Monday, May 20, 2013

Only the inviter should stretch

Introduction

Only the inviter should stretch. This nugget of sound advice comes from Howard Piltch, formerly one of New England's great players and teachers and now living in Indiana.

First, let's think about it for a moment. Most bridge players want to play higher-scoring contracts if they can so that there is a tendency to stretch a bit, often by both sides of a partnership. This is especially likely when we are vulnerable, although presumably everyone can see the vulnerability. Unfortunately, the result of both partners stretching is often an unsuccessful contract. Normally, the inviter knows that the partnership is within two (or three) points (high cards and distribution) of being able to bid game when issuing the invitation. Often, but not always, the invitee has limited their hand to about a three-card range. [The exact numbers are not all that important.]

Here's a hand that came up recently (in fact there were two very similar hands, one at the club, one on BBO, where the auctions were identical and the result the same – the only difference was that the club hand was at teams).



Here, South ("Invitee" – moi, I am ashamed to say) thought they had prime values and accepted the game try. Unfortunately, the contract was une chute, as the French would say. There were two diamond losers and two spade losers. The "Inviter" was a BBO robot (GIB) and, although minimum, I think most people would consider this a "three-card limit raise" and bid the same. Was this pair unlucky? Yes, a little. But neither player had any useful distributional features and South knew that they had exactly eight trumps between them. Having eight trumps, as opposed to nine or more, typically leaves declarer with no recourse in the event of a bad trump split or honors in the wrong place. This is especially true of a 5–3 fit. In fact, 3NT is unassailable and perhaps I should have offered to play there, if accepting the game try.

The other similar hand went down because of a 4–1 trump split – again, no margin of error when things don't go right. On that occasion, I was the inviter and my partner the optimistic invitee. Actually, a speculative double by the holder of QJTx offside would have resulted in a 9-imp loss (somehow it went down only one at the other table).

Often it is the invitation itself that carries significant risk. It is frequently the case that an invitation is made when the bidding is two levels below game. Going to the next higher level (and stopping there) is risky because we get no additional reward for being one level higher, yet there is now a finite probability that we will not make any contract at all. Then we feel like the dog in Aesop's fables who, carrying a bone in its mouth and then crossing a bridge, looked into the water and saw another dog with a bone. He attacks the "other dog" and of course loses the bone in the process.

The invitee, on the other hand, while also taking a risk by going to game, has a much better risk-reward ratio because the game bonus now comes into play directly. Indeed, the partnership may already be too high (on account of the invitation) in which case the bump to game has very little downside (the extra undertrick is unlikely to be a big factor – unless doubled).

Having said all that, it may seem that it is the inviter that should not stretch since most of the risk, that is the un-hedged risk, is taken in that step. But if the inviter never stretched, there would inevitably be a lot of games missed. Feel free to skip the next section and go straight to the conclusion.

Detailed analysis

Let's look at some examples in no-trump contracts (because these are so much simpler). The simplest is when opener bids 1NT (let's say this shows 15–17 hcp) and partner invites by bidding 2NT. What are the risks? The following table shows the (approximate) average number of tricks you can take with the points shown in opener's or responder's hand. For the average tricks, I have used Matthew Ginsberg's analysis. I haven't adjusted for the fact that, for example, 10 opposite 15 will generally do slightly better than 8 opposite 17 because the hands are more balanced and therefore tend to offer more transportation. I have also assumed that if the average number of tricks is x, then the probability P(n) of taking n tricks is n+1x (80% where n = 8 and x = 8.2) and P(n+1) = xn (the 20% zone in this case). For simplicity, I have eliminated the possibility of taking n1 or n+2 tricks.

opener151617
resp
87.68.28.7
98.28.79.1

Below, we show the expected IMP losses playing against par and when not-vulnerable (figures in red are for overbidding and in green for underbidding). To explain some of the values, we can look at an example: we see that in the minimum case (15+8), we can take 8 tricks approximately 60% of the time and only 7 tricks the other 40%. So, with 8 hcp, responder takes a risk by inviting – scoring against par, he will lose 4 or 5 imps (depending on vulnerability) 40% of the time and break even the rest of the time. Net expected loss: 0.34 or 0.43, that's to say 1.6 or 2 imps multiplied by the probability of opener having exactly 15 hcp (44%) and by the probability of having 8 (48.5%) rather than 9 hcp (51.5%). If opener, knowing of his partner's taste for inviting with only 8 points, raises to game only with a full 17, then he will miss a making game 20% of the time (6 or 10 imps). If he also "stretches" and accepts with any 16, he bids too high 30% of the time (10 or 13 imps). When combined with the probabilities of the various holdings (15, 16, 17) we get the expectations of loss shown in the table.

Not-vulnerableopener:accept on 17accept on 16 or 17
responder:
conservative
0.27
0.19+0.55+0.71
0.55+0.51
aggressive
0.27+0.34
0.20+0.19+0.34
1.29+0.51+0.34

And the following table is the equivalent for vulnerable contracts.

Vulnerableopener:accept on 17accept on 16 or 17
responder:
conservative
0.45
0.32+0.91+1.18
1.18+0.66
aggressive
0.45+0.43
0.32+0.34+0.44
1.67+0.66+0.44

Conclusion

The following table shows the summary using approximations (for vulnerable contracts – the non-vulnerable numbers are lower but the relative values remain similar).

Vulnerableopener:conservativeaggressive
responder:
conservative
3
2
aggressive
2
32/3

So, we can conclude that the best a pair can do is to have a conservative player sitting opposite an aggressive player. It doesn't really matter which way around they are. "Normal" bridge is to invite with a (non-stretchy) 9 and accept with an (optimistic) 16 – minimizing losses at approximately 1.33 (not vul) or 2 (vul) IMPs. This is the time-honored strategy. Note that these expected losses are against a team of robots that peek, not only during the play but during the auction! How would we do against flesh and blood teams? Assuming that they are using their optimum strategy too, i.e. the same strategy, then we should push these kinds of boards, assuming best play and defense at both tables. If the other team has their optimist and pessimist sitting the other way, there may be swings, but they should even out in the long run. Note that the numbers have already take into account the fact that it's worth pushing for games, especially when vulnerable.

So, how does this relate to the subject of this blog? Well, we've shown one way of doing things: the inviter is conservative and invitee aggressive. This reduces the incidence of invitations but is reasonably accurate. How about when it's the other way around, following the rule "only the inviter may stretch"? It would work just as well, though more invitations means that the opponents have a better idea of what's going on. And, according to Howard, this is the right way to do it in all of those other less clear cut situations. I've always followed his advice. I hope my partners feel the same way.

There is an obvious point – that if inviter keeps quiet, it doesn't give the invitee even the opportunity of an acceptance. No doubt there are other scenarios, for example major suit game tries, where it does make sense that only the inviter should stretch. But the main moral of the story is simple: don't be aggressive – or conservative – on both sides of the invitation!

Wednesday, May 15, 2013

No-trump Preempts

One of the features that I like about the weak notrump, whether 10-11, 10-12, 12-14 (I've played them all at one time or another) is the preemptive aspect. Especially at "green" (not-vul vs. vul). It's very much harder for the opponents to come in and conduct a sensible constructive auction starting at the two-level, vulnerable because the opening bid is so well-defined and responder will know just when to apply the chopper.

But opportunities to preempt the opponents in no-trump don't begin and end with the opening bid. Take this hand from a recent team game at the club: ♠– Q754 QT2 ♣AJT843 (third seat at favorable vulnerability). Partner deals and opens 1NT (12-14) and the next player doubles. The opponents have at least a 9-card fit in spades and 17-19 hcp. Moreover, partner's values are going to be sitting under the opponents' strength. So, my hand might take a few tricks but partner will probably only score aces, assuming that the other hand can get the lead a couple of times.

In other words, there's just a little too much danger, in my opinion, of them making 620. I therefore raised partner to 3NT. This was followed by two passes and then double. Should I pull to 4♣? No, I decided to stand my ground (I had the feeling that the next player might take the double out). And so it proved. It went 4 on my left followed by two passes. Was I tempted to double? No way! That would have ruined the whole plan. The contract drifted off only one trick after I underled my ♣A at one point, allowing declarer to score her kingleton. But the damage was done. Our teammates were +620 at the other table giving us a 13 IMP swing.

As it happens, we could have made 3NT, at least with careful play and a peek (or an avoidance play) by felling that same club kingleton. Or we could have made 5♣ with the same peek (though there would be no particular reason for the avoidance play at a suit contract).

It takes a bit of a nerve to pull this kind of thing off. And you have to be ready with an apology to the rest of the team when it backfires. But the satisfaction of bringing off this sort of coup can turn an otherwise forgettable evening into something to smile about.

Monday, May 6, 2013

Cohen versus Stewart

There is an article in the May 2013 edition of the Bridge Bulletin which pits Larry Cohen against Frank Stewart in "2/1? Help or Hindrance."

While I certainly come down on the side of Cohen and 2/1, I think both sides have made a few questionable assertions.

In my opinion, one of the major benefits of playing 2/1 is that it's easier. Yes, you read that right (Cohen agrees). In fact, many 2/1 auctions are simply a question of patterning out – no thinking or judgement required. Yes, there are styles of 2/1 that require more judgement, such as requiring that a "high reverse" by opener (e.g. 1 2♦ 3♣) shows reversing strength. But most of the time, for example you just sit down with an unfamiliar partner and agree "2/1", you rarely have to worry about such subtleties.

So, I find it very strange that most (all?) bridge teachers of novices start them off in completely the wrong direction by teaching them "standard." How can so-called Standard American be considered "standard" when 80% of tournament players (according to Cohen) are playing 2/1?

I admit that I'm a little biased. When I first learned to play bridge it was lunchtime bridge at work, a close relative of "kitchen bridge." Bidding was natural. The more you had the more you bid. But within a few months I had graduated to duplicate. One of the most popular systems in those days was Kaplan-Sheinwold, essentially 2/1 with weak no-trumps.  So, I never really was "corrupted" by this "SA" nonsense.

Let's dissect SA for a moment. It's essentially a hybrid (bastard might be more a propos) of Goren (four-card majors) and Roth-Stone (five-card majors) with perhaps a bit of K-S thrown in (for example if you play inverted minors). In other words, it's neither fish nor fowl. I'll go further to say that it encourages really bad bidding because minimum balanced openings with 12 or 13 high card points will frequently be forced to open a bad three-card minor. This is also true of course for 2/1 if you're playing a strong no-trump.

But I do wonder about some of Larry's examples of troublesome SA auctions. Surely, a tenet of SA is that a two-over-one response promises at least one more bid, with the possible exception of a sequence where opener rebids 2NT. So, I can't imagine having to worry about the auction 1♠ 2 2♠ being passed out. Still, there must be some misguided souls who would pass here, otherwise Larry would not mention the possibility.

I'm also surprised that Larry should advocate a semi-forcing 1NT. Surely, the forcing 1NT (by an unpassed hand) is a cornerstone of 2/1? I have written previously (for example, The Forcing 1 Notrump) about the benefits of being able to show a game-going hand that does not want to suggest a slam by starting with 1NT. Obviously, you can't do that if 1NT isn't 100% forcing.

But these are relatively minor quibbles compared with my thoughts on Frank's contribution. He makes a comment which I believe shows a shocking misunderstanding of the 2/1 style. In an auction which begins 1♠ 2♣ 2, he says "I suspect many Easts in 2/1 would bid 2♠ 'to save space.' Well, if I were West with such mangy spades, wild horses couldn't get me to cooperate in a slam hunt opposite a tepid preference" [my emphasis].

What he seems to miss is that East bidding 2♠ after an initial 2♣, promises three card support (not a tepid preference). Not only that, it shows (at least if you play my style) a hand that will willingly cooperate in a slam hunt, and with good clubs too (else East could have started with a 1NT response and then jumped to 4♠).

He also points out that with both partners potentially making minimum bids at all opportunities, how does a partnership know how to get to slam? Again, this seems to miss the point. As soon as a fit is discovered, one partner or the other can make a control-showing cuebid. Normally, there will be plenty of space available for these types of slam tries. It's true that some delicacy is now required. I have written about this issue before in Slam Tries. One solution is to employ either serious (or frivolous) 3NT. The other solution is to discuss the situations when a control bid demands another control bid. Admittedly, neither of these approaches is covered in the 2/1 texts, but the concepts are not difficult either.

In his auction 1♠ 2 3NT, he says that one player thought that 3NT showed "a powerhouse" and the other did not. But anyone who has read any book on 2/1 knows that a jump in a game-forcing sequence has a special meaning. Given that opener's 2NT rebid shows either a minimum balanced hand or a maximum balanced hand, there can only be one possible meaning for 3NT left: a medium balanced hand (15-17). So it seems that neither Frank nor his partner really knew the basics of 2/1. You can hardly blame the system for that!

So, is 2/1 (with 15-17 no-trumps) the best system out there? No, I don't think so. On theoretical grounds, I think that both K-S (or, if you like, 2/1 with weak no-trumps) and Precision are technically superior systems. But, 2/1 is certainly a lot easier than Precision and therefore a much better candidate for teaching beginners (which no-trump range to play is more a question of personal preference and familiarity).

Having said that, I will note that I've had all my best results playing with my favorite partner – and we play 2/1 with 15-17 no-trumps.