Rejecting bad clams of too few tricks in Robot games

[uusancal]

In robot games, there are no directors. The computer makes a decision on claims by the human player. The computer is very good at denying claims for too many tricks and accurately denies them when there is a reasonable chance that a line of play will result in fewer tricks than claimed. I believe the same computer should reject claims where too few tricks are claimed.

 

Case in point from a few days ago (hopefully the image the hand loads below). The most any normal person would lose is the AS. Moreover, playing out the Diamonds would result in no further tricks being lost. A claim of losing two more tricks should be rejected, just like it would by a director in a director-managed game.

 

 

 

Click here to see the hand and claim

[pescetom]

The most any normal person would lose is no more tricks B-)

But as you say, a poor player might give up a spade unnecessarily, or even lose two tricks in various stupid ways.

How is the computer supposed to judge?

I imagine all it does is check that the number you claim is feasible double-dummy, and even that is rather generous.

 

In a Director-managed game outside of BBO, you need to state your proposed line of play, including the order in which the cards will be played, so it is unlikely that you will be making more tricks than you claim and if so you will probably be allowed to reformulate the claim.

 

In a Director-managed game on BBO, it's up to the Director what happens if you point out that you claimed too few tricks - I doubt she would be very sympathetic.

[smerriman]

I imagine all it does is check that the number you claim is feasible double-dummy, and even that is rather generous.

FYI, it's not; it's a single dummy algorithm. Not quite sure how it works but I suspect it's a bit like the Bridge Master idea, where it can rearrange cards between the opponents in order to try to generate the worst possible scenarios and see if that results in the claim failing.

 

That's why it always rejects the claim if, say, the only way you can go down is a 5-0 split, even when the actual split is 3-2.

[pescetom]

FYI, it's not; it's a single dummy algorithm. Not quite sure how it works but I suspect it's a bit like the Bridge Master idea, where it can rearrange cards between the opponents in order to try to generate the worst possible scenarios and see if that results in the claim failing.

 

That's why it always rejects the claim if, say, the only way you can go down is a 5-0 split, even when the actual split is 3-2.

 

Interesting, thanks.

So in this case it would not have accepted a claim of 12 tricks because W might have started with ♦J9543 ♣AK82, right?

[smerriman]

Yep. Barmar once said they system in place doesn't theoretically guarantee getting the claim result right (and that they were provided a different claim checker but it hadn't been integrated into BBO), but I've never seen it accept a claim which it shouldn't*. Compared to the play algorithm where extreme cases often don't come up in simulations, the claim checker algorithm always seems to find the exception.

 

*Lamford claims to have found an exception to the converse, but I still haven't been able to prove it to myself satisfactorily yet. But that's not really a downside.

[thepossum]

Deleted. Wrong forum sorry