[Jyrki_63]

Hi,

First time posting here. Hopefully including the link to the hand works out.

Misfit hands are dfficult to bid under the best of conditions (=with extra strength), so why is GIB overbidding without any known fit? After the latest upgrade the "2-suited invite 10-12" has occurred at least twice. On both occasions I would have evaluated the hand closer to 7 as a responder.

2-suited invite

Here 4C seems quite a stretch. Given that I had about an ace more than the bare minimum for my bidding GIB only went down 2. That was only -0.27 IMPs, because the field was in the same spot. Yet it seems ill-advised to go beyond 3NT to simply show one's shape with a hand that has very few HCP and no known fit. In the case were openere has showed the other two suits GIB is doing it in spite of the evidence telling that no fit exists.

I realize that it is very difficult to design a simple program that could bid reasonably well. But many a bidding sin would not be committed, if GIB simply refrained from adding points for shortness unless an 8-card fit is located (or it owns a solid 6-7 card suit itself). Why does GIB thank that the example hand is worth pushing to the four level given that I am expected to have at least 9 cards in the reds? If it wants to invite, why not 2NT?

[georgi]

Thanks for reporting it.

For GIB 2NT would be 10-12 HCP and invite.

Going up to 4th level as misfit seems risky indeed. As the idea for stop there or pick 3NT must remain as choices.

Looks more reasonable if the invitational call goes there to be rebidded the lower one suit, like 2♠ here or return to 2♥ with 2 cards.

It requires tweaking for sure, so it will be corrected for the next update. If not all, at least most common sequences.

[calm01]

I agree with jyrki-63 that GIB sometimes does not wait for a 8 card fit to revalue its hand. This is a potentially fatal flaw for its bidding.

Fixing some sequences will help in the short term but will not solve the underlying problem. So the proposed fix will add more complexity to its code and make further fixes that more difficult, more risky, more costly and slower to implement and require more skill.

Such fixes therefore bring forward the time GIB will need to be retired.

Doing the job properly - teaching GIB basic bidding principles - is cheaper in the long run assuming BBO want to stick with GIB come what may.

But I feel sure that BBO knows this but continues to take repeated short term views to address the users frustration. The long term cost of this approach is very high. May God have mercy on the BBO computing soul as BBO appears to be on a path to a software development buffer with no reverse gear.

[barmar]

calm01, on 2011-December-26, 20:11, said:

But I feel sure that BBO knows this but continues to take repeated short term views to address the users frustration.

You don't know what our long term plans are, and we won't discuss them here.