GIB takes false preference with minimum and five card support

[wbartley]

Just play bridge robots. Human South.

 

This has probably been reported before.

 

[hv=pc=n&s=skqj3hat7dkqj2cjt&w=s54h865da43cq8754&n=st6hkj93d8765c632&e=sa9872hq42dt9cak9&d=n&v=e&b=9&a=p1sp1np2cp2sppp]399|300[/hv]

 

Since 2♠ invariably ends the auction, why doesn't GIB pass 2♣?

[pescetom]

Agree with your comment.

But think the human in south who passed instead of bidding 1NT needs even more urgent fixing :)

[virgosrock]

Agree with your comment.

But think the human in south who passed instead of bidding 1NT needs even more urgent fixing :)

 

i also agree. why did human south not bid 1N?

 

vrock

[jdonn]

Even if passing 2C is better than bidding 2S, which is far from sure, this would not qualify as a bug, just a slight judgement matter.

[Stephen Tu]

Even if passing 2C is better than bidding 2S, which is far from sure, this would not qualify as a bug, just a slight judgement matter.

 

It looks pretty sure to me. Would you bid 2S? I think taking false preferences can be taken too far and this is over the edge.

 

My older version PC GIB simulates and passes, and doesn't seem to think it's particularly close (it would rather raise to 3c than false pref to 2s).

 

And just because it might LOOK like a judgment matter, doesn't mean that it automatically doesn't qualify as a bug.

 

In my view if it automatically bids 2S with 2S/5C regardless of hand strength (I'd want 8+ HCP personally), no simulation, that is a bug, especially if it is not doing this out of a deliberate decision by the programmers to change its behavior. If before it was considering pass, but now it is not considering pass because of a side-effect of some other rule change/addition, that is counter to the programmer's intent and thus a bug even though the symptom looks like "bad judgment". Now if you guys specifically discussed this sequence, and thought 2 spades is automatic on even the worst 2/5 blacks, and this is an intentional change, then it would not be a bug by definition, but in my view would be very questionable design.

 

In any case I strongly suggest that it's worth the programmer looking into even if you think it's a "judgment issue". GIB doesn't appear to be using any judgment here, it is not considering passing an option. Maybe it's a basic vs. advanced bot thing? But on my PC version the book bid is pass as it should be IMO.

 

Also experimenting a little it seems it wants to raise to 3c on as little as xx xx QJxx Kxxxx, "8+" total points is enough to raise. Anything less than that it passes 1S, then once you get over the line to respond with this shape it wants to raise, no range wants to just pass. Seems quite a lot to me especially after spade opening with no "impossible spade" to distinguish courtesy raises from inv raises. I'd probably want something like min 9 HCP, 10-13 total points?

[virgosrock]

It looks pretty sure to me. Would you bid 2S? I think taking false preferences can be taken too far and this is over the edge.

 

My older version PC GIB simulates and passes, and doesn't seem to think it's particularly close (it would rather raise to 3c than false pref to 2s).

 

And just because it might LOOK like a judgment matter, doesn't mean that it automatically doesn't qualify as a bug.

 

In my view if it automatically bids 2S with 2S/5C regardless of hand strength (I'd want 8+ HCP personally), no simulation, that is a bug, especially if it is not doing this out of a deliberate decision by the programmers to change its behavior. If before it was considering pass, but now it is not considering pass because of a side-effect of some other rule change/addition, that is counter to the programmer's intent and thus a bug even though the symptom looks like "bad judgment". Now if you guys specifically discussed this sequence, and thought 2 spades is automatic on even the worst 2/5 blacks, and this is an intentional change, then it would not be a bug by definition, but in my view would be very questionable design.

 

In any case I strongly suggest that it's worth the programmer looking into even if you think it's a "judgment issue". GIB doesn't appear to be using any judgment here, it is not considering passing an option. Maybe it's a basic vs. advanced bot thing? But on my PC version the book bid is pass as it should be IMO.

 

Also experimenting a little it seems it wants to raise to 3c on as little as xx xx QJxx Kxxxx, "8+" total points is enough to raise. Anything less than that it passes 1S, then once you get over the line to respond with this shape it wants to raise, no range wants to just pass. Seems quite a lot to me especially after spade opening with no "impossible spade" to distinguish courtesy raises from inv raises. I'd probably want something like min 9 HCP, 10-13 total points?

 

Agree with you Mr. Tu.

Though GUB's addled innards have this in built rule for cases like this - have seen it multiple times - if is "marginally" shy of bidding 3C will try to keep bidding "alive" with 2S. many many many times have i seen this. after all 2c has a wide-ish range.

 

vrock

[wbartley]

I agree this isn't a bug. Neither is it an error in judgment. It's a misconfiguration. With two spades and five clubs, without some special agreement that 2♠ doesn't constitute a preference, bidding spades shouldn't even be in the list of options.

[Stephen Tu]

I agree this isn't a bug. Neither is it an error in judgment. It's a misconfiguration. With two spades and five clubs, without some special agreement that 2♠ doesn't constitute a preference, bidding spades shouldn't even be in the list of options.

We don't know if it's a bug or not. Whether it's a bug depends on programmer's intent. It's either a bug, or "by design". If it's by design I question the design. The fact that earlier versions pass suggest to me that it is more probably an unintended side effect of another rule change, and thus a bug, than a deliberate design choice, especially since I've never really seen anyone advocate false pref with this big a length discrepancy on this minimum of a hand, so it's hard for me to believe that it was deliberate. That it sort of looks like a judgment error from a human point of view doesn't determine whether it's a bug, since computers have designed behavior to mimic human judgment, and it is either performing as intended design or not.

As for whether false preference should be allowed on this shape, to me that is up to debate. I personally think it's OK, with 2-5 shape to have rules that say:

• pass with ~6-7 HCP
• false preference with ~8-9 HCP
• raise with 10-12 HCP

or some approximation of this.

 

My reasoning is you don't want too wide a range for the 3c raise, and with 8-9 you want to give partner a chance if he is strong 6-4 or 5-5, 16-18ish. Possibly the 8-9 HCP will give you enough strength to scramble 2S home on power enough times, so that the games you find by keeping the auction open compensate for the stupid 5-2s going down when you have a 8/9/10 cd club fit that makes.Gib definitely should not raise with 8-13 total points as it apparently is currently designed to; that's too wide a range to handle intelligently IMO. Raising on xx xx QJxx Kxxxx is way too aggressive. Now it's POSSIBLE that the false pref on 2-5 shape with 8-9 range is a bad strategy, and that one should just pass with 6-9 TP and raise with 10-13, or some other split of the range, and have no false preference range. That is something it would be nice for the programmers to test, try a never false pref strategy vs one that does, and see which prevails in the long run on these auctions. It's possible the answer differs between IMPS/MPs, though it's more practical to just pick one strategy for both.

[helene_t]

Since 2♠ invariably ends the auction

It doesn't. If E has say 16-17 points, it will bid on.

 

This could be a reason for taking false preference. With this hand, I would pass in order to avoid a 2NT or 3♠ bid by East, since I wouldn't accept any invite so I might as well stay low. But maybe GIB judges differently.

[zhasbeen]

Just play bridge robots. Human South.

 

This has probably been reported before.

 

[hv=pc=n&s=skqj3hat7dkqj2cjt&w=s54h865da43cq8754&n=st6hkj93d8765c632&e=sa9872hq42dt9cak9&d=n&v=e&b=9&a=p1sp1np2cp2sppp]399|300[/hv]

 

Since 2♠ invariably ends the auction, why doesn't GIB pass 2♣?

 

As far as I can tell this problem either can't be fixed, or will NOT be fixed. Without exaggeration I've seen more than 100 sequences like this one. I call it the "5-2 magnet".

 

We need a new Robot!!

[virgosrock]

As far as I can tell this problem either can't be fixed, or will NOT be fixed. Without exaggeration I've seen more than 100 sequences like this one. I call it the "5-2 magnet".

 

We need a new Robot!!

 

To repeat zhasbeen:

1. GUBBO is a tad short of 3C

2. In such cases it prefers no to drop the bidding and make a temp pref for spades.

3. There could be a ton of hands where opener may have 16 hcp but cant open 1nt and cant jump.

4. I am ok with this treatment as long as known.

5. might even give a good MP score LOL not that GUBBO cares.

 

vrock