. The expert submits the data to the reviewer.

The initial investigation is complete. Approximately 100 boards will have been reviewed by one expert. The boards will have been reviewed twice. First, for the opening lead based on knowledge of only one hand. Second, based on the opening lead with knowledge of two hands. It is only the boards where there is a difference on the lead selected that are relevant. This has nothing to do with double dummy analysis.

The reviewer must decide, based on past experience and the data presented, it is appears that the pair is making use of additional information on the opening leads.

The reviewer can continue by making more leads available for an expert to review.

The reviewer may choose to use some number of experts (it may be as small as one) to review the boards, and to remove all boards where the lead is the "obvious" lead. This will substantially reduce the overall work involved. As many as 50-80% of the opening leads may be removed using this method.

The reviewer now has to decide how many experts to use for a full analysis. The reviewer now has to decide if all boards, or only a subset of boards are going to be examined.

The decision if a pair is collusively cheating is based on the ratio of boards where the opening leader always happened to pick the correct lead when given a choice. This is a subjective decision. It is not something that can easily be thrown into a statistical model.