[lycier]

---

[/size]

Now you see Gib east easily discard ♥K only in the second round. This isn't a acceptable for arrangement.
I guess there might be something wrong with Gib playing server.

[Stefan_O]

lycier, on 2016-May-26, 21:54, said:

---

[/size]

Now you see Gib east easily discard ♥K only in the second round. This isn't a acceptable for arrangement.
I guess there might be something wrong with Gib playing server.

From my experience, declaring against Gib, when the simulations show two or more cards make the same result double-dummy, it will usually (always?) discard the highest.

Since East here can see dummy's Heart suit, and also infer that West has a very weak hand, it might well be that the double-dummy simulations here all show it doesnt matter if it discards the King or a small -- therefore choses the King. Since North's 3♥ bid shows 3 hearts, declarer will always score 4 heart tricks anyway.

This strategy of throwing the highest of equivalent cards often works OK, and sometimes can actually fool you, particularly as long as you as declarer has not totally learned about this robot-habit.

But not so well, in all situations, where declarer would otherwise have some kind of guess, of course.

It looks silly on hands like this, but the current robot-algorithm has no concept of when throwing a high or low makes life harder or easier for declarer.

But I think, yes, BBO should definitely review, if this "discard-high" strategy is really an overall winning one? (Maybe they already did, I dont know....)

It certainly has happened, that I was in a tricky high-level contract and pondering what suit to attack, then draw some extra round(s) of trumps, and Gib would needlessly throw away a sensitive card, that saved me from guessing.

I'm sure "discard-low" also has its issues, though you might not actually notice them so readily, because they dont look so conspicuous as this example.

And if you randomize between high-or-low discards, well, then you will still see behaviour like this deal, some of the times.

[johnu]

Stefan_O, on 2016-May-28, 15:43, said:

From my experience, declaring against Gib, when the simulations show two or more cards make the same result double-dummy, it will usually (always?) discard the highest.

Since East here can see dummy's Heart suit, and also infer that West has a very weak hand, it might well be that the double-dummy simulations here all show it doesnt matter if it discards the King or a small -- therefore choses the King. Since North's 3♥ bid shows 3 hearts, declarer will always score 4 heart tricks anyway.

This strategy of throwing the highest of equivalent cards often works OK, and sometimes can actually fool you, particularly as long as you as declarer has not totally learned about this robot-habit.

IIRC, GIB randomly throws high or low cards when it thinks it makes no difference.

You have misread the auction. South is declarer, so the good 4 card heart suit cannot be seen.

The question as to whether GIB should randomly throw discards has been discussed many times and I am convinced that if GIB is randomly throwing cards (as opposed to discarding for a specific purpose, e.g. some kind of unblocking play), it should play the lowest card, or the lowest of equals to the best of GIB's analysis.

EDIT: GIB can/should play random from equals, not always lowest. So with a 10 & 9, GIB can play either (assuming BBO never implements any kind of signalling for GIB), but with K & 2, it would play the 2

In shorthand, GIB should play lowest from unequals, random from equals.

[1eyedjack]

I am pretty confident that GIB is *not* randomly throwing cards, although this is often quoted as the reason for its not playing lowest. I base this on my observation that GIB plays its lowest card with a frequency which (I believe) is much less than would be consistent with random policy. I have not kept empirical records to support this observation, so feel free to opine otherwise.

If my suspicion is correct, and without doing the work I accept that there has to be an element of doubt about that, then it is not hard to understand the mentality behind the programmers' policy. Intermediate players and below, as well as a sizeable proportion of "advanced" players, are used to (human) defenders playing lowest. If declarer assumes that GIB is following the habitual trend of their usual human opponents then GIB will gain from the exchange by departing from expectation (possibly enhanced by declarer occasionally losing track of the spot cards).

Unfortunately, such a policy would be a losing strategy against good opponents who rumble the policy and do not lose track of the spots. Indeed as a policy it would be just as bad as always playing the lowest card, even if it never threw an active card inadvertently when pitching high.

So it may well be that always playing lowest would be an improvement on current policy of never playing lowest (if that is indeed the current policy, as I observe). Because that way it would never accidentally pitch an active card.

It would however only represent a marginal improvement, the margin being those hands where it is tempted to pitch an active card, which is not that frequent. The frequency of pitching active high cards appears to be higher than reality, simply because they invite being posted in the forums, where routine good plays do not get the same publicity.

Furthermore, it would lose its current edge over the beginner/intermediate player who currently assumes that GIB is playing lowest.

---

A routine bread and butter hand. You cash the AK, to which all follow, East with the Ten on the second round.

If GIB were programmed always to play the lowest card when not trying to win, and there were no other indicators from the bidding or play, you might be tempted to play for the drop, and you would be right to do so fractionally better than half the time.

If East instead were to follow with the Jack, you would know to finesse, with 100% success rate. You would probably finesse anyway in real life, but without that level of confidence. Indeed the reduced confidence in real life may indicate a completely different line of play.

This effect would totally distort the playing odds so as to change the nature of the game utterly.

The above hand is no contrived construction, but is the sort of decision that you are faced with time and again, much more frequently than the occasional pitch of a significant card that is GIB's current foible.

Under GIB's current (my suggested) policy of never playing lowest, you would finesse if the Ten appeared, and play for the drop if the Jack appeared, for which your long term success rate would be the same as if it always played lowest. But a defensive strategy of truly playing at random would improve its long term success in defence compared with either of those alternatives.

Let me finish by asking this: If GIB were to change its strategy to that of always playing lowest spot card, how many posts to the forum would you expect to see of hands like the heart suit in this thread when declarer successfully plays for the drop when the real life odds favour finessing? I suggest that the posts would be less frequent than the occurrence at the table.

[Stefan_O]

johnu, on 2016-May-28, 20:44, said:

You have misread the auction. South is declarer, so the good 4 card heart suit cannot be seen.

You're right. My bad.

So, back to the OP mystery, then....
How can the discard of ♥King make sense?
Is it a BUG after all?

[lycier]

How to explain my another issue " Gib throws useful high card " ?