[1eyedjack]

---

MP, robot tourney, best hand South.

With an air of resignation like a lamb to the slaughter I bid 3N in the vain hope that having pre-empted and upon hearing a non-forcing bid from partner, GIB might actually for once decide that it has nothing more to say.

But it was not to be. Deja vu? Absolutely. Witness:

http://www.bridgebas...post__p__757440

Anyhoo, robots misdefended and let it through, so all's well with the world, I guess.

[Bbradley62]

Georgi: Why does the description say that the 3NT bid shows 2-5 hearts?

[Vampyr]

I often wonder why people play with robots, but I have been thinking lately... The robots are always available, they won't leave, they won't sass you, you don't have to make conversation with them, they are nowhere near as good as the average real thing but are better than the lower end of the real thing... hmmm...

[1eyedjack]

Vampyr, on 2013-November-06, 13:11, said:

I often wonder why people play with robots,

Our competition on these hands is other humans sitting in the same seat as us who are playing with effectively the same partner. In that context I wonder why you find the appeal surprising.
It brings to mind my own surprise at players in duplicate events who bemoan that they have been dealt weak hands all night.
At least your success or failure in the field cannot be attributed to anyone other than yourself. There is no partner to shoulder the blame (or share the laurels).

Vampyr, on 2013-November-06, 13:11, said:

they are nowhere near as good as the average real thing

This obviously depends on whence you draw your population.
For example, if your population is drawn from those who habitually play with robots on BBO, the robots are fractionally better than average.
http://www.bridgebas...post__p__756610
Whether that is representative of "the average real thing" I leave to others to assess.
But it may be worth stressing that the hands that get posted in this subforum are a far cry from being representative of GIB's standard. We do not report when it gets things right. That requires no adjustment to its programming.

[Bbradley62]

1eyedjack, on 2013-November-06, 13:42, said:

Vampyr, on 2013-November-06, 13:11, said:

... they are nowhere near as good as the average real thing...

This obviously depends on whence you draw your population.

Agreed. GIB is clearly better than the "average real thing" found in the MBC or Express tournaments.

[1eyedjack]

Bbradley62, on 2013-November-06, 15:48, said:

Agreed. GIB is clearly better than the "average real thing" found in the MBC or Express tournaments.

Would you say that the average human combattant in robot tourneys (the actual population used) is better, worse, or about the same as in MBC or Express tournaments?

[georgi]

Explanation tweaked so it will show now 2- cards.

That will make GIB more reluctant to pull to 4major based on unknown fit, while previously it was 9+ cards promised.

[1eyedjack]

Is it so hard (or undesirable) to program GIB to insist that it passes having pre-empted (or for that matter whenever he has already completed the description of his hand), unless specifically invited (or forced)?

A similar example situation (mentioned in an earlier thread some months ago) is the uncontested auction
1N-2♣-2♦-3♥-4♥-??
GIB will ALWAYS convert to 4♠ despite that its partner has already heard that he has 5-4 and has made his choice.
(Actually this may no longer be true - not tested since version 28 released)

[georgi]

Not hard. For some of the cases it's the common behavior, but sometimes when percentage of simulating over expected behavior is too high needs to be adjusted.

Sometimes are added typical "human" rules.

Not for this hand for sure which is clear sequence, but sometimes when only human could navigate GIB to some auction and GIB is forced to pass, etc.

[1eyedjack]

It may be a language problem but my understanding of your response is that you accept that you could, if so minded, impose a rule that GIB should always pass 3NT, but that you disagree with my preference for that rule to be imposed and prefer to allow GIB the discretion to rebid. You have made some concession to my preference by reducing the frequency with which it will do so.

I agree that I have no right to insist. Mine is just an opinion, and just one in a crowd, at that. I like to think that my preference is backed by logic.

There are two example hands that gave rise to this discussion. One is at the head of this thread. the other is in an earlier thread but linked from the first post in this thread.

On the earlier hand, 3NT was superior because we had 9 tricks in notrumps but 4 losers in trumps, despite that we had a 10 card trump fit. I diagnosed this likelihood when bidding 3NT. Pre-empter could not have known this.

On the later hand 3NT was superior because I had a source of running tricks in a suit outside of partner's pre-empt suit. I required nothing from partner's pre-empt suit other than the safety of knowing that it was guarded against the opponents running it. Again, I was able to diagnose this when bidding 3NT. Pre-empter could not have known this.

These two hand types are vastly different. If (as I suggest) the pre-emptor cannot know which of these hand types has prompted me to bid 3NT (and there may be other hand types), how can it possibly decide on an informed basis whether to leave the contract there? I suggest that it cannot possibly make an informed decision, nor should it be required to: the decision has been made, by responder. Responder is not immune to error, but the decision is his alone. It may not always be the right decision, but if opener is free to reject that decision based on a complete lack of evidence then this will likely reduce the success rate of the hand.

Your solution to this has been to suggest to GIB that responder has a misfit when bidding 3N, and so (apparently) reduce the frequency of its removing 3N. It is true that when responder is bidding 3N based on his own source of running tricks, then it is likely that he has shortage in the pre-empt suit. Instinctively I feel that this hand type has a lower frequency than the first example hand, where responder expects to run opener's suit, in which case he is more likely to have a positive fit with the suit. With a misfitting hand that is simply strong enough for game but no source of tricks, responder will be more inclined to raise the pre-empt to game in that suit, for fear of communication problems in 3N.

Accordingly I remain unconvinced that the imposed solution is a significant improvement. But if I am the only one who feels that way, then I shall withdraw.

[georgi]

Even that small change reduces probability GIB to move from 3NT to almost never.

But putting 100% break is not always a solution.

With the regular preemptive hand ( supposedly with major ) - 3NT would be 99,99% of the contract.

But imagine opener ( GIB ) has a hand 1741, with :
x
AJ109xxx
Kxxx
x

Then would you be staying on 3NT with that hand or would prefer to pull to 4H for safer contract. Supposedly partner is not void.

Not all cases could be covered by GIB, but better to know it's 95% than aim for impossible 100%.

[1eyedjack]

georgi, on 2013-November-18, 13:56, said:

But imagine opener ( GIB ) has a hand 1741, with :
x
AJ109xxx
Kxxx
x

Then would you be staying on 3NT with that hand or would prefer to pull to 4H for safer contract. Supposedly partner is not void.

I just observe that with this hand opposite the South hand at the start of this thread, 3N remains by far a better contract than 4♥
I agree that 100% correct decision is an impossible dream. I just think that North can never know whether it is right to make that decision.
But I shall say no more on this, because I have nothing more to add that I have not said already, apart from one final parting shot:

I am just as keen that GIB should respond 3NT to *my* pre-empts, when that is the favourite action, as I am that it should pass my 3NT responses. It seems to me that by imposing a misfit requirement for the 3NT response you are effectively barring GIB from bidding 3NT on the first of the two example hands under discussion, should GIB ever hold the South cards in that situation, when responder actually had 3 card support for a 10 card fit and 3NT was the last makeable game. Perhaps you think that I just got lucky on that hand? That is certainly possible.

It does concern me that an undesired effect of this latest change (requirement of misfit for bidding 3N) may prevent GIB from bidding 3NT with a fit opposite a minor suit pre-empt, when 3NT is much more likely to be the right contract (than opposite a major).

[Bbradley62]

1eyedjack, on 2013-November-06, 22:25, said:

Would you say that the average human combattant in robot tourneys (the actual population used) is better, worse, or about the same as in MBC or Express tournaments?

I believe that the average human combattant in robot tourneys is better than in MBC or Express tournaments. Express tournaments are a little harder to gauge, however, because it appears that too many players simply do not realize that by registering they have agreed to play 2/1 with some specific conventions, so auctions get much more bizarre than they should, and it might be unfair to conclude that this happens due to the players being "bad".

[Stephen Tu]

GIB should not be allowed to pull 3nt on this auction, ever, IMO. In standard bidding, 3M-3nt and 2M-3nt are not asking partner's opinion. Partner may have void in major, stoppers, running suit of his own. He knows preemptor's hand a lot more than preemptor knows the 3nt hand. The only auction where preemptor can pull 3nt to 4M is 2M-2nt-?-3nt-?

Human with good hand and enough controls for game should be raising to 4M instead of 3nt on most hands with 1+ in M, and even some hands with void in M, if they have any clue what they are doing. 3nt is for hands with no fit, stops, and source of tricks other than M, just count on partner to stop M.

[1eyedjack]

Stephen Tu, on 2013-November-19, 13:23, said:

3nt is for hands with no fit, stops, and source of tricks other than M, just count on partner to stop M.

*OR* a high LTC hand, typically with no ruffing potential in the short trump hand, which has a good chance of running 9 tricks in NT via running opener's suit. Note that with such hands responder will normally have a fit with opener, - often a good fit.

Partner opens 3♥ and you hold

♠QJx
♥Axxx
♦Axx
♣Axx

What do you fancy: 3N or 4♥ (or even pass 3♥)?

[1eyedjack]

Here is another recent example, from a MP robot tourney, best hand South

---

As a matter of side interest, around half the tables simply raised 3♥ to 4♥, which I think is correct on this hand, pass being my second choice and a close one at that. At my table I chose to pass, which was singularly unsuccessful, and with the benefit of hindsight have changed my mind. Anyhoo, after 3♥ is raised to 4♥ East "balanced" with 4♠(!!!), got doubled for 1400 or so. So making 4♥ or 3N was only competing for about 50% of the matchpoints.

Leaving aside any issues arising from the 4♠ balance at other tables, I should be interested if Georgi would comment on whether, AFTER the fix mentioned in this thread

(a) Would GIB, sitting in the South seat, bid 3N?
(b) Would GIB, sitting in the North seat, pull 3N?

This hand is I think quite interesting because
(1) 3N is the right contract (particularly at MP), despite which I would not recommend South bidding 3N as a long term winning action. There is insufficient evidence that the Hearts are likely to run for zero losers, without which the hand lacks control.

(2) If ever there was a hand in which pulling 3N by North is justified, this hand may be it, and it got me to thinking, perhaps North SHOULD have some discretion, but only exercise it when his trumps are solid. It is ironical that the solidity of North's trumps is what makes 3N a good spot in this case. And yet if South bids 3N without knowledge of that solidity he cannot be bidding it with the expection of running the Hearts without loss. Ergo, he should not have that hand type when suggesting 3N. He COULD have a suit that he expects to run of his own, but that is probably North's void suit, in which case it may well not be running, and with a shape as extreme as 4-7-0-2 it may well be that 4♥ provides the better prospect.

I rather wonder whether North is perhaps a little strong for 3♥? Given GIB's tendency to pre-empt with a Jack-high suit on occasion? I read an article by Marc Smith several years ago arguing against disciplined pre-empts, on the grounds that yes, responder is placed in a difficult position, but there are two opponents for every one partner. Increasing the frequency of that arising would appear to be in your favour. I was not entirely convinced by the article or argument, because it fails to take into account that if the pre-empt is disciplined the opponents are just as much in the dark about their combined values as if it is wide ranging. Jury's still out in my head on that one.