Set of problems and articles I posted on rec.games.bridge in early 1994 about pattern recognition and bridge bidding tactics.

Problem #1

AKQJ865
AT3
97
3

IMPs, Not vul vs vul., 3rd seat.

Partner opens 3 in 1st seat, aggressive bidding style, RHO passes ...

In 1975, I played the then world champion computer chess program, operating in those days on a mainframe, and running very late in the night. I employed a strategy for human vs. machine: first, I took it out of the "book" [or known] openings, and then I launched a king side blitz, but only castling on the queen side later in the middle game. The latter was designed to confuse the position ratings of the computer (for each position it determines a "score" to reflect what it thought was the advantages and disadvantages of each player).

Chess programs have mostly consisted of a rule-based artificial intelligence approach. A set of rules is designed to evaluate a position. Good chess programs have a decent set of rules that can make the determination in the appropriate time frame. The strategy against these chess program is well known - simply "confuse the cat" - make the rule set evaluate the position incorrectly. In addition, do not let the program use any of its stored knowledge of openings, by taking it out of known openings.

Bridge players are pattern recognizers. This has its advantages and disadvantages. For example, expert bridge players often make good "futures" buyers, as they can spot patterns and find trends in the buying and selling of commodities faster and easier than most people. However, bridge players often expend considerable time and resources building up their empirical database of subject knowledge for pattern matching. One expert told me when I took up tournament bridge in the late seventies that it took fifteen years to become an expert. I have wondered since then whether is it not the number of years, but a certain number of hands that must be played to build up the database to a sufficient level.

Bridge questions are often a matter of pattern recognition. Bridge magazine bidding problems are usually hands that don't fit a particular pattern. One then learns what experts would do when they must find a bid for an awkward hand. This information is then "stored" in the database.

What bidding strategy should one employ against "pattern recognizers"? What I would like to suggest here is that you want to place your opponents in situations outside the scope of their empirical database, in which patterns are either hard to recognize or are perhaps misapplied to the situation.

Against inexperienced players (i.e. with a small database of bridge experience) this is easily done by an active style of opening light, preemptive raises, lots of overcalls etc. Against poor pattern recognizers (i.e. bad players) it is best to just let them be. Against expert players, it means from time to time being creative, by inventing new situations.

This last week I posed a bidding problem on the newsgroup:

AKQJ865
AT3
97
3

IMPs, Not vul vs vul., 3rd seat.

Partner opens 3 in 1st seat, aggressive bidding style, RHO passes ...

Answers from the net, were in three groups:
a) bid game, usually hearts;
b) preempt the opponents with a 5 bid;
c) make an asking bid of some sort.

Note first how players interpreted the hand differently, each concluding it fit, in some way, a certain type (i.e. we have game, we may have slam, or we need to preempt as they have game and maybe slam). I suspect one reason for the different answers is the lack of experience in holding this hand of this type after a three level opening bid.

The follow up question of 3-Pass-4-4NT(minors); Pass-6-?, did not find much enthusiasm, maybe because the players that wanted to say they would pass or double and lead a spade, had seen the pattern before of a subsequent post showing how this would result in the minor slam making. As it actually happened, LHO (the 4NT bidder) had a 2-0-5-6, RHO had a 2-4-2-5, and opener, the aggressive 3 bidder had 2-6-4-1, with bad diamonds and clubs. If the bidding had gone 3-Pass-4-4NT(minors); Pass-5-Pass-6 I wonder if opener would have found the spade lead with or without a double?

Playing against strong opponents, after it has gone 3-Pass-?, one can expect that a 4 bid would only "kick in" their pattern recognition of bidding over preempts - even though you are bidding four hearts to make! A more creative approach may be to first pass three hearts (as it actually happened, if 3 is passed out, it will score more than six clubs, undoubled down one). If as expected, the opponents reopen the bidding, a number of interesting situations may arise. If the bidding goes 3-Pass-Pass-4; Pass-4, one can now call 4. This will make pattern recognition difficult even for strong opponents. You introduce many chances for the opponents to go wrong, including doubling you in a game that will often make (they may double you expecting you to run to five hearts). If the bidding were to go 3-Pass-Pass-Double; Pass-3NT, one will welcome this development as well. If the bidding were to go 3-Pass-Pass-4NT, you may have not lost anything. And if the bidding were to go 3-Pass-Pass-3NT, you may double if it gets back to you and wait for the heart lead.

If the bidding problem I posted appeared in a Bridge Magazine, an answer of pass would not score high. I feel this is because the pass does not fit any particular recognized pattern or tactic. On the other hand, bidding 4 does, as well as 5, as does making an asking bid or psyching a minor suit. But a tactic that bridge players may do well to recognize is when one can create a situation where the opponents have no experience to guide them.

Problem #2 Matchpoints, vul. vs. vul, first seat, you hold:

Q8763
---
Q654
Q932
You Pass, and the bidding continues Pass-Pass-1.
You play Standard American, with weak two's fairly undisciplined in 3rd seat. Your call?
-
In a previous post I discussed how bridge players use pattern recognition. One problem with pattern recognition is that it sometimes dismisses opportunities for creative bids and plays.

Say, playing near the end of a close match in a long knockout event (IMP scoring), you hold:

KT96
32
J9865
Q3
You are not vulnerable, the opponents are, and LHO is the dealer.
The bidding goes 1-Pass-2NT, where 2NT was Jacoby, a forcing raise in hearts.
Your bid? Or put another way, how would you describe a bid of 3?

Kaplan, writing in the Bridge World, described the bid as "brave". Perhaps he would have used another word if it hadn't worked. It was made by Eric Rodwell in the semi-finals of the Montreal Vanderbilt in 1985, and kept the opponents out of slam.

Matchpoint events are a fertile field to make creative bids. If the number of players is large, you will need an extremely high score to win. From time to time, you will see chances to make things happen. If you view the probability of obtaining a good board by taking a certain action as high, and the risk relatively low, you should do it.

I posted this problem earlier this week:

Matchpoints, vul. vs. vul, first seat, you hold:

Q8763
---
Q654
Q932

You Pass, and the bidding continues Pass-Pass-1.
You play SA, with weak two's fairly undisciplined in 3rd seat. Your call?

The vote was overwhelming for a Pass over 1, which is not surprising as this is what good pattern recognition should recommend. The 1 overcall is just plain awful. Other bids are even worse.

But instead of using pattern recognition, let's do some thinking. Partner is not likely to hold six or longer hearts. So the opponents have a heart fit. As the opponents have not bid hearts yet, the likely fits are 4-4, 5-3 or 5-4 (LHO with 5). In addition they have at least near game values, and are most likely to have just this. Where do we want the opponents to play? In hearts, and as high as possible, for the bad heart split will cause them considerable misery.

If we pass what are some bad auctions we don't wish to have heard? Try:
Pass-Pass-Pass-1; Pass-1-Pass-Pass and even if one balances the opponents stay low.
Pass-Pass-Pass-1; Pass-1-Pass-1NT; All Pass and the opponents stay low.
Pass-Pass-Pass-1; Pass-1-Pass-2; All Pass

How can we encourage the opponents to bid hearts and at a high level? The recommended bid is (drum roll please) 2! The opponents' first priority in bidding will be to locate any heart fit, and this they will do in most cases. In addition they will often reach game or sometimes even slam, with auctions like:
Pass-Pass-Pass-1; 2-3-Pass-4 or
Pass-Pass-Pass-1; 2-Double-Pass-3; Pass-4 or
Pass-Pass-Pass-1; 2-Pass-Pass-Double; Pass-3-Pass-4

Of course sometimes the 2 bid will not work (e.g. 2 is doubled, or the opponents reach 3NT and even with the bad heart split and without the minor suit queens, it makes) and partner, examining the hand from a pattern recognition perspective, will not describe the call as "brave". In fact even if this bid is successful, partner will seriously question whether the 2 bid is a sign of playing with a bad pattern recognizer.

However if you want to win in large pair events you need to take chances from time to time. Would you take a 75% play for an overtrick in a matchpoint event? Would you make a bid that has 70% chance of scoring very well, 20% of breaking even, and 10% of scoring poorly? If you want to do consistently well, don't take many chances - your opponents will hand you good results. If you want to win, take chances when you see them.

Is this a "crap shoot" style of playing bridge? Yes, you are rolling the dice when you see the opportunity, but you are playing with your loaded dice. Opponents will be amazed how "lucky" you are, and what you got a way with.

The setting - a sectional tournament. You have been drinking, and your opponents can notice. On the first of two boards against a husband and wife pair, your partner passes in first seat, RHO, the husband, opens 1 vulnerable, and you, not vulnerable, jump to 3, preemptive, with only KQJxx and nothing outside. This has the effect of steering them by the cold three notrump into four of a major on a 4-3 fit down one. There is considerable hostility about your 3 bid.

On the next board, RHO, the husband opens 1 in first seat not vulnerable, and you hold, vulnerable:

KQ3
762
---
AKQT965

Your bid?

This took place a number of years ago, when sectional tournaments were like small-scale regionals, instead of large-scale club games. We were playing in the gym of a large complex of connected buildings, and the table where the husband and wife were seated was next to an exit door. If one when out the door, it was about a good ten-minute walk to go around the complex to the other side, where the main entrance to the gym was.

When I held the hand, it seemed a good idea to try three clubs. The wife and my partner passed this, and the husband reopened with a double, which was passed out. The wife led a spade to her husband's ace, and when she couldn't ruff the spade return and her club jack dropped under the third high club, I made my contract.

The husband was irate at his wife: "why did you pass the double?" She silently glared back at him, as if to say, "why did you think?" The husband ignored the telling stare and repeated his question loudly, but his wife remained fiercely silent. The husband then got up and stormed out the exit door, which could not be opened from the outside. I went and got another beer from the bar, and when I returned, as the next round was beginning, the wife was still at the table and there was no sign of the husband. We had started at the next table when there was a soft set of knocks from the exit door. The wife remained oblivious of the knocks.

Then a set of stronger knocks, which only I seemed to hear. Finally I excused myself from the table, and went to open the door.

Tournament bridge players are excellent pattern recognizers. This talent is not only used in ordinary ways like judging what bid to make with a hand or determining how to play a card combination, but is also used to assess what the opponents are likely or not likely to do. To play against pattern recognizers, one must be careful not to fall into known common patterns and not to establish any personal pattern of behaviour that can be taken advantage of. Known common patterns include (these lists will be far from inclusive):