[frank0]

Consider the bidding in IMP

---

Even the double is not strictly defined as penalty, we expect partner pass most of the time whether we double or pass, so our decision should be based on how often/much the contract could be set.

Imagine we have

---

Double with this looks crazy but according to DD result double is +EV bid.

Set up
We give South exactly this hand
We give East 6-11 HCP with at least KQTxxxxx in D(at least 8 cards, 2 out of top3 and 3 out of top 5)
Otherwise hands are random
Calculate the DD result of 10000 hands E playing D contract(using deal 3.1.9).

Result Frequency Pay if we X(IMPs, number in bracket means opp redouble)
Make7 119 -7(-13)
Make6 1051 -6(-11)
Make5 1839 -5(-9)
Down1 2764 +2
Down2 2568 +5
Down3 1411 +8
Down4+ 248 +12

If we assume opp always redouble when they make 11+ tricks we still have about +0.3IMP expectation, it becomes +1.6IMP if they never redouble.

Any comment is welcome.

[johnu]

How are you planning to bar partner from bidding? I can pretty much guarantee that partner will not be happy if they take out the double and see that hand as dummy.

[BillPatch]

johnu, on 2013-October-11, 23:40, said:

How are you planning to bar partner from bidding? I can pretty much guarantee that partner will not be happy if they take out the double and see that hand as dummy.

If partner believes that you will only double with the current standard for a double of a four level or five of minor double(the HCP of a strong NT opener), he will overbid. If he knows that your minimum is that of the OP(2 QT, 10 hcp outside of diamonds, at least 3 cards in each major) partner will probably be able to bid more good games overall at the 5 level than using the old agreement due to greater frequency of minimum doubles.

With a fair 5 card major partner would be able to bid game on a semi-balanced hand with 20 hcp, with the same major and a small singleton 13 hcp outside of diamonds, with a void in diamonds approx. 11 hcp outside of diamonds.

I base the 13 hcp outside od diamonds on Sam Stayman's dictate in "Do You Play Stayman? c. 1963 book that with 23 hcp between the other three suits an a singleton wth sufficient trumps you should have a good play for holding the trick loss outside of the singleton suit to one trick. Assuming another trick lost in the simgleton, 5 of a major is gin.

[nate_m]

Seems like the results indicate about what one would expect, namely that on balance it is unlikely that one side will take 11 tricks. I would imagine that the dd results would show similar results doubling on 2 aces and out. They might well show gains holding even less. In practice, I think it's a huge loser to make the meaning of X in that auction "Practically forced. It is a-priori unlikely that your side will take 11 tricks." I'd also guess that the results of the dd sim don't really show any advantage to doubling with the given hand because on a lot of the hands where you get it for a number your partner might have hit it playing a normal doubling style.

[MrAce]

Here is my problem with the statistics of OP; he deals thousands of hands and comes up with some results. Even if those results are % 100 accurate, in real life you do not face 5 minor opened in front of you more than (i am guessing) 30-40 times in your life time. When you do not face as much as simulation hands in real life, or anything close to, the statistics of thousands of hands may fail you big time in small pool. With very limited number of hands in real life, you need to be able to make more accurate decisions. It is not something like 2-3-4 level preempts, which you can take action relying on statistics and expect to be up in the long run. Simply because there is no long run vs 5 level preempts for 1 pair.

My point is, when you decide to defend against very rare type of preempts with the statistics, you are taking a risk of getting random results. Last thing you want is to go for random results as an expert player. Because this type of hands makes the difference between you and non expert players. Your experience, hand evaluation, knowledge outweighs the statistics by a lot when the situation you are facing or likely to face in the future is very rare. To be honest, i would never set a style in my pdship where we double 5m with a huge contrast of hands. Because i don't know how many months, if not years, i will wait for the next 5 minor preempt by opponents. Last thing i want to do would be relying on statistics in the case where they open 5 m.

[PhilKing]

If only someone had a searchable database of hands from top level competition that might shed light on this situation. Oh wait - they have!

Try this one:

♠KT74
♥AK63
♦AQ
♣T98

You are at red, of course. Lefty opens Five Diamonds and pard doubles. Righty passes - your go. Partner is Nunes if that helps.

[awm]

This is why I don't think opening 5♦ on a "random" weak hand with eight diamonds is the right call. Opening 4♦ works surprisingly well, as opponents will much more often try 4M and go down (possibly doubled). The problem with opening 5♦ is that you take away a tough choice from opponents most of the time (whether to play their own game or defend). You need to be pretty sure that playing 5♦X will be a good result.

This is not to say that I would "never" open 5♦ with only eight diamonds (although I prefer nine); it's more that I would need a hand with truly no defense (♦KQJxxxxx and out at NV) or a hand with extra ODR (say an 8-4 hand).

[PhilKing]

PhilKing, on 2013-October-12, 17:41, said:

Try this one:

♠KT74
♥AK63
♦AQ
♣T98

You are at red, of course. Lefty opens Five Diamonds and pard doubles. Righty passes - your go. Partner is Nunes if that helps.

Companion hand was:

♠A982
♥QJ94
♦873
♣AJ

5♦ goes for 1100. Fantoni drove slam and recorded -100 in 6♥.

That's all from the database, which is now in lockdown mode!

[gszes]

your statistical analysis is quite correct and an X of 5d at MP is much more practical since
we rate to gain MP 70% of the time. At IMPS however, such a strategy is hopeless since
(even sans xx) we are expecting mostly small gains measured against large losses. By
passing with these weakish hands and saving our x for strong playing hands we may be
able to confidently bid small and even grand slams that are ridiculous to bid if we use the
weak standard shown as the minimum for x. At IMPS it would be vastly more preferable to
suffer moderate to small losses 30% of the time while picking up moderate to huge gains
the other 70% of the time.

[BillPatch]

gszes, on 2013-October-13, 19:14, said:

your statistical analysis is quite correct and an X of 5d at MP is much more practical since
we rate to gain MP 70% of the time. At IMPS however, such a strategy is hopeless since
(even sans xx) we are expecting mostly small gains measured against large losses. By
passing with these weakish hands and saving our x for strong playing hands we may be
able to confidently bid small and even grand slams that are ridiculous to bid if we use the
weak standard shown as the minimum for x. At IMPS it would be vastly more preferable to
suffer moderate to small losses 30% of the time while picking up moderate to huge gains
the other 70% of the time.

So we are to refrain from doubling unless slam is 70% on a random hand; I thought the norm was
the HCP of a strong NT. I am assume that if you have the cards for a strong try, you will
usually be able to defend for approx. the same level of score, only a slam will produce sizable
gains over 70%. Or a double game swing.

I also disagree that the statistical analysis of the model is a correct approximation of the real
world problem. If we pass preemptor's partner has a good knowledge of the opening bid, will often
bid the good slams. The imp value of 5 ♦making 6 is about the same as bidding
and making slam. For those hands, no negative imo ev for our double. Also, if we pass balancer
will frequently be able to make a double, producing no positive imp ev for our bid, as the contract
will still be the same. I think adding the analysis of decision rules for these two actions will
provide a better estimate of the imp ev of the double for the given hand. I think that the modeling
these two effects would produce

Once we know whether this hand is + imp ev, we can test similar hands to refine a model of minimum
+ev hands with similar offensive values, so partner can accurate know what his expectations of bidding
over the double.

Finally we will test the model by comparing auctions comparing our new ranges versus the assumed standard
ranges bidding random hands after the given 5 ♦ opener.

[Lorne50]

frank0, on 2013-October-11, 23:10, said:

If we assume opp always redouble when they make 11+ tricks we still have about +0.3IMP expectation, it becomes +1.6IMP if they never redouble.

Any comment is welcome.

I got similar results but there is one thing missing from the analysis. If you pass it is not certain the other two will pass. It may not be possible to simulate but I guess that a lot of the time partner will double if you pass so the bidding gain is smaller than you think.

Also some of the time you double and they make 12 or 13 they may have been bidding it so some of your redouble results should be compared to them bidding 6♦.

It is interesting though.