[gwnn]

Mostly for comic relief. Check out trick 2:

---

[sakuragi]

I have seen quite several of these as well in bbo challenge.
It seems that these happens since the advanced robot bbo challenge ($0.25 one) is released.
Does the free BBO challenge robot become worse and the original (better) one goes to advanced robot bbo challenge?

[LBengtsson]

??? roboten behöver en ingenjör med en skruvmejsel!

[johnu]

On a good day, the spade finesse would have worked. This was a play that had a 50% chance of working so I can't criticize the play very much compared to some of the 0% plays that are out there.

[iandayre]

Why cannot GIB be prevented from making plays like this? Or better, why wasn't it prevented from doing so years ago?

[gwnn]

johnu, on 2017-October-06, 12:09, said:

On a good day, the spade finesse would have worked. This was a play that had a 50% chance of working so I can't criticize the play very much compared to some of the 0% plays that are out there.

Oh yea, I mean that's why I wrote mostly comic relief. It's even better than 50% after a takeout double, perhaps.

[Stephen Tu]

This particular one is actually not that hard to understand if you know how GIB thinks, for the first few tricks before the single dummy solver kicks in (for the advanced bots).

The problem is at trick 2, from double dummy perspective high spade from hand vs. low spade are equally good from its perspective, if it will always guess right subsequently then it doesn't really matter. So it randomly pick low spade (an error, because it's not looking ahead and realizing it can guess wrong, not analyzing from single dummy perspective). High spade is better as the discovery play.

But now, having led low spade from hand, if North has Txx spades, in order to get the max tricks it has to finesse the 9, otherwise ruffing hearts will eventually promote the T into a trick, or you try crossing in clubs to ruff diamonds and North ruffs a club at some point.

If it thinks South's double biases the deals enough towards void spade, vs. having stiff T/doubleton T, it's actually right to finesse the 9 having already made mistake of not leading high from hand. Without double it will surely not take finesse.

The solution is basically to kick in GibSon sooner I guess, but that requires a lot of resources with lots of tricks to go which is why it doesn't kick in for a few tricks.


Now for some of the other nutty plays posted recently, I think GIB just plays super stupidly if it is being run with too little memory/thinking time. Others I think screwy auctions really make it confused about what hands the opponents can have. On defense, it's just super handicapped by not understanding signalling at all, so it will be hopeless at this until they can implement, which is super hard vs. picking off bidding bugs.

[Stefan_O]

Stephen Tu, on 2017-October-06, 13:23, said:

If it thinks South's double biases the deals enough towards void spade, vs. having stiff T/doubleton T, it's actually right to finesse the 9 having already made mistake of not leading high from hand. Without double it will surely not take finesse.

Interesting thought, but can this really be the right explanation?

A priori,
North=Txx is only ~11%,
while South=Tx or T is ~39%, and the takeout (0-2 spades) does not exclude any of these...

[Stephen Tu]

It doesn't exclude, but South needs stronger hands to double without void than with, especially with North having HA. Also it expecting 3+ all the outside suit which is less frequent when south has doubleton. So much higher fraction of void spade qualify for takeout double than the balanced hand with doubleton. Whether bot calc is correct about this I don't know.

[gwnn]

Cool explanation!