[Stephen Tu]

lycier, on 2016-May-05, 04:02, said:

Very surprised, whenever I reply " No', some people often become unhappy, even sb is looking for my trouble . In a word, as long as complaining about Gibs, whatever it is correct or wrong, always is welcome, and many readers also regard it as a normal matter.
But I am unwilling to do such, No,No,No.

Lycier, it's OK to have contrary opinions. But the issue people have with you is in the manner in which you object. You are always posting a lot of irrelevant stuff thinking you are defending GIB but in reality you are not adding anything constructive to the discussion.

People post auctions where they think GIB did something clearly wrong. Where GIB makes a bid at some point in the auction where no good expert human player would make the same bid. Usually because one bid or another in the auction is defined poorly, or another bid would be more effective and should be higher priority, or something of that nature. We want a particular bid description altered, so that GIB will choose something else, or the priority of one bid adjusted higher than another, etc. The hope is that if this is fixed, if GIB has same auction in the future, it will have better choices available and bid more effectively. In cases like this, if you think the post is wrong, then you should not simply say stuff like "pity you are great wrong". That's really not constructive, and also can be construed as an insult to the human poster. Humans can be harsh when talking about GIB because machine does not have feelings to be spared. We are criticizing the bid, these are bugs, it is not intended as insult to the programmers (at least in most cases these days, in the past some posters here have I thought been overly critical of the programmers themselves, had unrealistic expectations, and in those cases I tried to defend the programmers). With human should try to be more diplomatic, and also be specific about what you find wrong about the argument. Like we say one particular bid should have altered definition, range different, or distribution different, or both, if you think we are wrong, you should say what definition you think it should be instead, or that you agree with GIB interpretation and choice, and give *bridge* reasons why. You aren't doing that, instead you are posting irrelevant stuff, so that's why we get upset with you.

You have a habit of posting all the times the board were played, where the auction was *different*. Where human bid something else, causing auction by GIB to be different. These are *irrelevant* to the discussion at hand. Looking at these other auctions do not contribute *any data at all* to whether the bid(s) on the poster's auction are bugs or not. The other auctions may reveal *other bugs*, but really they belong in separate threads, as they are *separate issues*, even though they arose from the same deal of four hands. We are trying to answer questions:
- what should bids mean, *given a particular start to the auction*
- with this hand *given a certain start to the auction*, what do you bid now?
Because this is all the data GIB ever has when choosing how to bid. It has no idea what went on at other tables. The other tables are irrelevant to the decision making process.

When you have a different start to the auction, this makes the questions *different questions*. It is as if we are discussing what the answers to questions #7 and #8 should be on a test, and then you start discussing questions #11 and #12 instead. #11/#12 might be interesting questions in their own right, and maybe related in some way, but really they should be separate discussions. When the start of the auction is different, the choices GIB have are different, it is different section of the rule database, and the information GIB has about the other hands is different. What GIB does on different auctions shouldn't affect what the choices should be on this auction.

Just because we think GIB does badly on a particular auction does not mean we think GIB is hopeless in general. We are just trying to make it better. Say you have a top student in mathematics class. He takes exam, gets 49/50 correct, top score in class. Does teacher ignore the one mistake, not mark it, not let student know he got it wrong? No, he marks it so that the student can learn and improve. We are posting bugs so that GIB can improve. We don't need to post hand where GIB does well, because GIB is not human, it does not need positive feedback so that it feels good about itself. If it was a human child we would praise when it does well, but machine don't need this kind of encouragement.

I will now take a look at your previous post in the thread and show where you are going wrong ... Maybe you will then understand better how to more constructively post in the future.

[Stephen Tu]

lycier, on 2016-May-04, 06:46, said:

However west Human suddently became crazy, why didn't west open 1♣ or 1♥?
Here, I would ask everyone two questions?
Would you agree with opening 3♣ ?
Would you think opening 3♣ is clever?

We don't care if humans do something weird. It doesn't matter what we think of the merit of a 3c bid. (I think it's not "crazy", with short spades and such suit disparity between the clubs and hearts a club preempt has some merit, if you open 1H you get to the wrong suit way too often, if you open 1c, bidding bad heart suit at high level is not attractive, and you end up bidding clubs again unilaterally or passing since you don't know if partner has spade stack; he doesn't know about your long clubs so doesn't know when to sacrifice. A club preempt will often be more effective if opponent's hand in spades. The time it works worse is if it is your hand for a heart fit, e.g. swap North and East and ew may get to a club slam instead of heart slam which is minor issue at imp but disaster at MP).

The question is what the GIB does after the action, whether *GIB* actions on the hand given the available information are reasonable or not, doesn't matter if the human actions were reasonable or not.

Quote

If your answer is Yes, Yes, now I would take a good example to show my excellent skills.

---

Result : 4♥N=
Am I clever?

So how is this relevant to the discussion? You psyched, lucky partner doesn't have spade support and instead you pick off the opponent's spades, get good result. GIB take somewhat reasonable actions (well, 2d is sketchy on suit quality, pass is better IMO), most people don't have good way to expose psychic opening and neither does GIB. What's your point?

Quote

BTY,if it opens normally, you will see how clever Gib's decision is.

---

Result : 5♥N-1
How bravooo Gibs !!!

Again, I don't see what your point is, how this is relevant to the original discussion. Also, I don't know if the GIBs are really deserving of praise on this auction. The double of 4s from north doesn't have any good definition. Really since NS have established fit in hearts, the normal meaning of this is penalty. So that double should be better defined. And South, having a very standard opener, really doesn't have any reason to pull the penalty double IMO. North shouldn't be doubling 4s with no aces, no defense, double fit heart/club.

Quote

Third issue :
Look at the "OP' hand :
Who is crazy? Whose fault?
After crazy opening 3♣, it is normal for north Gib to make takeout,ok? North Gib never become insane

No one's complaining about the first double. The complaints are whether pass is the right call over 5c (North should surely bid 6c or 6s at this point instead?), and whether it should redouble 5s on the actual auction in preference to choosing 6c. This is such a crazy rare auction that I probably wouldn't have bothered posting it, since I don't think penalty redoubles in slam auctions are something that come up enough for it to be worth the programmers spending time fixing compared to mistakes on more ordinary auctions. The call over 5c should probably be addressed though.

Quote

It just is human to become insane, obvously it just is human's fault !!!

Again, we don't care if human's bids are insane or not. It doesn't affect whether GIB's actions are optimal or not. Surely it shouldn't pass over 5c, no? And surely if EW are going to double 5s, redouble is the best call?

Then you post the 3c-dbl-4c-ppp auction.
1. Now that human pass, this is different auction, it's irrelevant to the discussion. It's a different sequence, a different bug. The fact that GIB pass south hand is irrelevant to north action when human south does not pass and west doesn't pass. We don't care whether human south has their bid or not.
2. Surely such a strong hand as North should take another call, no? Still very likely to make 4 of a major, defending 4c undoubled is unlikely best spot. I would double again planning to bid 4H if partner bids diamonds.

[lycier]

What you said are denial.
I have to say even you can't carefully read the wall of my evidences, you never accept those data, just like making some rediculous joking. In another word, I can say that you don't know how to use the hands data to study Gibs effectively. I am unwilling to say more, orelse, I am afraid it might hurt you, so would better give you some dignity.

Return to this hand of the " OP".

---

Both of 4♠ and 5♠ in complex competion tell north Gib twice that human have "5+♠, 10+hcp, 11-16TPs ", in fact, if south human really hold "5+♠, 10+hcp, 11-16TPs ", I believe declarer is very easy to make 7♠ in the almost of situations.
Who is telling a lie?
Who is the real instigator of the lie?
In fact, if your partner really tell lies repeated, you, of course, are same as Gibs, it turns out you become fool !!!
You have a famous opinion that Gib is a kind of rigid, facing to repeated lie, Gib only choose to believe partner's story, if loses 7♠ contract, all of you must say Gib is silly, this is a bug. It turns out human' true lies hurt Gib, how does Gib believe what human partner said?
If believe, it turns out it is a bug, if not, it also is a bug.
Where is the problem?
I think this issue came from its definitions including TPs evaluating approach.
How to improve its definitions?
I have no way to do it. the reasons :
If we give a quantitative definition, this 1hcp hand but with game values will show such definition is wrong.
If we give a non-quantitative definition, it turns out that it is very difficult to handle with the hands with quantitative descriptions.

Even I don't know the Gibs programming, IMO, I believe this is not a "BUG" conclusion which can resolve the problem.

[Stefan_O]

HA-HA-HA!!
You are just TOO MUCH, Lycier!
I start believing you are a ROBOT, too!!

[Stephen Tu]

lycier, on 2016-May-05, 16:19, said:

What you said are denial.
I have to say even you can't carefully read the wall of my evidences, you never accept those data, just like making some rediculous joking.

That's because your wall of evidence doesn't show anything relevant. You are focused on things that *don't matter*. It doesn't matter what other people did on the same board on different auctions. It doesn't even matter what the other 3 hands actually are!

Imagine that you *cannot* see the other 3 hands, focus only on actions of North GIB, at one particular decision point. North can see his own hand, and the auction only. Say you have that strong hand and see 3c-dbl-4c-4s-5c-? to you. Is the right call pass at this juncture????

Quote

In another word, I can say that you don't know how to use the hands data to study Gibs effectively.

You are the one that is unable to grasp the concept of looking at GIB mistakes in isolation, regardless if other people in the hand (human player, or another GIB) make a bad bid. That South make mistake on the auction does not mean North's bidding does not have bugs in it. South's bidding may absolve North of blame for the final result, but it does not mean individual bids by North are automatically all correct. We are trying to focus on what North should do assuming South is bidding normally. Even if South make complete random psychic 4s bid on doubleton, at that point of the auction, North should bid slam in spades or try for grand. Even if South being complete fool and lead to bad result, it doesn't mean pass 5c is correct.

*Unless* GIB is playing forcing passes on some auctions, which I don't think it is, and is probably difficult to program, and I think in that case North's pass over 5c ought to be alerted for human benefit.

Quote

Both of 4♠ and 5♠ in complex competion tell north Gib twice that human have "5+♠, 10+hcp, 11-16TPs ", in fact, if south human really hold "5+♠, 10+hcp, 11-16TPs ", I believe declarer is very easy to make 7♠ in the almost of situations.

Right. So why is North passing 5c, risking playing 5c undoubled, when he should be thinking about bidding grand? Say South actually had a stronger hand and 5c got passed out. Isn't North passing 5c a mistake? So passing 5c is still a mistake when South doesn't have his bid. Whether or not South has what he is promised *doesn't matter* to the question of whether North should pass 5c!

Now South doesn't have his bid here, but it's perhaps worth adjusting the range for 4s to include hands with less HCP but good playing strength. Maybe 7+ HCP and 10+TP is enough for the bid. *BUT regardless, North should be bidding over 5c, not passing!*

Quote

You have a famous opinion that Gib is a kind of rigid, facing to repeated lie, Gib only choose to believe partner's story,

I think you are applying things I have said in the wrong context. The time I described GIB as rigid I remember was only in saying GIB doesn't like to open 1nt/2nt on semi-balanced hands, and should probably be adjusted to allow open with minor suit 6322 or 5422 type shapes.

As for dealing with lies, I think most of my comments were talking about how in defending the hand, it should account for declarer possibly being off shape and stronger/weaker than bidding supposedly shows. I don't think I ever posted anything about repeated lies by partner in the bidding.

As for human partner lies/bad bids, GIB really has no alternative but to assume partner has their call and bid accordingly! Can you not see that if South human is to be believed, that North should not pass over 5c?

Quote

if loses 7♠ contract, all of you must say Gib is silly, this is a bug. It turns out human' true lies hurt Gib, how does Gib believe what human partner said?
If believe, it turns out it is a bug, if not, it also is a bug.
Where is the problem?

The problem is the pass over 5c. Gib should believe the human. The final result is irrelevant, South's hand is irrelevant. The bug is that North is almost about to pass 5c undoubled when slam is right if human actually have their bid.

Quote

How to improve its definitions?
I have no way to do it. the reasons :
If we give a quantitative definition, this 1hcp hand but with game values will show such definition is wrong.
If we give a non-quantitative definition, it turns out that it is very difficult to handle with the hands with quantitative descriptions.

It's pretty easy to improve definitions here.
1. On the 4 spade call, probably require 10+ hcp is too much. It should allow bid on fewer hcp if compensating distribution points.
2. distributional point counting can be improved by allowing more than just the shortness adjustments, can add some points for long suits, especially with such extreme shapes as South.

3. We should figure out what is preventing North from bidding over 5c. Like we should try to see what the definitions of 5s/6s/6c are at that point of the auction, and why pass is being chosen over those alternatives.

Lycier, you really have to learn the concept of not looking at the other hands to evaluate whether or not the bidding of a single hand is correct or not. Just because partner made a mistake doesn't mean your bid not also a mistake. When improving bidding in a partnership, you try to decide what the right bids are if partner has his bids. If he doesn't have his bid, then the final result and most of blame will often go to partner, but it doesn't mean you should stop trying to figure out if you also made a mistake. For improved bidding both partners must make better bids. But in this forum we don't control the humans. People aren't asking about the mistakes they may have made. So we are focusing on what GIB should have done, ignoring if the human did something nutty. We are focused on what it should choose to bid at a particular point on the auction, not on the final result, or who is most to blame for the hand as a whole.

That is why some posters here have started posting hands where you see only one hand. Because one hand is all you have to look at, and all one needs to determine if there is a bug or not! To try to stop you from posting irrelevant arguments based on looking at all 4 hands, other auctions.

[lycier]

Stefan_O, on 2016-May-05, 17:12, said:

HA-HA-HA!!
You are just TOO MUCH, Lycier!
I start believing you are a ROBOT, too!!

I am pretty glad to hear wonderful words so much.
Of course, I would better accept some expert's opinions.

Many thanks to Stefan_0, especially to Stephen Tu.

[lycier]

Let's go on discussing.

Stefan_O, on 2016-May-04, 17:28, said:

I don't think the Double over 3♣ is wrong -- generally you can double with almost any distribution 19+hcp, since you can then come back with your longest suit next round.
The problem on this specific hand, of course, is that South with a bit different hand might jump to 5♦ over the Double, expecting North to have a least secondary support in diamonds -- that would be awkward to say the least....

But on the bidding in your diagram, I think it's wrong from North to pass 4♣ -- with North's hand it must be almost certain that N/S has game in one of the majors.
North should rebid 4♥ on the second round, and South presumably correct to 4♠ -- hopefully that's not a cue-bid

Hi Stefan_0 :

You are very brave, but you still got screwed this time.
After preemptive 3♣, what's the best next for north Gib on the Gib CC ? See my example hand.

---

The solution is cuebid 4♣ to show strong two suit with 4 losers, especially points to majors, I used to call 4♣ 4-loser convention and used to call jump cuebid 5♣ 3-loser convention.
How wonderful Gib CC .
Just everyone see, turns out, my bidding sequence is most reasonable.

I am thinking if everyone can learn Gib CC seriously, where is the Bug? how many terrible misunderstandings can there be in your hand?

Any ideas?
Would you tell us your best approach?

[lycier]

Actually it is third time to meet such style hand.
First time,I can't remembered it.
Second time, I found its answer.
Third time, I saw Stefan_O's hand.
So I guess the maximum probability for this hand is 1/30,000 .Its minimum probability is 1/50,000.
Now there is a problem related :
Assume this is a real bug, but with 1/30,000 - 50,000 probability,is it necessary for the Gib programmers to fix it?