Monday, December 31, 2012

What makes a good penalty double?

Back in April, I wrote about different types of penalty double – the second installment was Penalty Doubles (part 2). There, I observed, that the traditional penalty double was an endangered species because in practice it doesn't come up very often. Note that I am excluding what I call "equity-preserving doubles" which cover such things as doubling sacrifices. By the "traditional" penalty double, I mean the type where you double a contract that the opponents have bid willingly (and it appears that the hand "belongs" to them). Such doubles typically occur in an auction that begins competitively and, frequently, they arise after partner of doubler has opened the bidding. And note also that I am talking here about doubles that are strictly penalty in meaning, not cooperative doubles that are converted by passing.

What makes a good penalty double? I think there are several factors:
  1. lack of a fit for partner, otherwise a relatively balanced hand – in particular you must not hold a long suit (six-plus cards) of your own;
  2. trump quality and position – your tricks are sure tricks and there's no danger of tipping off declarer as to the correct line of play – but the position of your honors in the other suits must appear to be favorable to you too;
  3. an expectation of setting the contract at least two tricks – or, at match points, your judgment that the contract is normal (four or five of a minor is generally not normal at MPs) and that your peers are likely to consider doubling (if the contract isn't normal – perhaps they've clearly gone off the rails – you are almost certainly getting a good board if they go down even without a double);
  4. no place for the opponents to run.
Let's look at these in turn, although we will go in reverse, starting with point four. Several years ago in a world-wide simultaneous pairs event, I was the happy holder of ♠AK on lead against 6NT. Clearly, this pair had gone off the rails and I was likely to get a very good board simply by "staying with happiness". But the temptation for that world-wide top overcame me. I doubled, even though I wasn't sure that they didn't have a place to run. They ran to 7♣ and naturally I doubled that too. It made, of course, and instead of getting the world-wide top I was hoping for, I got a well-deserved world-wide bottom!

Point three: even though our hand might look fairly normal the other hands at the table might be goulash hands, with long suits and voids. Having a safety margin is definitely worth-while. Furthermore, the penalties really don't get interesting until the second trick (again, we're not talking about equity-preserving doubles). Let's say we are playing IMPs and the contract is 4 (vulnerable) going down one at both tables. The opponents double at the other table but we don't. We lose 3 IMPs. Not such a big deal. Now, let's say it goes down two at both tables. If we don't double and they do, we are losing 7 IMPs. Now, that's getting more serious.

Point two: it's fairly obvious that we need trump quality because otherwise declarer will simply draw our trumps and run his winners. And, equally we need to be sitting over the trump length lest our honors be finessible. Well-placed trump tricks are tricks that can't disappear. Note that if trumps appear to be 5-5 (or even 4-4) we won't always be able to tell where our honors are in relation to declarer's honors.

Point one: perhaps the best indicator of the success of a penalty double is our fit for partner (assuming that partner has bid which is typically the case). Obviously, a void is the best, but a singleton is pretty good. Why should that be? Well, whatever the number of total tricks is on the hand (and the fewer the better from the point of view of the penalty doubler), each card that we have in our hand in partner's suit is likely to increase the total tricks by one. This presupposes that partner has a real suit (one of at least five cards in length). A good fit for partner means that we might even be better off playing our own contract than defending their contract. And it increases the chance of shortness in one of the opponents' hands.

Now for some examples to help summarize what I've said. Let's take an example from a recent club game which also happened to be a world-wide simultaneous event. You hold this hand (only we are vulnerable) ♠876 T843 J5 ♣QT84. Partner deals and opens 1♠. RHO doubles and you pass. LHO bids 2♣ and partner rebids 3♠. RHO now raises to 4♣. Your call?

If you chose double, it means that you haven't been paying attention! Let's look at the factors one by one:
  1. we have an excellent fit for partner who appears to have a good six or seven spades;
  2. our trump quality is OK though nothing to write home about – but we don't even know where the trumps honors will be found – furthermore, with honors potentially on both sides we could easily be tipping declarer off to the correct line, if there is one;
  3. we don't really have an expectation of beating this two – let's say partner has 16 hcp, we have 3 – that's not even half the deck – plus we have no other surprises in store for the opponents;
  4. do they have a place to run? how about diamonds? Partner's hand is spades and more spades.
When I played this hand (as LHO) we got to 5♣ and I went down. There was no double and I mistimed the play. When Kim played the hand (also as LHO) the player with the cards shown above doubled 4♣. Not only did Kim make the contract, she actually made the overtrick (playing double-dummy).

As it turns out, N/S can take eight tricks in spades with their 18 hcp, although East has to underlead the A to get West a ruff to avoid the overtrick. E/W can take eleven tricks in clubs. That's nineteen total tricks – two nine card fits with a club void in the long spade hand. In fact, a diamond contract doesn't do so well thus there's no good runout there, but of course none is needed. See below for the layout.

Here's another example of a bad penalty double, which came up just the other day, although this time I was the happy recipient of the gift. My opponent held this hand: ♠T98 JT82 54 ♣QJ83. All are vulnerable and your RHO (the author) opens with 1. You pass and LHO bids 1♠. Partner bids 2. RHO bids 3♣ and you compete to 3. I don't recall for sure what happens next but, if I recall correctly, the opponents end up having the auction to themselves and bid 5♣. Do you double? This would be a true penalty double because the opponents have voluntarily bid 5♣ and it is their hand.

Let's look at the various factors: we have a very good fit with partner [full contraindication]; trump quality is good and the position appears to be good, although there is a possibility that declarer might be able to come up with a winning line [perhaps half an indication]; we really don't have a solid expectation of a two-trick set but it's conceivable (one heart, two clubs perhaps and maybe partner will contribute another trick) [half an indication]; might they have a better spot than 5♣, though – diamonds for instance [half a contraindication]?

On balance, this is not a good double. Nevertheless, the holder of this hand doubled and, even though 5♣ was cold (though not six), the opponents ran to 5 which partner now doubled. This made easily (although RHO – moi – neglected to redouble and make the overtrick for a clear top – another pair actually bid and made 6).

And now my final example for this particular chamber of horrors: I recently made a very poor decision to double a competitive 4♣ call by my RHO, holding this hand: ♠QT92 QJ9876 ♣T2. All were vulnerable and partner had opened proceedings with 1. RHO overcalled 2♣ and I had no good call so passed. LHO bid 3♣ and partner came back in with 3. RHO bid 4♣. At the time, I intended it more as a "suppressant" double (a species I apparently failed to mention in my earlier discussions). This type of double is used solely to warn partner not to bid again. Yet, it is made in the expectation of a set, even though you aren't sure that partner could have made his contract (thus it is not strictly for equity preservation). Therefore, it is a penalty double and at least some of the other rules need to be followed. One of the rules I broke here, not having really thought about it as a rule before this occurred, is the long-suit rule. Just as you might conceivably pull partner's penalty double when you have a long unbid suit of your own (see George Rosenkranz' excellent Tips for Tops), so you should not make your own penalty double with an unbid long suit. Even if partner has only three cards (a suit he will not normally bid on his own), there's a 10% chance that one of your opponents is void and only a 40% chance that two rounds of the suit will stand up.

As always, I'm interested to hear comments. Have I missed some aspects of what makes a good pure penalty double? I know that I double too much in practice (although most of those are "equity preservation doubles" where the rules given above typically don't apply) so it is important for me to have some good guidelines. The long-suit exclusion is the latest enhancement to my personal checklist.

Best wishes to all of my readers for 2013.  May all your squeezes and endplays produce extra tricks.

Monday, December 24, 2012

A few observations on Walsh and XYZ

I find that I wrote a draft on this subject some time ago (during 2011 I think). I'm going back to it now because I want to make a few remarks on Walsh and XYZ [although I don't mean to imply that these two conventions must go together – but as we will see, it helps].

Continuing our discussion related to opening with a (possibly prepared) minor bid in Prepared Bids (part 1). Now, Mr. Walsh comes along and makes life yet more complex. Playing Walsh, both partners agree that, holding a four-card major, there is little point in responder bidding 1 over 1♣ unless he has at least invitational strength. If responder has a weak hand and a four-card major he simply bids it. This maximizes the chances of finding a four-four fit and being able to stop at the two-level. If responder has at least invitational values (i.e. enough to make a "check-back" bid at his second turn), he can afford to start with 1 if he wants to. Partner will assume temporarily that there is no major suit fit and rebid 1NT with any balanced hand that didn't fit in the range for an opening 1NT [yes it's OK to rebid a major if the hand is unbalanced]. With a four-card major and invitational values, responder now makes a check-back bid (New Minor Forcing, XYZ, etc.) and if there is a major suit fit, it will be found.

Playing Walsh, therefore, even the second hand (♠84 KQ65 KJ6 ♣A963) must rebid 1NT. If partner has a weak hand with four hearts, he would have bid 1 to begin with. If he has invitational or better with four hearts, he will now make his checkback bid and all will be well.

Not everyone feels that Walsh is an advantageous system – so are there really any drawbacks, assuming that you have a good check-back system? One disadvantage is that with two minimum hands, our best contract might well be two diamonds on a 4-4 fit and we could miss it. But if you think about it, the chances of us actually being allowed to play in this ideal contract are negligible. If we have an eight-card fit and around half the points in the deck, the opponents are very unlikely to allow us to play there even if we do bid it.

In some cases, where we have a safe harbor in two diamonds, XYZ can come to the rescue. Suppose you have ♠3 AT83 QT832 ♣Q7. Playing Walsh, you bid 1, eschewing your five-card diamond suit. Partner now rebids 1NT. While 1NT may be the ideal matchpoint contract when non-vulnerable, it can be a liability when the opponents have at least an eight-card fit in an unbid suit (spades in this case). Although there are a lot of hands types that partner can have, he certainly has at least two diamonds since if he was unbalanced with a singleton diamond we would either raise hearts (possibly on three), bid 1♠ or rebid 2♣. It's also even possible that opener has 4-4 in the minors (see below) in which case his shape is most likely to be 3244 (although 2344 is also possible). In all of these cases, two diamonds should be a playable contract if you can get there. Having agreed XYZ (this also applies to two-way checkback), over his 1NT rebid, you bid 2♣, partner bids 2 and you pass!

Now, if the opponents come in over 2, you won't mind so much: you weren't that thrilled with the contract anyway. And even then, opener might raise with a diamond fit.

We touched briefly there on opening 1♣ with four-four in the minors. Since writing the draft of this, I published a blog entry specifically on that subject: Opening with 4-4 minors.

Now, let's concentrate on how XYZ works. I don't know who first developed it, but I think it is one of the best conventions around. I was fortunate to learn it in my early days of bridge so to me it is second nature. I'm still awaiting the book on XYZ (David Metcalf, Howard Piltch, are you listening?) but there is a reasonable summary on the internet here.

Like two-way new minor forcing, which operates over opener's rebid of 1NT, XYZ operates after any sequence that starts with three bids at the one level, including some which involve competition. The basics are easy of course. Responder's rebid of 2♣ forces a 2 response by opener, although in the case of a major-suit rebid by opener, who at that point is almost unlimited in strength, it is permissible to skip 2 if the hand is too good to allow a drop at 2. Because there are many more sequences available than in the special case of the 1NT rebid, there's a lot to discuss with partner about follow-ups. But to be honest, in ten years of playing the convention, I don't think that it's ever mattered that we didn't have all the follow-up sequences defined.

One thing I do like is this: after the invitational sequence 1X–1Y–1Z–2♣–2, responder now gets to show directly which potential "asset" he is interested in hearing about from opener. This allows more auctions to end at the two-level.

Let's look at an example: opener holds ♠KJ93 T87 A3 ♣KQT4 and responder ♠84 KQ652 JT6 ♣A93. Opener starts with 1♣ (assuming a 14-16, 15-17, etc. notrump opener) and responder bids 1. Opener rebids 1♠ (admittedly, 1NT is also possible). Responder bids 2♣, opener 2 and responder 2. Knowing we have an eight-card heart fit and and fewer than 25 hcp, we can decide to pass and we are still at the two-level (yes, I'm sure there's a few of you out there who will re-invite with 3 or jump straight to 4).

As always, however, we must ask ourselves if the natural meanings for the artificial bids wouldn't be more valuable. Basically, we are giving up the possibility of giving preference to partner's clubs at our second turn. We can play three clubs by going through the relay and then bidding 3♣ to play. What about a weak responding hand like ♠8 KQ62 JT6 ♣98732? If, after 1♣–1, opener rebids 1♠ or 1NT we could, playing standard methods, bid 2♣ to play. When we can't do that it's occasionally a concern, but I can't actually remember the last time it caused me any kind of a problem. Again, if we have a good fit in clubs, we will never get to play it.

To summarize, I have found that playing Walsh and XYZ together is simple, effective and feels very natural. Like any convention, if you choose to adopt XYZ you must realize that you'll have a couple of disasters before it shows its worth.  But I really think it's well worth the trouble in the long run.

Thursday, December 13, 2012

Dodging a bullet – or two

Imagine that you are playing the convention known as suction. I'm talking about the ACBL-legal version that operates over an artificial forcing club bid (1♣ or 2♣ depending on system). White on red at matchpoint scoring, you pick up ♠T96 87 ♣KJT9643 and, after partner passes as dealer, you hear RHO bid 2♣ (strong, artificial and forcing). You bid 3♠ which is essentially a club preempt or a two-suited hand with both red suits. The next player bids 4 and partner doubles. This is passed back to you. What's going on?

In fact, we have an explicit agreement about this type of situation (see DSIP rule summary), but for the sake of argument, let's actually think about the logic. LHO presumably has a good heart suit (this doesn't seem like a good time to be psyching!) and RHO has some kind of a monster hand. So, what kind of hand would partner have to have in order to make a penalty double here? Shortness in clubs, very good hearts and perhaps a couple of aces on the side. Pretty unlikely, right?

But in any case, we're forgetting something. Partner doesn't actually know whether we have the club hand or the red-two-suiter hand. Well, if he has the hand we think, with all of those hearts and aces, he can presumably infer that we don't have the red two-suiter. But that hand is so unlikely, it's just not in the realm of possibilities. There are only 40 points in the deck: opener has at least 20, we have four, LHO presumably has some, more if his hearts aren't very good. That doesn't leave a lot for partner's penalty double.

And, if partner does have this miracle hand and they end up playing in 4 instead of in a better fit, wouldn't we just be thrilled to get a positive score.

No, the only sensible thing that partner's double can possibly mean is this: "I have offensively oriented cards and am planning to compete – but I don't know for sure which suit to bid or how high until I get clarification of your hand."

Why wouldn't partner simply accept the "transfer" to 5♣ and let our hand decide? For the simple reason that, in competition, that would show – guess what? – clubs.

The more I experience what makes a good penalty double and what doesn't, the more I realize that, unless all hands are balanced, in which case a double will do well provided that our cards are sitting favorably and we have more of them, it's important to have no fit for partner. So, it's absolutely essential, in the case of a distributional hand, whether partner is single-suited (which suit?) or two-suited (need to know both suits).

Opportunities for this kind of double abound. Let's suppose that we are playing DONT against 1NT openers. LHO opens 1NT and partner doubles showing a single-suited hand. RHO bids 2 (transfer). Our hand is something like KTx Qx KJxx Qxxx. We definitely want to compete here but we don't know where yet. So, we double, showing "cards" and not especially diamonds. In the unlikely event that LHO passes and partner's suit is diamonds he will pass too (this actually happened once). Double to ask for partner's suit is so much more useful than doubling to show diamonds.

So, any time that partner has left us in the dark about a suit (or, in the case of suction, their whole hand) double is best used to inquire. It's worked well for me.

So, what about dodging those bullets? Well, par on this board was -500 (6♣X down 3). My hand, by the way, was ♠82 76 AJ432 ♣A852. As you can see, my hand was eminently suitable for a sacrifice just so long as I can be sure which minor to bid. So, after my RHO bid 4, I doubled. LHO passed and so did partner! That could only mean that he had the red two-suiter. Had my RHO chosen to pass at the point, we would have been -890 for an absolute bottom. But fortunately, she also thought my double was for business. So she bid 4♠. Naturally, I bid 5. See how much I trust partner? Although I suppose if I had trusted him 100%, I might have bid 6! If either of my opponents had pulled out the red card over 5, we would have been -800 for 1 matchpoint out of 17. Fortunately, LHO bid 5 and there we let it rest for a 43% board. Not as good as the 73% board we might have achieved for the par result, but at least not a zero.

Sunday, December 2, 2012

Romanized Blackwood follow-ups

Well, I seem to have been laboring under a misunderstanding for the last ten years or so. The error in my ways was finally brought home to me at a club game with my partner Len a few weeks ago. It was actually the second time in maybe a year that this had happened (different partners) and I began to sense a pattern.

The issue arises because the Romans have an unusual counting system – and I don't mean all of those Xs, Ms, Ls, Is, etc. No, they count like this, using what is technically a modulo-three system with an inversion of the second and third digits. 0 (or 3 or 6, etc.), 2 (5, 8...), 1 (4, 7...). No wonder they have so many car accidents – they have no idea how fast they're going.

I'm kidding of course. This counting scheme, which so far as I know is only used in bridge and even then only when checking controls for a slam, is actually very well-suited to the problem at hand, especially when counting key cards [as in Roman key-card Blackwood], as opposed to just aces, of which there may be any number between zero and five. The two (and five) key card responses are imbued with an extra piece of information: possession of the trump queen (after the other responses, asker must inquire if he needs to know).

After a (Roman) 4NT Blackwood enquiry elicits one of the ambiguous responses: 5♣ or 5, there is a potential problem. Whose responsibility is it to continue bidding when responder actually has the higher number of "aces"? The books are generally quiet on this point, although some of them imply that it is asker's responsibility because they make comments like "Asker should know from the auction what the response shows." Only Fred Gitelman (as far as I have seen) casts any doubt on the efficacy of this idea by observing that there have been some spectacular grand slams doubled down 3 after this occurrence.

For my part, I learned somehow – I know not whence or from whom – that asker should assume the lower number and sign off. And therefore that responder would continue bidding with the higher number. This wouldn't work however when clubs or diamonds is the agreed suit and responder bids our suit at the five-level. Now, opener must step up and make the decision because a pass will end the auction unilaterally.

But when a major suit is agreed, the notion that responder will continue with the higher number is still plausible (though apparently not guaranteed). As an example, here is our auction from the most recent debacle: 1  1  2 4 5 5 – all pass. 2 showed 15-17 "dummy points" since we play 12-14 no-trump; 4♠ was kickback for hearts; 5♣ showed zero or three key cards. Is it conceivable that I would be asking for key cards with none of my own? What kind of key-card-less hand could I possibly have that wants to be in slam opposite a likely balanced hand 15-17? On the other hand, could partner possibly have a hand with no key cards? Not in this universe.

I was adamant that it was responder's responsibility to get us to slam with the higher number and that it was not required for asker to have to try to work out responder's possible hands. But unfortunately I have found no support for my position in the literature.

In this particular case, it was obvious to each of us that we had all or most of the key cards. My hand: ♠QT AQ84 KT6 ♣AK54 (two key cards + Q). Partner's hand: ♠AK53 KT75 AJ5 ♣J2 (three key cards). Assuming that partner did indeed have three, I would have played in 6NT rather than hearts. In fact, it makes 7NT with careful play. But it makes only 6 because the trumps split 5-0. In practice, I didn't play the hand well and only just made my contract for a poor match-point score (although not a zero). I would have likely have made 12 tricks in no-trump.

So, let's take an objective look at the situation. There's no possible hand for responder that has zero key cards given this auction. So let's look to see if there's any bad hand that asker might have had consistent with the bidding. For 5 to be the proper contract, asker would have to have no key cards at all and yet be invoking the key-card ask. Asker might have this hand: ♠QJT2 QJ84 KQ6 ♣KQ. Is it conceivable that this hand would be trying for slam opposite a hand with15-17 dummy points? It is not. But what if asker had a more distributional hand such as ♠Q2 QJ98432 – ♣KQ54. Note that splintering or asking with "exclusion key-card Blackwood" would not be appropriate here since opener began with 1. Could this hand be looking for slam by jumping straight to 4NT? Yes, it's just possible but extremely unlikely.

I think there's another, more useful, principle that can be applied here, and which I wrote about a while back in Aceless Wonders. This principle can be generalized to the following rule: do not indicate a strong hand without an ace. Or, to put it another way, all aceless hands are (relatively speaking) weak. So, strength-showing bids (incidentally, these are almost the same set of actions which Eddie Kantar considers to be triggers of the 1430 asks versus 3014 asks in his system of Roman Key-card Blackwood) all promise at least one ace:
  • opening with 2NT, (strong) 2♣ (or 1♣ if playing a big club system);
  • (opener) reverses;
  • strength-showing jump (in the case of responder, this would have to be a strong jump shift);
  • (opener) makes a game-forcing cuebid in competition;
  • asking for key-cards (if none of the previous triggers have already occurred).
Using this principle, I don't think it is possible to get into a bad slam (the type where you are missing three key cards). In other words, key-card asker can always bid on – assuming the higher number of key cards – unless key-card responder has never implied an ace by making a strength-showing bid. In such a case, it now becomes responder's responsibility to continue bidding when holding the higher number of key cards. There is however still an outside chance that each partner has just one key card and that the partnership bids up to the slam level – but this would require both partners to have overbid their hand in the same auction.

The more significant problem with this rule is that both partners have to be on the same page (this page, in fact). I suppose then that a practical solution is the following: each partner has the responsibility of bidding on unless it is probable that the partnership is missing three key cards.