BBO Discussion Forums: New GiB record? - BBO Discussion Forums

Jump to content

  • 2 Pages +
  • 1
  • 2
  • You cannot start a new topic
  • You cannot reply to this topic

New GiB record?

#21 User is offline   diana_eva 

  • PipPipPipPipPipPipPip
  • Group: Admin
  • Posts: 4,998
  • Joined: 2009-July-26
  • Gender:Female
  • Location:bucharest / romania

Posted 2016-October-21, 10:37

 MrAce, on 2016-October-21, 10:13, said:

Diana, here is an example of tournament GIB


http://tinyurl.com/zxghuoz

Check the 9th trick


That's not nearly close to bot losing a trick from AKJT9x vs Qxx. That;s the kind of thing I was referring to when I said rented/tourney bots won;t do that.

#22 User is online   smerriman 

  • PipPipPipPipPipPipPip
  • Group: Advanced Members
  • Posts: 4,029
  • Joined: 2014-March-15
  • Gender:Male

Posted 2016-October-21, 14:20

 diana_eva, on 2016-October-20, 05:38, said:

Happens only to challenge and daylong robots, you won;t get these when renting basic robots or in robot tournaments.


Does that include the free weekly instant tournament? Just saw this at another table:



See trick 11. Similar to MrAce's, except there are only 3 cards left, two suits, and it plays the card guaranteed to be worse than any other.
0

#23 User is offline   diana_eva 

  • PipPipPipPipPipPipPip
  • Group: Admin
  • Posts: 4,998
  • Joined: 2009-July-26
  • Gender:Female
  • Location:bucharest / romania

Posted 2016-October-21, 15:40

Sorry I even attempted an explanation. I shd leave this stuff to programmers.

Anyway, no, it doesn't apply to instants of any kind. Only to challenges and daylongs. I was trying to reassure people that absurd plays like the OP, or losing a trick from KQ9x vs AJxx like here, or playing low-low-low with KQxx on AJTxx like here won't occur with rented robot or in robot tournaments. I didn't expected as counter example defense goof ups that I could be making at the table :P Not that those aren't bugs, but they aren't 100% impossible like the OP IMO.

We're testing advanced robots in Challenges which should hopefully fix these bugs, and in general the robots shd be smarter. If you register for the BBF Challenge event #4 barmar will enable advanced robots for participants so we can get some feedback on whether the advanced robots in challenges are better than the bots used now.

This post has been edited by diana_eva: 2016-October-21, 15:52


#24 User is online   smerriman 

  • PipPipPipPipPipPipPip
  • Group: Advanced Members
  • Posts: 4,029
  • Joined: 2014-March-15
  • Gender:Male

Posted 2016-October-21, 16:58

 diana_eva, on 2016-October-21, 15:40, said:

We're testing advanced robots in Challenges which should hopefully fix these bugs, and in general the robots shd be smarter.


Right - that's definitely a good thing, but is unrelated to the actual issue here - why is basic Gib making these defensive mistakes, when according to the definition of how it defends, it should be 100% impossible (more impossible than the declarer play, where the algorithm involved is a lot more complex). And will BBO ever look into it?

In terms of defense, the only difference between advanced Gib and basic Gib is an increased number of simulations - but even with a small number of simulations, the above mistakes should never happen, implying there's a larger fault here.

(Even if the programmers looked into a single example, and said the reason Gib lead the worst possible card is because it only ran 10 simulations, by a 1 in a million fluke, all of the 10 happened to not include the usual distributions, so it didn't think the card it played mattered. That would satisfy me.)
0

#25 User is offline   nige1 

  • 5-level belongs to me
  • PipPipPipPipPipPipPipPipPip
  • Group: Advanced Members
  • Posts: 9,128
  • Joined: 2004-August-30
  • Gender:Male
  • Location:Glasgow Scotland
  • Interests:Poems Computers

Posted 2016-October-21, 17:01

 diana_eva, on 2016-October-21, 15:40, said:

If you register for the BBF Challenge event #4 barmar will enable advanced robots for participants so we can get some feedback on whether the advanced robots in challenges are better than the bots used now.
Good! that should remove MrAce's objection :)
0

#26 User is offline   diana_eva 

  • PipPipPipPipPipPipPip
  • Group: Admin
  • Posts: 4,998
  • Joined: 2009-July-26
  • Gender:Female
  • Location:bucharest / romania

Posted 2016-October-21, 17:19

 smerriman, on 2016-October-21, 16:58, said:

Right - that's definitely a good thing, but is unrelated to the actual issue here - why is basic Gib making these defensive mistakes, when according to the definition of how it defends, it should be 100% impossible (more impossible than the declarer play, where the algorithm involved is a lot more complex). And will BBO ever look into it?

In terms of defense, the only difference between advanced Gib and basic Gib is an increased number of simulations - but even with a small number of simulations, the above mistakes should never happen, implying there's a larger fault here.

(Even if the programmers looked into a single example, and said the reason Gib lead the worst possible card is because it only ran 10 simulations, by a 1 in a million fluke, all of the 10 happened to not include the usual distributions, so it didn't think the card it played mattered. That would satisfy me.)


My reply refers to the OP which is a report on how GIB fails to cash out an all winner suit in a Challenge.

The rest of the examples are from different type of games, some use different type of robot and will probably filed as separate bug reports.

I have no idea what is the definition of how bot defends, sorry, all I wanted to point out is that a rented robot (even the basic one) will not mess up the suit play in examples like those I gave above.

  • 2 Pages +
  • 1
  • 2
  • You cannot start a new topic
  • You cannot reply to this topic

1 User(s) are reading this topic
0 members, 1 guests, 0 anonymous users