[lycier]

---

[johnu]

IMO, it is a cry for help from GIB "Programmers - please fix me"

[steve2005]

Is it just me or is this happening more often? Cause I have seen some.
A hand was played by another player, so I can post the hand.
They played 3N from wrong side so K2 is on board
Gib actually led a suit partner had bid!
8 lead 2 from dummy 7 from RHO 3 from declarer.
Gib then switched!
RHO had AQJT976 and could have cashed first 7 tricks if it overtakes. RHO should overtake leader as it could be singleton, but it wasn't!

[Stephen Tu]

Josh said something about defaulting to ducking if run out of thinking time on lycier's other thread. Perhaps number of users have increased to the point where the servers are sometimes overloaded.

[johnu]

Stephen Tu, on 2017-September-07, 08:48, said:

Josh said something about defaulting to ducking if run out of thinking time on lycier's other thread. Perhaps number of users have increased to the point where the servers are sometimes overloaded.

I don't really care what happens for "free" GIB basic robot games because you are getting what you paid for.

If you have paid for GIB robots either as a rental or by playing in a paid tournament, running out of time should never happen (or there should be a vastly increased time limit to avoid this happening so you can get a valid result). If the paid GIBs get hopelessly stuck in some kind of infinite loop of logic, a dummy double completion of the hand should be implemented after a suitably long period of time.

[steve2005]

johnu, on 2017-September-07, 12:25, said:

I don't really care what happens for "free" GIB basic robot games because you are getting what you paid for.

If you have paid for GIB robots either as a rental or by playing in a paid tournament, running out of time should never happen (or there should be a vastly increased time limit to avoid this happening so you can get a valid result). If the paid GIBs get hopelessly stuck in some kind of infinite loop of logic, a dummy double completion of the hand should be implemented after a suitably long period of time.

When programming it is very dangerous to not have a time limit, memory limit, cpu limit etc. or 1 user can take entire resources of a system. Usually if one user is taking over 50-60% of some critical resource the whole system will become unworkable, At our university when they reset the network sever there was one limit was left off in error. One users program quickly took up all available resources.

I would be surprised if the advanced robots don't have limits on them and they won't be huge. If Gib hasn't figured things out in 20-30 sec (I don't know exact time) it wont figure it out. So this problem likely applies to the advanced robots also and should be addressed.


But in any case, if your trying out the robots and it makes a mistake a person who learned bridge the day before wouldn't make; Do you think they are going to say sure lets spend money and play with the robots? I saw a hand posted where opening lead was K and AQJ3 are in dummy and the 3 is played. If Gib doesn't have enough time to run the number of simulations it needs, then run a smaller number or even just 1 random hand. At least then it wouldn't make this silly duck. Defaulting to ducking is as bad as another default passing when not knowing what to bid.

[steve2005]

johnu, on 2017-September-07, 12:25, said:

If the paid GIBs get hopelessly stuck in some kind of infinite loop of logic, a dummy double completion of the hand should be implemented after a suitably long period of time.

If Gib does get stuck in an infinite loop, by definition it won't be able to calculate a double dummy solution without knowing the opponents hands, which would violate the rules of bridge.