[Stephen Tu]

Auction:
1♣-p-p-1♥;
1♠-2♥-p-p-;
3♣;

3♣ showed "4+ sp; 4+cl". Instead of (4 sp; 6+ cl). Partner took an ill-advised preference to 3 sp with 3 sp - 2 cl, down 2 instead of making.

Seems like opener rebidding the minor freely in comp without partner raising should be unambiguously 6+, I've encountered several spots like this where it only showed 4 or 5, leading partner to take adventures in finding other trump suits.

Would be nice if bypassing a chance to bid spades twice denied 5-6, though I suppose that could be harder to program.

[pigpenz]

I had the same problem dont know if it were the same hand
all the lucky people just kept bidding ♣ till the cows came home

---

[barmar]

GIB has more than 20 bidding rules that match this auction. Some show long clubs, but not all, and the explanation of the bid is the least common denominator of all of them.

BTW, GIB thinks the way to bid the South hand is 1♣ ... 2♣ ... 2♠. But it also thinks this only shows 4+ ♣.

I think the problem in both cases is that there are some low-priority default rules that don't check whether the club rebid is restating a suit that was bid previously -- the same rules would match 1♦ ... 2♣.

[Bbradley62]

pigpenz, on 2012-April-03, 16:12, said:

I had the same problem dont know if it were the same hand

---

Same hand. What're the odds?

[Bbradley62]

barmar, on 2012-April-05, 07:06, said:

BTW, GIB thinks the way to bid the South hand is 1♣ ... 2♣ ... 2♠. But it also thinks this only shows 4+ ♣.

I think the problem in both cases is that there are some low-priority default rules that don't check whether the club rebid is restating a suit that was bid previously -- the same rules would match 1♦ ... 2♣.

Does that mean this can easily be fixed? What hand with only 4C would rebid 2♣?

[barmar]

Bbradley62, on 2012-April-05, 14:41, said:

Does that mean this can easily be fixed? What hand with only 4C would rebid 2♣?

None would ever bid it, because there are higher priority rules that would produce some other bid. It's coming from rules that would be used to interpret bids by other players. The rule essentially says "If someone balances in clubs, they show 4+ clubs" -- it SHOULD say "If someone balances in clubs, and didn't previously bid clubs or their partner didn't raise clubs, they show 4+ clubs".

[Bbradley62]

barmar, on 2012-April-06, 08:19, said:

The rule essentially says <something>...
it SHOULD say <something else>...

At the risk of sounding repetitive... Does this mean you're going to fix it, so that explanations reflect reality?

[barmar]

Probably, but until Georgi posts "this is fixed in the next version", I don't make any promises. There are lots of things on his plate, he can't fix everything we identify right away. I can identify problems like this, but fixing them is his bailiwick.

To understand problems like this in general, consider the following in a generic programming language (this is NOT how GIB bidding rules are written):

if X == 1 then A
else
if X == 2 then B
else C

Any human can tell than if you get to C, X can't be 1 or 2. But GIB is unable to figure this out automatically, so to make it understand this, we'd have to write something analogous to:

if X == 1 then A
else
if X == 2 then B
else if X > 2 then C

But we have to be very careful, because there could be another set of rules like:

if Y == 1 then D
else
if Y == 2 then E
else C

Both decision trees include C, but the first has implications about X, while the second has implications about Y. So to fix this, we have to split C up into two rules, one for the X tree and another for the Y tree. In the case of default rules, there may be quite a few decision trees that lead to them.

[cloa513]

barmar, on 2012-April-06, 10:53, said:

Probably, but until Georgi posts "this is fixed in the next version", I don't make any promises. There are lots of things on his plate, he can't fix everything we identify right away. I can identify problems like this, but fixing them is his bailiwick.

To understand problems like this in general, consider the following in a generic programming language (this is NOT how GIB bidding rules are written):

if X == 1 then A
else
if X == 2 then B
else C

Any human can tell than if you get to C, X can't be 1 or 2. But GIB is unable to figure this out automatically, so to make it understand this, we'd have to write something analogous to:

if X == 1 then A
else
if X == 2 then B
else if X > 2 then C

But we have to be very careful, because there could be another set of rules like:

if Y == 1 then D
else
if Y == 2 then E
else C

Both decision trees include C, but the first has implications about X, while the second has implications about Y. So to fix this, we have to split C up into two rules, one for the X tree and another for the Y tree. In the case of default rules, there may be quite a few decision trees that lead to them.

Why do you have two rules for the same starting sequence?
Why not one rule per starting sequence?


By the way, I think
1C,-,-,(1H),1S (should reasonably show 4 clubs for strange hands like 4144 but usually 5 clubs)
If it was
1C,-,-,(1H),2C,2H,-,-,2S is 5+ clubs and 3 spades for a strong hand with 3 spades AKx.

[barmar]

cloa513, on 2012-April-06, 15:33, said:

Why do you have two rules for the same starting sequence?
Why not one rule per starting sequence?

Because it's not practical to list every sequence in its own rule -- there are too many possibilities. So to the extent possible, we try to write rules that reflect general principles.

This is the whole problem of writing a bridge bidding program: how do we translate what humans do so naturally in terms of applying a handful of principles into computer language? It turns out to be extremely difficult (AI researchers were similarly surprised in the 60's and 70'd to discover how hard it was to write computer programs that could understand natural language -- they initially assumed it was just a matter of writing the appropriate grammars and dictionaries, but much goes unsaid due to assumed common understandings).