[CoreyCole]

In a tournament the other day, the auction went 3C on my right, Dbl by me, all Pass. I expected to find my partner with a club stack, but here were the actual hands:

9 8 7 5 3
Q 9 6 5
6 3
A 6

K 10 6 4
10 8 7 A 4
Q J 9 7 5 4 K
Q J 7 K 10 9 8 5 4 2
A Q J 2
K J 3 2
A 10 8 2
3

Why would North pass the double with that hand instead of responding 3S with his 5-card suit? I have had this happen on at least two previous occasions. This time I took the time to record all four hands to I could report it. The occasion was an ACBL MP robot duplicate tournament.

[BunnyGo]

Here's the link.

[jamegumb]

I posted two of these hands in the past weeks:

http://tinyurl.com/7aa7q5b
http://tinyurl.com/72hj9d9

Yes, this is a known GIB "feature".

[barmar]

I'm not sure what happened. The book bid is 3♠, and in my simulation tests it considers Pass, 3♥, and 3♠, but always chooses 3♠. But I see that the traveler is full of 3♣X contracts.

[jamegumb]

If this is indeed a bug, I'd sure appreciate some sort of fix. This is far and away my number one peeve playing with the robots.


(Actually, it may be my only complaint; well, outside those times it follows a takeout bid by bidding a minor and then later a major which it describes as having 4+ of the minor and 5+ of the major, even though it often has only 3 cards of the latter.)

[cloa513]

Obviously jamegumb you haven't been playing enough with GIBs- read the other threads.

[jamegumb]

cloa513, on 2012-February-15, 03:58, said:

Obviously jamegumb you haven't been playing enough with GIBs- read the other threads.

Oh, I've played plenty with bots and I've read many threads. I think I've just come to terms to play well within my partner's limitations, with this one little exception. (I should also note my experience with GIB is mostly related to the Robot Races. I'm not expecting it to play that well - after all, I don't. I'd just like it to remove takeout doubles to five card unbid suits.)

[barmar]

As I mentioned above, the bidding rules say that it should bid. I redid my simulation test under conditions closer to the way tourney bots work, and I was able to reproduce the pass.

What's going on is that it doesn't think it's likely to get to game -- even if your hands can make 4♠, you'll probably just be in 3♠. But on many of these hands, you're setting them 1 or 2. The thing that got it to misjudge was West's pass. It expects West to bid 4♣ with 3-card support (when dealing hands for a simulation it assumes everyone makes the book bid, but in this case West's simulations told it not to raise), so it's convinced that you have a doubleton ♣, and that makes it more likely that you only have 3 ♠. In some of the hands it dealt, you have ♣Qx behind East's KJxxxxx, so you're getting two ♣ tricks on defense, but only one on offense. And then there were some hands where you're doubling because you have a bid hand with 5 ♦, so you're going to bid 4♦ and go for a number when you could have beat 4♣.

We've been discussing giving the book bids more priority in these auctions.

[Bbradley62]

barmar, on 2012-February-15, 20:38, said:

We've been discussing giving the book bids more priority in these auctions.

Please give book bids more priority across the board. It appears to me that the majority of the time that players are unhappy with GIB's bids it is because GIB simulated instead of making the book bid.

[barmar]

Bbradley62, on 2012-February-16, 19:09, said:

Please give book bids more priority across the board. It appears to me that the majority of the time that players are unhappy with GIB's bids it is because GIB simulated instead of making the book bid.

While those may be the majority of the complaints, we don't hear about all the times when the simulations figured out the RIGHT thing to do when the book bid would have been wrong. So it's not clear that the simulations are causing more harm than they help.

[mgoetze]

Bbradley62, on 2012-February-16, 19:09, said:

Please give book bids more priority across the board. It appears to me that the majority of the time that players are unhappy with GIB's bids it is because GIB simulated instead of making the book bid.

Here's an example of me being unhappy because GIB made too much of a "book bid":

---

Surely a few simulations would have helped GIB figure out that offering to play a 5-0 trump split is not a good idea.

[Bbradley62]

I don't see that as an offer to play, but rather as a description of his hand so that you can make an informed decision as to the best contract.

[jamegumb]

Bbradley62, on 2012-February-22, 07:24, said:

I don't see that as an offer to play, but rather as a description of his hand so that you can make an informed decision as to the best contract.

I dunno. Give south:

J82
AJx
AJxx
AQx

The 5-0 split has chances, while the spade contract could be down on two heart ruffs.

[AAr]

I should point out I've had GIB pass out my takeout doubles several times on inapproprate hands, and usually I get poor results (usually opponents making doubled) when it does happen. Occasionally, I do get a +500 with no game on with GIB correctly passing out a takeout double (Of course, other tables are having the same happen so it's not a cold top when it does.) Usually initial takeout doubles do get taken out approately. It's reopening doubles that are for takeout that GIB tends to leave in error. It still won't stop me from making reopening doubles.

But, what the kind of hands GIB does passout takeout doubles makes me reluctant to for myself to pass out penalty doubles. I recall one hand (Don't recall the specific hand or details) where I went for, IIRC, -1100 for taking out a penalty double, ending up in a weak 4-3 fit with no play whatsoever (They ended up drawing our, trumps and cashing their side winners at the end.) when passing out the double would have left them in a 4-2 fit against a 5-2 break, likely giving us at least a +800. But, giving the hands GIB pa out takeout doubles, and the way GIB declares much better than it defends, and the way it usually bids, I thought the contract would likely make and our contract would be two down at worst.

Of course, one I thing I can't complain about is GIB's overbidding (or perhaps my misbidding) of slams. I have been pushed to nonmaking grands before (most often bidding 7 of a suit over 6NT, or 6NT or a slam in a high-ranking suit in which I wanted to be in a slam in a lower-ranking suit), but I've also gotten lucky with GIB pushing me to a grand I did not want to be in, then needing a finesse to win and good suit breaks, and getting the lucky layout, and won a couple of GIB duplicate tournaments this way.

[cloa513]

jamegumb, on 2012-February-22, 12:23, said:

I dunno. Give south:

J82
AJx
AJxx
AQx

The 5-0 split has chances, while the spade contract could be down on two heart ruffs.

Why spurn 3NT which is far easier?

[ahydra]

I think 3H on that sequence should show heart shortness and a 6th spade - as, like you say, playing against a 5-0 split is not a good idea. So it's not "making a book bid" that's the problem, but perhaps the book bid is just wrong here.

ahydra

[jamegumb]

Maybe the natural bidding helps GIB sniff out psyches?

[Bbradley62]

Another reopening double that got passed.

---

[Bbradley62]

and this takeout double

---