[Stephen Tu]

---

Is it not strong enough to bid 2♥? Surely this should be adjusted?

[barmar]

It's willing to bid 2 of a lower suit than yours with 11 TP, but it wants at least 13 to bid a higher suit.

I agree that this is too much, I'll talk to Georgi about lowering it.

[Bbradley62]

 barmar, on 2012-April-11, 21:43, said:

... but it wants at least 13 to bid a higher suit.

From the Dead Horse Dept: Wouldn't most people count North as at least 13 total points? Stiff A, Ax and Kx should fully retain their shortness points in addition to their HCP. Would GIB open the North hand in first seat?

[barmar]

GIB opens that hand 3♥. It wants 13 TP to open 1♥; it will do it with as little as 10 HCP with a 7-card suit, but wants compensating shortness. And while GIB doesn't use these rules, it doesn't meet the rules of 20 or 22, either. While I'm sure there are people who will open it 1♥ (you, I guess), I wouldn't, and I'll bet I'm in the majority.

But competitive bidding doesn't have to show the same strength as opening, which is why I think we can lower the requirements for this response.

The big problem with GIB's calculation of total points is that it doesn't adjust for context. I was taught similar HCP adjustments when I was first learning bridge, but later on I learned that you only make these adjustments once you've found a fit, and then you can treat each suit differently depending on how they're likely to fit with partner's hand. But GIB doesn't know how to do this, and it's not ever likely to learn it -- it's too radical a design change. Figuring out how well hands fit together is a job for simulations, but we don't allow sims to override basic auctions like this one.

BTW, every lesson on hand evaluation says that honors are more valuable in long suits than short suits. Rather than adjust the values of the honors themselves, we accomplish something similar by downgrading the short suits containing honors when calculating TP.

[broze]

 barmar, on 2012-April-12, 00:48, said:

it doesn't meet the rules of 20 or 22, either. While I'm sure there are people who will open it 1♥ (you, I guess), I wouldn't, and I'll bet I'm in the majority.

I have found the rule of 20 to be a good guide but it falls down on these single-suiter hands. Change A♠ to A♣ and 6♣ to 6♠ and suddenly it fits the criteria, which seems pointless to me. Also in a hand with such an excellent 7 card suit it makes sense to give greater weight to shortage. I would open 1♥ in seats 1 and 2 at any colours - not that I'm claiming this would be standard.

Apologies for the off-topic post.

[barmar]

My guess is that it's an opener for 10-20% of players. Anyone who generally opens light would probably not even think twice, and if they're under 20 they'll accept an invitation to game with it. But more traditional players would probably side with GIB and open 3♥.

Regardless of that, I agree that it's worth a free bid rather than a negX.

[nathan2008]

this is ok... i find more often, GIB will overbid..... but i will probably bid 3c here

[ahydra]

I wouldn't open 3H on that, ever. 4H in 3rd, otherwise 1H, regardless of vul. And the hand surely does meet the rule of 20: 10 hcp + 7 hearts + 3 clubs.

Standard bidding says that to bid at the 2-level opposite an opening you need about 10+ HCP, i.e. slightly less than an opening. So regardless of whether GiB thinks this is slightly less than, or is, an opening bid, it ought to bid 2H. I suggest you bring the limit down to 11 TP to match that of the other suits (maybe even reduce both limits to 10?)

ahydra

[Stephen Tu]

If you are only adjusting point count on short-suits only, it makes sense to only make them after finding a fit, and adjust for degree of fit, which is what old books by Goren & Hayden Truscott used to do. But I think short-suit only adjustments are inherently flawed. Long suits, especially ones this strong, have inherent trick taking power, and the 7th+ cards are particularly valuable. A lot of long suit point counts add a point for each card over 4, the 7th+ cards to me are worth almost 2, although usually I am in "trick counting sim" mode with these type of hands. Counting only short-suit points like Goren vs. counting both long suit + short-suit like say Root recommends work out nearly the same on most common patterns. But with 6+ and especially 7+ suits I find the Goren method not reflective of the strength of the hand. Especially when you are also doing things like not counting shortness value in partner's suit without a fit, which is probably a good thing in general.

I can't imagine not forcing to game opposite a partner who can open. I also can't imagine opening 3h, except perhaps vul vs. not. You have 7 tricks, by the old conservative rule of 2 and 3 you are supposed to open 4 if choosing to preempt. Or open 1. It is a rule of 20 opener, 10 hcp + 7 + 3, and has 2 QT so meets the rule of 22 also.

If it's only counting this hand as 11-12 "total points", and thus not worth GF, I'd argue that it's evaluation is really flawed. It should be around 14 IMO. FWIW the K&R evaluator rates it as 13.5, it considers 12.5 a "mandatory major suit opener". I think way more than 10-20% open 1.

It's fine to bid 2h, then 3h non-forcing if the hand was say 2623 or 3613 with the same honors. But the 7th card makes it way too strong, I don't think it close at all. We want GIB to mimic good players playing down the middle, don't we, not mainstream average (bad) ones, no?

[Bbradley62]

 barmar, on 2012-April-12, 00:48, said:

While I'm sure there are people who will open it 1♥ (you, I guess), I wouldn't, and I'll bet I'm in the majority.

Hopefully the wager was "I'll change it if I'm wrong". Please see poll results: http://www.bridgebas...age__mode__show

[pigpenz]

it was amazing at looking at what the results were for this hand.

[Stephen Tu]

I personally see merit in both 1 and 4, feel anything else is a serious underbid. But even if poll is not representative, this is actually an area where GIB itself could figure out what is best, perhaps, if one had access to test GIBs vs. each other One could:

- for this particular hand, set the hand in question as dealer, force GIB to try 1h, 3h, 4h, as opener, play it out a large # of times, see what actually works best
- in general, test GIB vs. a tweaked evaluator version that upgrades a bit more for good long suits, at least enough for it to want to force to game opposite partner's opening, see if the evaluation function is worth tweaking.

[barmar]

We actually do have the ability to run test matches between different versions of GIB -- as part of testing a new version, we run a 1,000-board match against the old, to ensure that it's actually improved. And we can set dealer parameters if we want to focus on a particular type of hand, like this. For example, one of the changes that we were going to make in v22 was to change the auction 1M-1N-2M-4M; responder bid 4M in this auction with a 3-card limit raise. The test match showed that v22 missed many games that v21 bid and made, so we ran a match that focused on these hands, and confirmed that the old style was better.

So perhaps when we get some time, we can try a match that checks this particular tweak.

[Stephen Tu]

 barmar, on 2012-April-11, 21:43, said:

It's willing to bid 2 of a lower suit than yours with 11 TP, but it wants at least 13 to bid a higher suit.

I agree that this is too much, I'll talk to Georgi about lowering it.

If he does lower it, also look into auction 1m-(1s)-2h-?-3h. This should have an upper limit, probably 12 total pts. Right now it doesn't appear to have upper limit.

I'm getting 14+ total points for both, now, and 3h is described as "strong rebiddable". I think "twice rebiddable" is enough.