[mikeh]

 PhilKing, on 2015-July-29, 15:52, said:

FWIW I do not believe WesleyC is being as naive as is perhaps being assumed, but I do not agree with the approach (the outcome has some interest, in that it hugely favours bidding, where I would assume the actions were close, even with a bludgeon sim).

One has to weigh the stated limitations of the sim accordingly, and he sets them out clearly. Personally, I would always go for a human-filtered sim in cases like this.

I wasn't kidding when I said I appreciated his effort. It is pretty easy to critique a simulation after the fact, especially when, as should always be done, the simulator has been careful to describe his constraints. The fact that I think, with the benefit of seeing the results, that the simulation is seriously flawed doesn't diminish the effort. Indeed, it remains useful in the sense that it seems to suggest that getting our side to 5♥ is probably a good idea more often than not. Unfortunately, I do think that it doesn't advance analysis of who should get us there...us or partner.

Doing the subjective analysis required to deal with 'what happens if we pass' is incredibly tedious and unreliable, since no two players will likely agree on all 10,000 hands....we will argue over the 5♦ bid and partner's choice. I have done 'eyeball' quicky sims of 100 hands on occasion, and 100 is about the most I'd ever do, which sample size, even if I were to be seen to be a perfect decision maker, is pretty small. I just don't think this kind of problem is susceptible to meaningful analysis with the tools with which I am familiar.

[WesleyC]

 apollo1201, on 2015-July-29, 13:32, said:

Could you pls give us the frequencies of the number of tricks made? If you still have them ;-)

In the case where Partner holds 11 or fewer HCP, our side makes on average 9.85 tricks in hearts and they make 10.18 tricks in diamonds. Partner average length in hearts is 4.68.

East playing Hearts
Tricks - Frequency

4 2
5 9
6 48
7 211
8 560
9 1071
10 1403
11 1143
12 466
13 87

South playing Diamonds
Tricks - Frequency

6 2
7 23
8 216
9 1047
10 1820
11 1322
12 565
13 5

[WesleyC]

 mikeh, on 2015-July-29, 14:05, said:

Wesley, I appreciate the effort that went into the simulation, but (and I mean this) with a great deal of respect for the effort, I think your simulation is so flawed as to be basically meaningless.

In short, we have to exercise a great deal of judgement to determine the outcome of any choice, and sims are no good at that: at best a sim provides plausible hands to the decision-point, but thereafter everything becomes subjective.

Cheers Mike.

I can understand your skepticism, and you raise many valid points. The fact that i've presented the results as a summary is definitely a limitation because much of the value comes from being able to eyeball a bunch of deals which are consistent with the scenario to get an insight into the kinds of hand shapes and point distributions that are most likely.

Trying to decide what constitutes a 5D bid is really difficult, because players styles vary so widely. At these colours my 5D bid would usually be an 84 or a 9 card suit. Even holding a strong 8221 with solid diamonds and outside values, I'd prefer to keep 3NT in the picture. However against opponents that always take saves and don't double enough, I might widen my range considerably. At the end of the day, glancing over a sample of the hands, most looked reasonable for a 5D bid. Finally, i'm pretty sure that increasing the playing strength of the 5D bidder would actually make it MORE likely for bidding to be the winning action. I'll mess around with the conditions a bit and get back to you on this.

Defining the hands on which partner will choose to act also depends heavily on your style and agreements. As a counterpoint, i'm guessing that most people play partner's double in this spot as 'values' with a penalty orientation. Given you're always planning to pull the double, this might cause you to reach 5H when you were better off to defending, and not reach 5H when it was the right but partner didn't have enough defense to double. Again glancing over a sample of the hands from part 2 of the sim (where partner holds a maximum of 11 HCP), i'd guess that at least 80% are clear cut passes.

However, I think you're missing the main point of the exercise, which is simply to apply some amount of concrete analysis to a decision that would otherwise be purely based on hand-waving arguments.

[WesleyC]

I redid the sim, giving South a hand more consistent with a conservative 5D bid.

If South holds only an 8c Diamond suit, they always have a void in a side suit. Holding a 9c diamond suit there is no shape restriction. South always has at least 12 HCP.

I maintained the condition that partner holds 11 or less HCP (in order to exclude many of the cases where they will be able to act).

In this scenario, the opponents averaged 10.6 tricks in diamonds and we averaged 9.7 tricks in hearts.
This significantly INCREASED the value of bidding 5H.

Tricks in Hearts
2 -2300 2
3 -2000 1
4 -1700 2
5 -1400 12
6 -1100 53
7 -800 241
8 -500 636
9 -300 1168
10 -50 1345
11 450 1042
12 480 431
13 510 67

Tricks in Diamonds
7 400 9
8 300 157
9 200 518
10 100 1523
11 -600 1770
12 -620 987
13 -640 36

The total point expectancy of bidding 5H compared to defending 5D is +217.
Converting to IMPs gives an expectancy of +3.5 IMPs/board.

And finally i've uploaded a sample of 100 deals to eyeball.

100 Sample Deals

[mikeh]

 WesleyC, on 2015-July-29, 23:25, said:

I redid the sim, giving South a hand more consistent with a conservative 5D bid.

If South holds only an 8c Diamond suit, they always have a void in a side suit. Holding a 9c diamond suit there is no shape restriction. South always has at least 12 HCP.

I maintained the condition that partner holds 11 or less HCP (in order to exclude many of the cases where they will be able to act).

In this scenario, the opponents averaged 10.6 tricks in diamonds and we averaged 9.7 tricks in hearts.
This significantly INCREASED the value of bidding 5H.

Tricks in Hearts
2 -2300 2
3 -2000 1
4 -1700 2
5 -1400 12
6 -1100 53
7 -800 241
8 -500 636
9 -300 1168
10 -50 1345
11 450 1042
12 480 431
13 510 67

Tricks in Diamonds
7 400 9
8 300 157
9 200 518
10 100 1523
11 -600 1770
12 -620 987
13 -640 36

The total point expectancy of bidding 5H compared to defending 5D is +217.
Converting to IMPs gives an expectancy of +3.5 IMPs/board.

And finally i've uploaded a sample of 100 deals to eyeball.

100 Sample Deals

I'm not sure where I'd draw the line for partner not reopening, and I appreciate that simulations require an arbitrary cutoff. Using hcp as the only metric probably cannot be avoided, but I think 12 as the minimum needed to reopen is too much.

However, that is a minor quibble and I thank you for your efforts.

[mikeh]

On a quick read of the sample hands, I think I would double as partner on roughly 28 of the hands. That means that my pass as opener would mean defending on 72 of the 100 sample hands, on which you would be playing. I will see if I can find the time to figure out whether on those 72, bidding is better than passing. It does IMO confirm both that the subjective element is very important and that using an 12 hcp requirement for partner to act is wrong. Put another way: give me a 4423 hand with both major aces, and how on earth can we defend an 11 trick contract opposite an opening hand undoubled? Of course, opposite that double I get to a hopeless 5H just as you do my only point is that I think your cutoff really, really skews the sim in favour of bidding over 5D.

[WesleyC]

Yeah, good point. Aces are extremely important when defending against high level contracts.

I can easily plug a different metric for HCP into the sim and see if it makes a difference.

What about Aces = 4, Kings = 2, Queens = 1 and you need 8 points to reopen?

[billw55]

 WesleyC, on 2015-July-29, 23:25, said:

I redid the sim, giving South a hand more consistent with a conservative 5D bid.

...

In this scenario, the opponents averaged 10.6 tricks in diamonds and we averaged 9.7 tricks in hearts.
This significantly INCREASED the value of bidding 5H.

This is not entirely surprising, as the better south's suit is, the less chance partner holds wasted values there - and hence the more chance that he holds working values in the black suits. Essentially, this improves the odds of getting the magic shortness-opposite-spots layout that produces well on offense.

[mikeh]

 WesleyC, on 2015-July-30, 00:24, said:

Yeah, good point. Aces are extremely important when defending against high level contracts.

I can easily plug a different metric for HCP into the sim and see if it makes a difference.

What about Aces = 4, Kings = 2, Queens = 1 and you need 8 points to reopen?

You have put a lot of effort into this.

I am going to set out my remaining concerns, and these are all, imo, connected with the inherent problems doing sims on these situations, and not aimed at criticizing your efforts. I just think that the tools we have aren't sufficiently sophisticated.

While I think your idea of assigning different values to the A/K/Q are a step in the right direction, no single metric will do.

I looked at your sample hands. On quite a few partner held diamond values and length. I may be kidding myself, through my desire to have double mean what I want it to mean, but my view is that partner stays fixed on those hands, and doesn't make the 'obvious' penalty double, precisely because while obvious, it isn't penalty

Having a single metric for partner to reopen means that we lose the ability to defend undoubled whenever partner hits the metric even tho irl passing is the sensible action. It's analogous to 1♥ (4♠) when we hold KQ109 in spades and no other high card. We can't double because double invites partner to pull with shape.

Another concern is the 5♦ action. Iirc, you swung from the 'too weak' minimum of KQ10 eight to 8+ in the suit and 12+ Hcp.

That 12+ hcp eliminates from your sample a myriad of hands on which most would bid 5♦: hands such as xxx x AKQxxxxxx void.

The result of giving the 5♦ overcaller a powerful hand in terms of hcp is to significantly increase the likelihood that 5 or even 6♦ will make. The result is that in your example of 100 hands, there were a large number on which we need to bid 5♥ as a save, rather than to make.

I am not saying that any of your sample hands wouldn't bid 5♦. I am saying that you have arbitrarily and inadvertently skewed the sample in favour of hands where 5♦ makes, while also increasing the relative frequency of hands where 5♥ fails. You have, imo, caused the sample population to contain relatively few of the critical hands on which no 5 level contract makes.....and on ALL such hands, passing 5♦ is the clear winner.

Thus it is no surprise that, having removed from your sample many hands on which passing is correct, you are left with a sim that suggests that passing is wrong.

The problem is almost insoluble. Lower the hcp constraint on the 5♦ bidder and you increase the number of 5♦ bids on hands on which most players would find another call. Increase the constraints and you get the problem I just described.

The only solution that I can think of is to lower the constraints for partner's reopening, lower the constraints for the 5♦ and then so an eyeball evaluation, with all of the subjective problems that that creates.

Anyway, if nothing else I think this exercise is a very good illustration of the power and the limits of simulations in competitive bidding situations, and I thank you for the effort.

[WesleyC]

For what it's worth setting up sim like this takes hardly any effort at all. I've already got templates ready to customise so it only took about 15 minutes to setup the criteria and check that the hands were approximately suitable and then another 15 minutes to analyse it. The time consuming part is arguing with you about the validity of the results

You claim that I'm somehow skewing the range of hands to exclude hands where 5D and 5H both go down, and cite an intermediate hand like [xxx x AKQxxxxxx void] as an example. However this is EXACTLY the kind of hand where if partner has too little to reopen, we risk a double game swing. For example: [JTx AQxxx xxx Qx] or [Qxxx QJTx xxx Ax]. On the hands where partner is passing out 5D, it matters very little how you divide up the HCP between the two opponents - they will often have enough to score +600.

Moving away from the simulation now and back to a theoretical argument. The fundamental flaw with your analysis of this hand is ignoring the fact that when a reasonable RHO bids 5D at these colours, they are bidding it with the expectation of making 11 tricks fairly often. Their plan might be based on HCP, extreme shape or some combination of the two. Holding a high card minimum, 4 card support and a void in the opponents suit, you have no reason to think that 5D isn't making and also a very strong preference for declaring a heart or club contract. Because of the vulnerability, even if bidding turns out to be wrong, a phantom sacrifice risks few IMPs compared to the potential gains if either 5H, 5D or both are making.

[aguahombre]

So, what you are saying then is that yes if you bid 5H partner will bid six; but you don't care because it is just a slightly more expensive sacrifice.

[apollo1201]

 WesleyC, on 2015-July-29, 21:06, said:

In the case where Partner holds 11 or fewer HCP, our side makes on average 9.85 tricks in hearts and they make 10.18 tricks in diamonds. Partner average length in hearts is 4.68.

East playing Hearts
Tricks - Frequency


9 1071
10 1403
11 1143
12 466
13 87

South playing Diamonds
Tricks - Frequency


9 1047
10 1820
11 1322
12 565
13 5

Thanks Wesly. So then, with a weak partner, we go down 2 times out of 3 at the 5 level while they make the contract 1 time out of 3, so we tend to turn a plus into a minus. However, considering your requirements for opponent's 5♦ hand were too low, it means they probably make more than half of the times (see the huge score of 10-trick hands that turn into 11 if you add an A), while we might almost always go down, but 80% by 3 or less, so a good save against the likely -600.

If partner is stronger, I believe 5♥ turns into a favorite for making (don't have the stats though), so we should bid it, preferably I understand from other comments by pulling partner's X to 5♥.

Now we have to assess the probability partner has above or below 11 (just kidding).

Overall a very interesting thread to read, bearing in mind that opponents (and partner) don't know the hands double-dummily, so could act wrongly (defending at 6♦ against 5♥) as well, and that we have a limited time to decide at the table so no time to run sims ovbiously...