[the_dude]

..I just tried the Robot tournaments this week .. so maybe this is just expected behavior and I'm just new .. but I was under the impression that GIB could at least count it's top winners. Is that not true?

---

I made a bidding miskey and the robot bid NT first .. we ended in 6NT by North with no opposition bidding and a small heart was led. Diamonds are 3-1 so obviously you have 12 top tricks. In addition East holds both 4 spades and the A of clubs so even a novice (or robot) couldn't help but squeeze him while running his red suit winners.

Somehow my robot went down 5.

Worse still, a glance at the score sheet shows that every other robot that played the hand in NT made 7. Those that played in Diamonds made 7 too. How is this possible? There was NO opposition bidding that might even confuse it. How come my robot can't count tricks and every other one can?????

ps: sorry for the rant .. that 0% (instead of 90%) made a big difference in the final results .. I guess I'm kinda annoyed right now ...

[wank]

chill, dude

[ArtK78]

The more you play in robot tournaments the more you will see things like this.

Recently I made a takeout double holding 1-3-4-5 distribution after a 1♠ bid on my right. My LHR bid 2♠ and my Robot partner bid 3♦. After a 3♠ bid on my left the auction was passed around to me. I competed with 4♦. My Robot partner now bid 4♥! I corrected to 5♦ and the Robot bid 5♥. I gave up. The Robot had 3 hearts and 5 diamonds.

I am very hesitant about making penalty doubles when I know that I have the opps beat because the robot frequently pulls penalty doubles.

On the other hand, I frequently pull the robot's penalty doubles because when I don't my opps make overtricks. This is happening less often - the robots are getting better - but I still have a lot of nonvulnerable -690 scores and the like.

Another thing to look out for is the soundness of the robot's preempts. I have seem some of the soundest preempts in the history of bridge perpetrated by the robots. Be wary when doubling opposing robot's preempts. Also be aggressive when your robot partner opens a vul weak two. I frequently raise my partner's weak 2 bids to 4 on hands that I would never consider acting on in real life.

[Rain]

The same robots are used in each tourney. But bidding difference is one factor that can result in differences in simulation, leading sometimes to different results.
Also, in MP, GIB currently always tries to get a top.

http://www.bridgebas...ername=the_dude

I think he tried to finesse to get 6NT+1.
All the other making 6NT contracts had similar bidding and all elicited a spade lead, sparing the robot from an embarrassing finesse.

[xxhong]

This strategy is really bad. For all MP players, the goal of MP is not to shoot for tops, but for highest expected value in MP.
For example, if you bid 6S and don't think it is a very popular contract because most other players will stop at game, making it will be good enough and shooting for over tricks can be just absurd. Suppose only two tables bid 6S, 8 tables bid 4S or 5S, so you get 8 match points if both of you make 6. Suppose it's a 100% slam to make 6 and 50% to make 7 and 50% to go down. So the EV for making 6 is 8 match points, the EV for making 7 is 9*0.5=4.5.

Of course, it is often difficult to assign the accurate MP for each hand when gib plays, but this kind of thinking process is the only sound one that one has to apply. Shooting for tops would cost players tons of MP.

Rain, on 2011-July-29, 09:46, said:

The same robots are used in each tourney. But bidding difference is one factor that can result in differences in simulation, leading sometimes to different results.
Also, in MP, GIB currently always tries to get a top.

http://www.bridgebas...ername=the_dude

I think he tried to finesse to get 6NT+1.
All the other making 6NT contracts had similar bidding and all elicited a spade lead, sparing the robot from an embarrassing finesse.

[Bbradley62]

Rain, on 2011-July-29, 09:46, said:

Also, in MP, GIB currently always tries to get a top.

I think this is a gross overstatement; GIB tries to maximize his score. If he thinks there's a good chance of getting the overtrick, he'll go for it; he doesn't do it just because there is any chance that it will work.

This hand demonstrates a lead problem that I've pointed out before: GIB leads low from 3 small against NT, although his convention card says that he leads top of that holding. So, North misanalyzes things based on the misinformation he's given about East's leads.

[Bbradley62]

the_dude, on 2011-July-29, 07:40, said:

I made a bidding miskey and the robot bid NT first

Your misclick actually right-sided the slam contract. Yes, GIB butchered this hand. Note: 3N is a horrible bid over 1♦!

[xxhong]

Bbradley62, on 2011-July-29, 15:33, said:

Your misclick actually right-sided the slam contract. Yes, GIB butchered this hand. Note: 3N is a horrible bid over 1♦!

If you allow gib to declare a slam, no matter what, it is wrong sided.

[pigpenz]

Rain, on 2011-July-29, 09:46, said:

The same robots are used in each tourney. But bidding difference is one factor that can result in differences in simulation, leading sometimes to different results.
Also, in MP, GIB currently always tries to get a top.

http://www.bridgebas...ername=the_dude

I think he tried to finesse to get 6NT+1.
All the other making 6NT contracts had similar bidding and all elicited a spade lead, sparing the robot from an embarrassing finesse.

what i dont understand is I have the commercial version of GIB and it is very competent. The robots seem to be the bastard sons of GIB. I am curious by default what does the old interface of GIB default too as far as thinking? And does the the amount of CPU power have any influence on the simulations done by GIB or is it all old progamming as far as power goes, to me it would seem cpu power would come into play.

Matchpoints: I tend to agree part of a good MP player is determining how good your contract is in relation to where the field will be not to play for the maximum number of tricks on each hand, if all the GIBS play the hand then I guess their mistakes for overtricks are all an average board.

[the_dude]

Rain, on 2011-July-29, 09:46, said:

The same robots are used in each tourney. But bidding difference is one factor that can result in differences in simulation, leading sometimes to different results.
Also, in MP, GIB currently always tries to get a top.

http://www.bridgebas...ername=the_dude

I think he tried to finesse to get 6NT+1.
All the other making 6NT contracts had similar bidding and all elicited a spade lead, sparing the robot from an embarrassing finesse.

Thanks for the explanation .. I get it I guess. It just seems to me that risking *any* contract for an overtrick should not be part of GIB's arsenal. If GIB is not capable of determining the subtleties of *when* an overtrick is worth the risk (and clearly it is not) then it seems to me the default behavior should be to *always* make the contract if possible. I mean, if you are trying to make it simple and foolproof .. the most basic MP concept is "plus is good"

Bbradley62, on 2011-July-29, 15:33, said:

Your misclick actually right-sided the slam contract. Yes, GIB butchered this hand. Note: 3N is a horrible bid over 1♦!

Actually .. i was trying to bid a very normal 1♥!

I'm new to the robots so I was mousing over many of the bids to see what they would mean to GIB - a very very handy feature I might add. (and I confess I *was* checking out what 3NT would mean) but then I went back to hearts .. it just wasn't supposed to be 3!

In the future tho .. based on this I'm going to have to agree with xxhong and make sure I always bid NT first regardless

[barmar]

the_dude, on 2011-July-29, 20:29, said:

Thanks for the explanation .. I get it I guess. It just seems to me that risking *any* contract for an overtrick should not be part of GIB's arsenal

This is a standard part of matchpoint strategy.

If you're playing in a good field, and you're in the normal contract, most of the other players are going to do the same thing. So if you go down, so will they, and you'll all get an average.

And if you're in the WRONG contract, you often need to do this. E.g. if you expect the field to be in 4♠, while you're in 3NT, and both games make, you need to make an overtrick in 3NT to get a good board. Just making your contract will be a bad score, so it won't be much worse if you go down.

But this is something GIB can't analyze. So it simply uses the strategy of trying to maximize the expected number of tricks in its contract.