Thursday 22 February 2007

DBT9: Homogeneity is Bad!

A homogeneous bid is good for giving immediate information to partner. However, you must also make proper use of your ability to refine the description of your hand later.

It would be very silly to play a system in which no hands were ever opened 1S. After all, we only have a very small number of opening bids to work with, and not to use one of them would be a terrible waste. That much is obvious. But in a similar way, suppose that in your system there are no hands which, after opening 1C, would rebid 3S after the auction started 1C : (1D) : pass : (3D). Surely that is wasteful as well. In order to make maximum use of the space available to you, you would really want to make sure that not only are all the opening bids given meanings, but also all the rebids as well.

This is why homogeneity is not always a good thing. If you tell partner all the important information with the opening bid, then that's great for helping him make an immediate decision, but it leaves you with nothing else to say later.

Of course, it is not always possible to make full use of opener's rebids. If you expect partner to pass your opening bid a large proportion of the time (as for example with a 1NT opening), you cannot expect to have the chance to rebid. So the rebids cannot be defined. Having this sort of opening bid in your system effectively decreases the total amount of space available to you. This may be a problem that you are willing to accept (I quite like natural 1NT openers myself), but it is still a problem.

The difference between one-bid and two-bid hands comes up again here. Opener's high-level rebids can only be used on two-bid hands. So, if you want to make full use of the space available in each opening bid, then these opening bids must each contain a fair number of two-bid hands.

Limiting the strength of your opening bids is bad from this perspective. An extreme example would be a 1H opening in a Strong Diamond system, which might show about 8-12 HCP with 4+ hearts. Hands in this strength range are hardly ever going to be two-bid hands. So after any reasonably high-level competition you would have to make disproportionate use of pass. The other rebids may get used very occasionally - and would therefore be very descriptive when they do come up - but the vast majority of hands will not be further described. To put it another way, there will be fewer sequences in your system available for the genuine two-bid hands (because none of them start with 1H), so these will not be described so well.

In contrast, a "standard" 5-card major opening, with a range of 11-21 HCP or so, makes much fuller use of the two-bid sequences (arguably too much). The beauty of this bid is that it is on the one hand a nice homogeneous bid (partner knows about the 5-card suit immediately) but there is also plenty of potential for making descriptive rebids.

It's important not to confuse this with arguments about the frequency of the opening bid. Indeed, trying to analyse systems by looking at the frequencies of their opening bids is very unreliable. For example, suppose you were to play a 1C opening showing a balanced hand with 10-13 HCP. Then this has a frequency of about 15% in first seat, much higher than standard systems which typically have 1C openings with a frequency of 10% or less. However, this 1C consists of a single hand type with no two-bid hands, and so hardly any of opener's rebids are utilised. So we might say informally that "you are not opening 1C enough", but really this argument is not about pure frequency at all, rather the number of different hand types in the opening. You could easily add, say, all 20+ HCP hands to this 1C opening. Indeed this could result in quite a respectable system, as we shall see later.

Wednesday 14 February 2007

DBT8: Homogeneity is Good

Let's consider a particular opening bid, say 1H. A bidding system will typically require somewhere between 5% and 10% of hands to be opened 1H in first seat. Of course, if you were to choose 5% of hands at random, they wouldn't make a very good 1H bid: we have to choose them so that the opening bid provides useful information to partner. Particularly useful is if all the hands which are opened 1H share a property (or properties) in common, as then partner will be able to make deductions without having to wait for further clarification.

For example, suppose that our 1H bid promised exactly five hearts. Then if partner holds an average hand with four hearts, he can deduce that it is likely to be safe to compete to the three-level in hearts. This is based on the Law of Total Tricks, which admittedly is far from infallible, but is a good guide when you know exactly how many cards your partnership has in the suit.

It is fairly unusual to have a bid which shows exact length in the suit. More common would be a bid which shows five or more hearts. In this case, partner would still be justified in bidding to the three-level on an average hand with four hearts, but now if opener turns out to have more than five hearts there is a fair chance that bidding only to the three-level would be under-competing the hand. Similarly, if the 1H opening could occasionally be only a four-card suit, if responder still bids to the three-level there is a danger that this will be too high. Of course, there are many other factors which affect how high we should be competing, but trump length is very important, and so the more specific the information responder has about suit length, the better his chances of getting the decisions right. This is just part of the obvious general principle that a homogeneous bid - one where all the hands opened with that bid have some property in common - is good for providing information to responder. And the more specific the property is, the better.

Let's think a bit more about this 1H bid showing length in hearts. As mentioned already, in most systems the bid does not show exact length: rather there is a range of possible lengths. In this case, there are two properties of the opening bid which become important:

  • The minimum length promised by the opening bid; and
  • The average length promised by the opening bid.

For example, in Acol, where the minimum length for a 1H opening is 4 cards, the average length is close to 5. (I'm deliberately being a bit vague about what I mean by "average", but usually this would be interpreted as the arithmetic mean.) But other systems can have very different numbers - even if you consider only 4-card major systems, there is a lot of variety in how often 1H is opened on a 4-card suit.

This poses a problem for responder. If you are in a position where you have to make a decision without further help from opener, then you will have to decide - do you play opener for the minimum length, or for the average length in hearts? Playing opener for his average length will be right most of the time, but if he turns up with fewer cards than the average there is a danger that the partnership will be too high, or even playing in the wrong denomination. Whereas, if you anticipate opener holding his minimum length, you may very well be under-competing. There is not really any good solution to this, you are forced to hedge your bets.

But this is much more of a problem for some systems than for others. The key is this:

It helps to make the minimum length for your bids
as close to the average length as possible.

This is how you avoid responder's problem of what to play for.

And this explains the big advantage of playing 5-card majors.

5-card suits are much more common than 6-card or longer suits. So when you open 1H or 1S showing 5+ cards, the average length is still close to 5. The distinction between minimum length and average length therefore does not worry responder: he can happily play opener for a 5-card suit. From the point of view of homogeneity, a suit opening which promises 5+ cards is about as good as you can get, short of showing the exact length.

4-card opening bids are much worse in this respect. Assuming that you still open 1H or 1S when the major suit is five cards long, the average length for these opening bids is going to be closer to 5 than 4. This means that responder will find it difficult to judge hands with support for the major. (Of course the corresponding advantage that these systems have is that you get to find major-suit support more often in the first place.)

The worst possible case is if you play 4-card majors, always open your longest suit, and tend to open the lower of two 4-card suits. Then virtually the only time you open 1S on a 4-card suit is when you have precisely 4=3=3=3 shape. Here the average length is over one card more than the minimum length. Indeed, I would hardly say it is a homogeneous bid at all: it basically shows either a 4=3=3=3 hand or 5+ spades, and this is very bad news for responder since the first hand type is not particularly great for playing in spades at all, whereas the second type is excellent for spade contracts. How is he supposed to know what to do with spade support? I hope you would not be so foolish as to play this system - the balanced hands are so rare that it's much better to take them out of 1S somehow and guarantee five cards as the minimum. If you're going to play 4-card majors, you need to open them on four cards as often as you possibly can. This brings the average suit length down much closer to the minimum. This is another good reason why if you are playing a strong NT, it makes sense to open 1H or 1S on minimum hands with a 4-card major even if holding a longer minor.

The same principles apply when we are considering the strength promised by an opening bid, rather than the length. That is, we would ideally like to have a precise description of strength, but if forced to use a relatively wide range, we really want the minimum to be as close to the average as possible. For an opening bid which promises 12+ HCP or thereabouts, this usually happens fairly automatically, since hands with higher HCP values are less frequent. But it doesn't always work that way. For example, if you play a short 1C opening and a weak NT, the 1C opening is very often a "strong NT" hand type. This means that the average strength (perhaps it is more helpful to think of the mode here) is not so close to the minimum. Of course this analysis is rather too simplistic, as there are some definite advantages arising from the "multi-way" nature of this bid - opener is known to have either real clubs or extra strength - but having the most common hands so far away from minimum strength should still be a cause for concern.

Saturday 10 February 2007

System Regulations

As you can see from this blog, I am fascinated by bidding theory. However, when it comes to actually sitting down at the table and playing the game, I do not believe that people should be allowed to experiment with unusual methods without restriction. System regulations are a necessary part of the game.

In an ideal world, two things would happen:

  • People would be allowed to play whatever methods they thought were best; and also
  • People would have the opportunity to come up with the best defences to all of their opponents' methods.

However, in practice these two things are incompatible. In most events, players do not find out what their opponents' methods are until just before they begin to play, and if these methods are sufficiently unusual there is not enough time to discuss an adequate defence.

So if the two objectives are incompatible, what do we do? The anti-regulation lobby would have us believe that freedom to choose our methods is the most important principle, and that opponents should try to cope with this as best as they can. I do not agree with this: if forced to choose, I think the right to be prepared for opponents' agreements is the more important. This may well involve "generic" defences which work against a variety of possible conventions. However, it is not reasonable to expect people to have generic defences for everything the opponents might want to throw at them.

Fortunately, system regulations offer a compromise. The people whose methods are disallowed might not see it as a compromise, but really it is:

  • Players may only play methods which are permitted by the system regulations; but
  • Players must be prepared to defend against all the permitted agreements.

From this we get the important principle of what the convention regulations should be:

The permitted methods should be precisely those things which it is reasonable to expect pairs to be prepared to defend against, given the amount of time available for discussion.

So for example, in a duplicate pairs event, there is virtually no time for discussion at all, so the permitted methods should only be those things which players are expected to already know their defences to. Whereas, in a knock-out teams event where players can find out their opponents' systems in advance, there can be a much wider range of agreements allowed. Of course it also depends a lot on how good the players involved are: there can be much greater expectations on players in a top-level event than those in a club event. So the principle of having multiple "levels" of system regulations, depending on the event, is a good one.

So, I am very much in favour of restrictions, provided that they are for the right reasons. The right reason for disallowing a convention is if it is considered to be so unusual that players would not know how to defend against it. Note that the important question is whether the opponents would know what their bids meant. There are some conventions which make life difficult for the opponents simply because they are obstructive - this should not be a matter for system regulations: provided that the opponents can understand the methods, they should be allowed. What really annoys me is when there is a convention which is not difficult to defend against (in terms of the opponents having agreements about what their bids mean), but which is disallowed for some other reason. This should not happen.

The other big problem with convention regulations is when they are poorly written. It is very important that the regulations are clear and consistently applied, since players must be able to know whether their conventions are allowed before turning up at the event. The worst possible thing is when different TDs have different views on what the regulations say. Fortunately, apart from a couple of well-known issues, the system regulations here in the EBU are largely free of serious problems, though there are a few little corners that I wish the L&E would get round to tidying up.

The EBU's system regulations are unusual in that they are extremely detailed. But in terms of how much is permitted, we seem to be fairly near the average. (North America, in particular, is much more conservative, whereas countries like Australia are more permissive.) In fact I would say that the EBU has got the level almost exactly right. If anything, the recent decision to use Level 4 in the EBU's competitions seems to me to allow a bit too much for duplicate events. I think it would be better for the game if the standard level for tournament play was somwhere between level 3 and level 4; the rest of level 4 would be allowed in teams events provided that those conventions were disclosed in advance.

At the very highest level, such as the world championships, I do think that agreements should be unrestricted. My argument here is simply that these events are planned so long in advance that players would have enough time to prepare for all of the opponents' systems. In other words, the "ideal world" scenario really ought to be achievable here. Some players might still complain that even with all the time available it is still not possible to develop defences which they consider to be good enough, but I think there comes a point where the organisers have to say: perfection might not be achievable, but there is enough time to be adequately prepared.

Friday 9 February 2007

I've Learnt my Lesson

I can usually be relied on to have an opinion about any convention, whether I've actually played it or not. But I feel particularly strongly about the ones that I have played, and decided that they aren't any good. Here are some conventions that I would never agree to play again.

Kickback

Do not, under any circumstances, agree to play this convention. The idea is that you use the bid above 4 of the trump suit to ask for key-cards. However, if this bid is a suit, it's often far too difficult to distinguish it from a natural bid in that suit. You could have detailed agreements to try and ensure you always knew which was which, but no matter how good the agreements are there always seems to be more potential for confusion. Besides, I don't rate key-card asks as being particularly fundamental to slam-bidding, certainly not enough to warrant having all this extra space given to them.

Rusinow (leading 2nd from honour sequences)

Most of the time this is OK; however there are some situations in which you will want to lead unsupported honours, and then Rusinow doesn't work so well. You could agree to make exceptions, for example in partner's suit. But defining these exceptions carefully enough is not so easy, and the supposed advantages of Rusinow over standard methods seem to me to be far too tenuous to warrant all the effort.

Raptor 1NT overcall (showing a 4-card major and a longer minor)

I actually play this at the moment with Mark, but I wouldn't want to play it with a new partner. The problem is obvious - what do you do with a hand suitable for a natural 1NT overcall? Well, either pass or double, presumably, but I really dislike both of these options.

More generally, I really hate playing methods which give me difficulties on hands which "standard" systems would have no problem with. It's incredibly frustrating to be playing some clever gadget, and then have to admit that if you were playing standard methods you would have had no problem at all, whereas your gadget has made the hand unbiddable.

So my policy is to only play gadgets if they come at little or no cost. It doesn't matter if they have compensating advantages - if a convention puts me at a clear disadvantage compared to standard methods on a common hand-type, then I don't like it. That's why I won't play Raptor.

3NT opening showing a 4-minor pre-empt

The more traditional "Gambling 3NT" is OK, but when the minor suit need not be solid it doesn't work so well. There is more variation in the hand types, and so continuations are more difficult. Partner might not even be able to tell which suit is held. The convention would probably work well enough if you had enough time to discuss all the follow-ups and pass/correct bids. But for such a low-frequency opening you probably don't want to put that amount of work in.

For similar reasons I dislike any high-level multi opening. Even playing the multi-2D, which I've been using regularly for a long time and comes up a lot, I worry about all the unclear sequences that it generates.

Gerber

OK, maybe this one's too obvious. We all know Gerber is a bad convention. But I agreed to play it with my partner in the local league because he really wanted to, and I wanted to keep my partner happy. What I learnt was:

Never agree to play Gerber,
even if it's just to keep partner happy.