calm01

The bidding was 1C 1H X pass !N all pass GIB had passed his partners !N with 10 3 2 K J 9 8 7 6 4 J 2 J What went wrong with the simulation?

Barman, I like the idea of GIB simulations being equivalent to judgement calls. This makes sense as it seems encoded logic is employed - do a simulation - when no described bid seems to fit the hand. Humans make judgement calls when their bridge logic is not up to the job - it is often called taking a 'punt'. Beginners often make more faulty judgement calls as they have less bridge logic to fall back on, or worse hold inferior bidding rules compared to more experienced players that gets them into bidding problems more often than more experienced players. Here the underlying problem appears to be failure to encode better bridge logic - for example bidding only good 5 card suits when responding to a 2C opener is perhaps that missing bridge logic. Thus depending on simulations would be required less - the equivalent of making GUB a more experienced player. GIB would need to punt (simulate) less oftem. Time to add some better bridge logic into responding to 2C openers.

The 3S bid by the robot suggests four questions to me. The first three are: - does it make bridge sense for to mention other than decent 5 card suits (Q10xxx) or better as responder, as the partnership is likely to be in the slam zone in which case even the description promising a 4 card spade suit is deficient - but I guess that is a matter of partnership style, - does it make send to bid any three card suit (even AKQ) when partner may support with 4 to when playing in game will likely be in the wrong strain, - why suppress a decent 6 card suit with enough values exist to play in 4N unless the system always plays 4N as Blackwood. The fourth question is if GIB is actually showing partial stops for 3NT, not a suits at all. Not my preferred bidding style when partner has failed to rebid in NT but has some limited merit. My conclusion is- until persuaded otherwise by wiser souls - is that GIB is just wrong with its bid and/or its description. Perphaps GIB has been on the GIN and that will also explain the delays other are reporting!

Uday, Ari and team. The large potential value of this forum to Bridgebase is again demonstrated by the willingness of the members to spend time, effort and risk of embarrassment communicating with the support team. And the quality of that members communication is of a high standard. Whatever the outcome of this particular problem, I know how difficult it can be for one group of people to have empathy with another groups experience. There is always a difference in: - viewing point in space, - viewing point in time, - knowledge, - roles, - effort required to communicate, - openness to being wrong and authority required to communicate, - ... This is probably the support teams best recent opportunity to demonstrate rapport with its forum members. Fail and the forum world will judge unfairly - and the forum feedback will diminish; succeed and the world will love you for a while - and the forum feedback will improve further. I am betting you will select the latter path as your future prospects partly depends o the nature of your rapport. And rational support teams usually act in their long term best interests.

Bbrafley62, Good idea about bringing attention to GIB behaviour in the documentation. While Udat and the team manage the bridge bidding expectations regarding the GIB handling of the GOSH, there is also the common failure to appreciate understanding that almost all GIB doubles are for takeout. I have lost count of how many times I play in something like 3C -1 or -2 after honouring partners takeout double with my 1 point 4432 or 4333 hand for what would normally be a poor or average result only to find half the competition have defended 2H X making (or similar with or without overtricks) to give me +6 or more undeserved imps. Updating the system definition including a a date of last change requires a marked change in documenting behaviour of the support team. The trouble with such change is that sometimes such alterations in behaviour are seen as admission of failure of past behaviour of the team colleagues. Improvements thus sometimes only occur when the ego is relaxed. Uday and his team surely do not suffer from this problem.

Part of my job as a software project manager was managing user expectations. So Uday, if you want your frustration at unrealistic expectations of improvement to be lower then: - post a list of known unfixed bugs with an ABC priority for them, - post a list of recent bug fixed and the latest GIB release number, - ideally ask for vote on relative priority (ABC) from members of this list. Explain that the members vote is a consideration but so is available resources, risk assessment, frequency of occurrence/impact and the skills required for fixing each bug. In other words engage with the community even more than you do and you will have a more supportive and helpful community. I for one have stopped logging bugs which are not high occurrence - I suspect others may have have frustrated into giving up earlier. So I applaud the original poster for still being willing to take the time to share. Without such increased engagemen,t the GIB forum will be mostly limited to new contributors reporting known bugs who in turn get frustrated and give up. This does not do you and the teams effort justice - does it?

Your view of little happening to GIB is perhaps both unkind but perceptive. My perception is that the developers are reluctant to engage on a major upgrade of any kind, preferring a small tweak here of a simple test there. You will note that there has not even been a reply to my requests for a change log and GIB version number. A simple addition of a fourth pinned discussion would be the ideal place for such a bug fix/improvement log. I have previously been reluctant to suggest how the GIB version and change log could be implemented because telling how to implement something can demotivate. Over 20 years ago i was a software team project manager. Based on this experience and the nature of developers recent responses, there are three probable problems: - there is little or no budget from Fred for the purpose of GIB improvement and the guys are doing their best in this context, - the code is so opaque that even undertaking a minor change is a high risk or significant task, - there are competing requirements for GIN that leave the developers in almost a no-win position. If even one of these apply to some degree then we must have patience until the time is right for GIB improvement. Perhaps Fred needs to be be taken up on his request for feedback.

I agree - in 2/1 fame force you normally bid you distribution when in a forcing to game situation if a fit has not been agreed. So i would take it as clubs and hearts and since neither NT or fit is shown by responder it must be must be at least 5-4. There are lots of suspect descriptions - but I do not think this is one of them. Ari recently indicated that correcting descriptions and associated coding will improve GIB bidding - so it is worth continuing to report suspect descriptions

3 robots and the bidding was: S W N E P 1N 2D 2N P 3C 3D .... 3D was sescribed as "4+D; 5+H; 5+S; 18+ total points" It is amusing to play with a partner with 14 cards. Do you feel that th effort needs to be expended to ensure the descriptions make bridge sense?

This reply is of the form of tough love. My experience is similar to that of dennis708. I can recognise no pattern in GIB behaviour in giving preference. The worst case is preferring a 4-1 fit in spades to a 6-5 fit in diamonds. From what I have inferred from Ari and Barmar helpful responses is there is much hope for improvement in many areas of GIB bidding and play. A small tweak here and tweak here and there could make a bid difference apparently. But until a small group of programmers and bridge players take it in hand and put in the required effort to exploit the opportunities for substantive change - on a regular and structured basis - that potential change goes mostly unrealised. My current approach is that i enjoy the 24 hour availability of playing with 3 robots as I can answer the 'phone or make a sandwich etc., without having to worry about upsetting three human players. So I accept the poor bidding, palooka defence and erratic declarer play. I make adjustments to my bidding as I would with a poor/erratic human player. I choose to laugh rather than to cry by making the best of my partners foibles. But I would appreciate knowing when my palooka partner has improved in one area so I can stop compensating for his/her poor bridge education when he/she has learnt something new. I need to know when the version has changed and what has changed, This is essential feedback. I also need to see some steady improvement not to lose heart. The version number and bug fix list is essential feedback and appears to be missing for GIB changes. Arigreen and Barmar please note is is not enough to try and improve the GIB program - you need to communicate the version and bug fixes as you implement them. With this encouragement I hope that your obvious enthusiasm and commitment will translate into more achievement as the potential your imply is realised and communicated.

Barmar, Thanks for the helpful explanation. But not mattering in a simulation is not an excuse for sloppy play. Perhaps, the programming needs to be altered so that no-lose plays override not mattering in a simulation. In this way an entire class of poor plays would be eliminated. And players would not have to trouble you again with an entire class of poor GIB behaviours.

North is the dealer (vulnerable against not). The contract is 3MT bu South after an auction of: N E S W 1D 1S 3NT all pass N, E, and West are robots. The lead is the 3 Spades (from 743). Dummy (N) has: K10 108 AK832 AJ65 The 10 is played and East holding QJ852 AJ4 Q4 Q103 plays the two! There is never any lie of the spades where allowing the ten to be a trick and giving declarer a tempo is the correct play. This is palooka level of defense. Surely the computer can do better. Ari, fixing this issue will probably eliminate a whole class of defensive errors made by GIB.

Barmar, I agree about human pattern recognotion taking many aspects of hand evaluation into account. but if you were to limit GIB to just two hand evaluation techniques I for one would vote for: - HCP for no-trump - losing trick count (less one for 3 Aces) for suit contracts with a 8 card or longer fit. You may have your favourite two techniques - why not apply them to GIB? At lease two techniques are required because: - HCP is very accurate for notrump as it distinguishes just one point difference for 3MT (25) and 2NT (23-24), - just using HCP can be a poor guide for suit contracts with an 8 card or longer fir and even less relevant when no fit is found where a measure of playing strength can be more important. I sispect Ari's test of 100 hands suffered from using HCP for those hands which wer notrump (little problem perhaps) and also for suit fir/non-fit hands where it would be a poor guide. One size does not fit all -at least two techniques are required - one for each of notrump and suit-fit/ suit non-fit. I suggest three techniques would pay even higher dividends as losing trick count for suit non-fits does not do a good job. What do you feel is the best way forward to improve GIB bidding?

Ari, Thanks for trying.

How will we know that FIB has been updated?

TOTAL POINTS appear to have little brudge meaning. Before a fit is found total points are nearly pointless. After a fit is found controls, HCP, or losing trick count are all good guides. If no fit is found total points are entirely pointless. If notrump is the selected final contract high card points are probably the best guide with playing strength a secondary consideration. Perhaps the total points concept is a poor mans' playing strength estimate. Ari, if you cannot determine any practical value for total points, just remove all reference from the descriptions and remove all coding references to total points. This might improve the bidding judgement of GIB at one stride.

A GOSH (good one sited hand) that most bridge players understand is shown by a double of opponents suit opening and a subsequent bid or even jump bid of the good suit is not in GIB's vocabulary. I accept this and just simply overcall at lowest level or gamble 4 of major directly on a wider range of hands when playing with GIB than would with all but the ramkest palooka. Surely it would not be beyond Bridgebase programming skills to enable GIB to unlearn its initial understanding of tolerance for the three other suits when meeting a GOSH bid by partner. But correcting this commonly occurring and basic bidding mistake is not sufficient - you need to tell players about the change in GIB skills or the change will not have much value. So I asking for two high priority changes: - a clear indication of the release of GIB and am accessible list of recent bug fixes, - the ability to recognise a GOSH. Thanks for listening/

Playing with 3 robots, the vulnerability was N/S red and E/W green with West the dealer. In an uncontested action West (robot) opens with 1diamond and all pass. The robot East has passed his robots partners opening bid holding: Q7 K2 A 10 3 AK9863 12 tricks in diamonds were duly made by declarer. This is not a case of one wheel coming off the bidding wagon - all four fell off at the same time. How is this to be explained and fixed?

Both non-vulnerable, the bidding was: N E S W 1C x xx 1S P 3S 4C P P x P P P The double of 4C by a Robot was described as: "Takeout doUble 2- C; 3-5D; 3-5 H; 4- S; biddable S; 18+ total points;" West, also a Robot, passed this takeout double" and the contract was made. When you have a suit bid (although forced) and had it raised, a double of a competitive bid by opponents by either partner is for penalties as you have the choice of passing, bidding on in your agreed suit or bidding another suit or notrump. The description makes no bridge sense as being for takeout. So which Robot is wrong, the one with the strange description or the one that ignored the description? Currently playing against 3 Robots is a bit like Riuber Bridge against three palookas. While winning is nice as you have only one palooka partner but two palooka opponents, would it be nice to have a better challenge perhaps?

Sometimes the Robot cardplay defies any logic I can devise. Maybe you know better. The bidding was: 1D P 1H 2D 3C p 3N all pass The lead was the Queen of hearts. Dummy goes down with: AK95 J AJ ($ KJ84 Third hand (robot) opponent holds: 102 1076432 - 109652 plays the 10 of hearts!. This is handy as I hold AK985 hearts. Can you devise any reason for the robots card play other than random?

in 3C the robot declarer spade holding was 87 in hand and Q105 in dummy. Dummy has a guaranteed entry in trunps but that is another story.. Declarer led a spade towards dummy and played the Q when the 7 was covered by the 6. The Queen loses to the King. Winning the return, another spade was led towards the 106 in dummy, On the Ace the 10 was thrown. What a waste as my Jack, now singleton fell under partners Ace. One can question the wisdom of playing the Queen on the first round, but it makes no sense to play the ten on the second round. If it was a human declarer one would think of naivety or grandstanding - as it was a robot then its card play leaves something to be desired. Surely the lowest of losers must be standard choice unless unblocking or endplaying comsiderations override.

How true - when playing with a partner it pay to play partners system! But, in respect of doubling first with a strong one-suited hand, GIB does not seem to play "bridge as we know it"! As I have said before, in some ways GIB is so much more human than an emotional pick-up rubber bridge partner that one quickly learns to manage to become profitable. Takes me back to my youth! http://forums.bridgebase.com/style_images/1/icon1.gif

How true.

Bbradley62, You're right, I was confused about the actual vulnerability. javascript:emoticon(';)') Vulnerable, the double is perhaps a bid without purpose or carries too high a risk for the probable reward and so the best description might be "unused" when vulnerable. But the GIB bid descriptions do not seem to be in format that can suggest often 4-4 but maybe 4-3 in the reds and 11 points if 4-3 and 9-11 if 4-4. Nor do they seem to vary too much by vulnerability. The format on length nearest to what we both agree the non-vulnerable bid promises would seem to be 3+H, 3+D. This does not convey what we both agree that the bid promises. So in the restricted format used in GIB descriptions 4+H 4+D is perhaps the nearest it can convey to our common understanding. Given that restriction on length, description suggesting 11 points may be a good compromise. I ignore the total points component of the bidding descriptions but perhaps this is off-topic. So again "4+D; 4+H; 11HCP; 12 total points" makes bridge sense to me given the apparent limitations of GIB bid description format.

Seems fairly straightforward to me as we are non-vulnerable. If we were vulnerable the double fails to have any meaning for me. The clues are there: - I have passed twice so am unlikely to have opening values, - partner passed before so probably does not have opening values or 10 cards in the two red suits with 10+ points, - both opponents have limited their hand with bidding suggesting likely 8/9 club fit, - I am being offered two/three suits to select from at the two level, - partner cannot have the values to double holding more than 1 club so the opponents in this instance have a 9 or 10 card club fit, - for us to have near to, or actually have, the balance of points we must have a total of say at least 18 points with a very high probability of a fit. Based on this very common competing logic, it is only worth reopening in this non-vulnerable situation if the doubler has 9 points with a good 4 or decent 5 spades and 4-4 in the reds or 10/11 points if holding only 4-3 in the reds. So the description "4+D; 4+H; 11HCP; 12 total points" makes bridge sense to me. It is roughly what i would expect from a good partner, being non-vulnerable, playing duplicate pairs or any high standard game where the opponents can be trusted to have bid soundly. If the opponents are poor it would be too risky to double.