[han]

Over 400 hands:

One overtrick: 1 (0.25%)
making exactly: 18 (4.5%)
Down one: 58 (14.5%)
Down two: 100 (25%)
Down three or more: 223 (54.75%)

Average number of tricks: 6.415

[awm]

Han is running a different simulation than I assumed (i.e. he was assuming partner has a sound penalty pass). Yes, in these cases 3♦X often goes down. But it might be worth considering how often 3NT and 4NT are making opposite these hands as well. Even down three is a minus position if you can make a game or slam, and his point range (0-15) includes plenty of hands that can make a game or slam.

I'd expect that the hands where the penalty pass has those two diamond tricks and that's it, you get a lot of 3♦X making or 3♦X down one. On these hands you may be better off playing 4♣-1.

On the hands where the penalty pass has a lot of additional cards on the side, sure you are beating 3♦ by several tricks, but you pretty often had a making game too (and it might pay more).

[cherdano]

awm, on Oct 27 2008, 05:22 PM, said:

Han is running a different simulation than I assumed (i.e. he was assuming partner has a sound penalty pass). Yes, in these cases 3♦X often goes down. But it might be worth considering how often 3NT and 4NT are making opposite these hands as well. Even down three is a minus position if you can make a game or slam, and his point range (0-15) includes plenty of hands that can make a game or slam.

I'd expect that the hands where the penalty pass has those two diamond tricks and that's it, you get a lot of 3♦X making or 3♦X down one. On these hands you may be better off playing 4♣-1.

On the hands where the penalty pass has a lot of additional cards on the side, sure you are beating 3♦ by several tricks, but you pretty often had a making game too (and it might pay more).

Are you suggesting to bid 3N instead? How else does making 3N come into play?

[awm]

han, on Oct 27 2008, 09:19 AM, said:

Over 400 hands:

One overtrick: 1 (0.25%)
making exactly: 18 (4.5%)
Down one: 58 (14.5%)
Down two: 100 (25%)
Down three or more: 223 (54.75%)

Average number of tricks: 6.415

Okay, I am convinced that if partner has four diamonds to at least two honors, 3♦X will usually go down and will go -3 almost half the time.

How exactly is this meaningful?

Are you implying that:

(1) Partner will only pass for penalties holding four diamonds to two honors?
(2) On most of these hands we could not make a game somewhere?
(3) Penalty pass hands are sufficiently frequent that the call which works best opposite a penalty pass is highly likely to be the best call?

All three of these seem dubious to me.

[mikeh]

cherdano, on Oct 27 2008, 06:26 PM, said:

awm, on Oct 27 2008, 05:22 PM, said:

Han is running a different simulation than I assumed (i.e. he was assuming partner has a sound penalty pass). Yes, in these cases 3♦X often goes down. But it might be worth considering how often 3NT and 4NT are making opposite these hands as well. Even down three is a minus position if you can make a game or slam, and his point range (0-15) includes plenty of hands that can make a game or slam.

I'd expect that the hands where the penalty pass has those two diamond tricks and that's it, you get a lot of 3♦X making or 3♦X down one. On these hands you may be better off playing 4♣-1.

On the hands where the penalty pass has a lot of additional cards on the side, sure you are beating 3♦ by several tricks, but you pretty often had a making game too (and it might pay more).

Are you suggesting to bid 3N instead? How else does making 3N come into play?

I assume that he is thinking that partner will, on occasion, consider bidding a red v white game.. indeed, with a diamond holding such as KQx or AQx (and extras in the other suits), where he has two stoppers, but the opps rate to take 5 diamond tricks, it is easy to see why he might elect to try for 600 rather than 500 or 800.. plus, if a slam is available, he may get there by a value-showing 3N while passing leaves him with at best an 800 into his 1430/1440.

To balance 3N is sick... it may work, of course, but I don't think that Adam was suggesting that route to game.

BTW, I would be interested in knowing how many of Han's sample of 100 hands would/could have warranted a 3N bid rather than a pass. One of the problems with simulating this type of auction is accounting, if even possible, for the subjective element.. and it is almost impossible to do this objectively when the simulator knows the actual hand.

[mikeh]

I should add that I see this as an automatic double. Yes, it may not work out, for various reasons covered by others, but so what? 4♣ is a nothing bid.. you can't get to your two most probable good spots... 3♦ doubled or 3N by partner, and you have seriously reduced your chances of reaching a good major suit contract.. and for what? The dubious safety of a 10 trick +130? If you think that this call will let you reach a good club slam or a good 5♣.. maybe you are right... on a small percentage of hands.

[jdonn]

awm, on Oct 27 2008, 06:41 PM, said:

Okay, I am convinced that if partner has four diamonds to at least two honors, 3♦X will usually go down and will go -3 almost half the time.

It's not "-3 almost half the time." It's "-3 or more, over half the time." That is pretty different.

Quote

How exactly is this meaningful?

I don't see why you have a problem with han's analysis at this point. You invited it by saying things like

awm, on Oct 26 2008, 11:14 PM, said:

I will miss a few penalties on hands where we should defend, but I will also miss some 3♦X making results.

as though you were implying these results will happen an equal amount of the time and were therefore some kind of wash. This sim absolutely refutes that! If you believe it's poorly set up somehow and a penalty is not 20 times more likely than a make when partner penalty passes, you may be at least partially right. But you simply can't argue with a straight face that a penalty is not a LOT more likely than a make.

Quote

Are you implying that:

(1) Partner will only pass for penalties holding four diamonds to two honors?
(2) On most of these hands we could not make a game somewhere?
(3) Penalty pass hands are sufficiently frequent that the call which works best opposite a penalty pass is highly likely to be the best call?

All three of these seem dubious to me.

(1) If partner has four diamonds, he is not at all likely to pass without at least two honors. And if you are implying he left out cases of penalty passes working badly (such as four cards to one honor [and who is to say those would work badly?]), I respectfully submit he left out cases of penalty passes working well (such as five cards to one honor).
(2) If the game is 3NT, you can no longer reach it. If the game is 4M, double is certainly better for reaching it (I certainly don't believe partner will bid a mediocre five card major over 4♣). If you believe the game is 5♣ then your bid is certainly better for reaching it, but that's the least likely game (well at least in a sense, since it's the highest) and in those cases another game may make anyway.
(3) You included the possibility of a penalty pass in your argument too! You can't have it both ways.
(4) You criticized his sim. We still don't know what parameters you used...

[cherdano]

mikeh, on Oct 27 2008, 05:47 PM, said:

cherdano, on Oct 27 2008, 06:26 PM, said:

awm, on Oct 27 2008, 05:22 PM, said:

Han is running a different simulation than I assumed (i.e. he was assuming partner has a sound penalty pass). Yes, in these cases 3♦X often goes down. But it might be worth considering how often 3NT and 4NT are making opposite these hands as well. Even down three is a minus position if you can make a game or slam, and his point range (0-15) includes plenty of hands that can make a game or slam.

I'd expect that the hands where the penalty pass has those two diamond tricks and that's it, you get a lot of 3♦X making or 3♦X down one. On these hands you may be better off playing 4♣-1.

On the hands where the penalty pass has a lot of additional cards on the side, sure you are beating 3♦ by several tricks, but you pretty often had a making game too (and it might pay more).

Are you suggesting to bid 3N instead? How else does making 3N come into play?

I assume that he is thinking that partner will, on occasion, consider bidding a red v white game.. indeed, with a diamond holding such as KQx or AQx (and extras in the other suits), where he has two stoppers, but the opps rate to take 5 diamond tricks, it is easy to see why he might elect to try for 600 rather than 500 or 800.. plus, if a slam is available, he may get there by a value-showing 3N while passing leaves him with at best an 800 into his 1430/1440.

My understanding was that Adam was arguing against double. If partner has a natural 3N bid and we are making 9 tricks there, then that is an even stronger case against bidding 4♣...

[cherdano]

awm, on Oct 27 2008, 05:41 PM, said:

(2) On most of these hands we could not make a game somewhere?
(3) Penalty pass hands are sufficiently frequent that the call which works best opposite a penalty pass is highly likely to be the best call?

I think you have it completely backwards, Adam. It doesn't matter if we "have a game somewhere" unless your suggested alternative (I assume you want to bid 4♣?) would actually get us there.

Above, your strongest argument against doubling seemed to be that it won't work well if partner makes a penalty pass, so (3) is relevant to your arguments.

[han]

Since I'm heavily under attack, here is my defense:

I did not draw any conclusions. I presented my results, and then my parameters. There are some very smart people reading who can draw their own conclusions. Since matmat asked me, on the hands I dealt (the ones where partner has 4+ diamonds etc.) pass beats 3NT by about 0.305 IMPs per hand. I don't think this is a very significant.

I don't think that xxx xxxx J10xx xx is a very sound pass. The program I use does not allow me to specify spotcards lower than the 10, 2 of the top 5 seemed a reasonable specification to me.

With down at least 3 I meant down 3, down 4, down 5 or down 6. Yes, there was one hand where declarer was down 6. Kuddos to jdonn for understanding my English.

I also want to make the following comments:

Of course I did not allow for 5-card majors.

I don't believe that partner has 3 clubs on average when he passes our double.

No, 4441 is not the only shape where partner has club shortness. Partner could also be 3451, 4351, 4450, 3361, 2461, 4361, 3460 or 4360.

[xcurt]

han, on Oct 27 2008, 08:18 PM, said:

Since matmat asked me, on the hands I dealt (the ones where partner has 4+ diamonds etc.) pass beats 3NT by about 0.305 IMPs per hand. I don't think this is a very significant.

Don't sneeze at 0.3 IMPs per board. In another thread I argued that the difference between a team of, say, 4 Han's, and a top 10 seed in the Spingold, was probably about 1 IMP per board. Getting this kind of thing right is a not-insignificant part of the difference between the pros and the rest of us.

Although we're talking about advancer's problem and not balancer's problem, I think this suggests that balancer's hand is a clear double, since we have the most 3NT-oriented balancing double we could imagine, and about the worst defense we could imagine, and we're still ahead when partner chooses to play 3♦x given the set of logical alternatives {Pass, 3NT}.

[han]

Yikes, two people in one thread that try their best to insult me!

I did not sneeze at .3 IMPs per hand. I said it was not very significant here because we are comparing the merits of double compared to pass on a very select subsets of hands, and it says very little about the relative merits of 3NT (a call matmat flirted with) and double.

Morover, since most of the swings in the set were quite large (7-10 IMPs usually) and the specifications were chosen fairly arbitrarily, a different choice in parameters could easily lead to a quite different IMP difference.

[FrancesHinden]

xcurt, on Oct 28 2008, 05:52 AM, said:

han, on Oct 27 2008, 08:18 PM, said:

Since matmat asked me, on the hands I dealt (the ones where partner has 4+ diamonds etc.) pass beats 3NT by about 0.305 IMPs per hand. I don't think this is a very significant.

Don't sneeze at 0.3 IMPs per board. In another thread I argued that the difference between a team of, say, 4 Han's, and a top 10 seed in the Spingold, was probably about 1 IMP per board. Getting this kind of thing right is a not-insignificant part of the difference between the pros and the rest of us.

Although we're talking about advancer's problem and not balancer's problem, I think this suggests that balancer's hand is a clear double, since we have the most 3NT-oriented balancing double we could imagine, and about the worst defense we could imagine, and we're still ahead when partner chooses to play 3♦x given the set of logical alternatives {Pass, 3NT}.

I think you are misunderstanding the use of the word "significant".

of course a reliable gain of 0.3 imps a board is worth having.

But there are a lot of guesses made in setting up a simulation*, and it's not over all possible hands, and it's double dummy results rather than single dummy... all of these facts mean that a difference of 0.3 imps is not statistically significant and you shouldn't draw the conclusion that one call is better than an other just on 0.3 imps.

*For example, against one of my regular partners your parameters for the 3D opening in the sim are wrong

[keylime]

Double. I still have the ability to bid clubs if pard makes some noise. Since he/she didn't, time to try for 300/500.

[awm]

There are many types of hands you need to consider here. I would roughly divide these by frequency:

(1) Hands where partner has a four-card major but not five-card major and is not suited to penalty pass. Usually I would expect partner to bid his four-card major (at the three or four level depending on strength) in response to a takeout double. Partner would pass 4♣ or try one of 4NT or 5♣ depending on strength opposite a 4♣ overcall. My expectation here is that playing in 3M/4M is almost invariably bad. Typically 4♣ is better (as well as less likely to get doubled) opposite the bad hands. Opposite the good hands it is fairly frequent that 3NT is the best game but you will often be better off in 4NT/5♣ than you will be in 4M on the moysian (i.e. give partner ♠AKxx ♥xxx ♦Kx ♣xxxx; I would expect a 4♠ bid opposite a double. Clearly 3NT is best, but 4NT has good chances also and 5♣ has some play as well).

(2) Hands where partner has a five-card major. On the ones where partner's hand is bad, you definitely win by doubling since you get to play 3M on a 5-3 instead of 4♣. When partner's hand is good, it is more of a toss-up. Bidding 4♣ will often reach the good 4M game anyway, you will occasionally play 5♣/4NT instead of 4M (but this might even be better if the major breaks 4-1 or worse) and you will occasionally reach a good club slam that you'd never get to after double.

(3) Hands where partner has a penalty pass of 3♦ but not a lot of general values outside diamonds. Doubling, you will sometimes defend 3♦X making. More often you will beat 3♦ a trick or two (but rarely more). Bidding 4♣ normally gets you -200 (but it's hard to double with no trump cards). My feeling is that defending will be expectation-negative score here, although it's probably more like -100 than -200. Han's simulation doesn't really solve this case because he includes many hands where partner has a bonanza outside diamonds and 3♦ goes down multiple tricks.

(4) Hands where partner has a penalty pass of 3♦ and a bunch of outside cards. Bidding 4♣ gets you to 4NT, which sometimes makes and sometimes doesn't. Defending 3♦X gets you a plus score, usually +500 (but could easily be +300 or +800 depending on how things break etc). I'd expect that on these hands, 3NT is in expectation better than defending given the vulnerability and nature of our hand, but it is much less clear whether 4NT will be better than defending.

(5) Hands where partner has a 3343 or 3334 shape or the like, not really suited to penalizing. If you double and partner bids 3M on a three-card suit it will be absolutely awful. If partner finds a 3NT call with a diamond stop, you're in very good shape. If partner converts the double you are probably in trouble. Bidding 4♣ gets you to a relatively safe partial when partner's hand is bad and a slightly unsafe game (4NT/5♣) when partner has some cards.

As for my simulation parameters, keep in mind that I do not have a double-dummy solver easily available, and I am trying to look at all hands partner could have, not just "penalty pass" hands. All I did was to specify that opener has 6♦ with two of the top three honors or any 7♦, we have the given hand, and see what the layouts look like. I threw out a few hands where partner or RHO had an obvious call other than pass over 3♦ (these were quite rare actually). I can post a bunch of hands if people really care to see them.

[xcurt]

FrancesHinden, on Oct 28 2008, 10:37 AM, said:

xcurt, on Oct 28 2008, 05:52 AM, said:

han, on Oct 27 2008, 08:18 PM, said:

Since matmat asked me, on the hands I dealt (the ones where partner has 4+ diamonds etc.) pass beats 3NT by about 0.305 IMPs per hand. I don't think this is a very significant.

I think you are misunderstanding the use of the word "significant".

The post I quoted didn't specify if "significant" meant "0.3 IMPs is chump change" or "significant" meant "0.3 IMPS is a tiny fraction of the variance in outcomes. I read it for the former, but I might have misread that poster.

No I didn't mean to insult anyone. I'm sorry if I did. I'm not sure I understand the reference to two people trying to insult Han either. But 0.3 IMPs per board is a meaningful difference (unless swamped by the variance, which doesn't mean that 0.3 IMPs per board isn't worth picking up, just that your simulation hasn't told you much).

Also, I think the point of the simulation got lost. In fact I may rerun the simulation myself and post dealer code. I learned something from this hand. I think the point of the simulation is that responder should pass very aggressively in the auction 3♦-P-P-Dble; P-?. If you accept that, this hand becomes much less of a problem since we don't have to worry about partner pulling to a bad 4-3 major suit as often as we might if partner is of the "I take out my partner's takeout doubles" school.