BBO Discussion Forums: Gib shows 6S5H when has 6H5S and P pref H - BBO Discussion Forums

Jump to content

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

Gib shows 6S5H when has 6H5S and P pref H

#21 User is offline   gwnn 

  • Csaba the Hutt
  • PipPipPipPipPipPipPipPipPipPip
  • Group: Advanced Members
  • Posts: 13,027
  • Joined: 2006-June-16
  • Gender:Male
  • Location:Göttingen, Germany
  • Interests:bye

Posted 2016-April-23, 02:45

View Post1eyedjack, on 2016-April-23, 02:35, said:

Quote

And each had the chance to communicate said preference to the other.

FYP

FYFYP? :P
... and I can prove it with my usual, flawless logic.
      George Carlin
1

#22 User is offline   lycier 

  • PipPipPipPipPipPipPipPipPip
  • Group: Advanced Members
  • Posts: 7,612
  • Joined: 2009-September-28
  • Gender:Male
  • Location:China

Posted 2016-April-23, 03:10

View Postgwnn, on 2016-April-23, 02:32, said:

Your "method" is, when confronted with a bug, to point out that there are other sequences which do not have bugs. It's like if I called a plumber to fix my drain and he replied "there is no problem with drains in general. 73% of drains work." That is not a method for plumbing. That is just trolling.


Actually about a bug, it is a very complex issue, I believe that many issue is how to choose bids instead of Bug since I have some the data to support my point, how about your data?
Nothing.
When you want to make a point, you may need to have the related data to support it. In fact, you have no exact methods , you are making joke. it is rediculous for you to deny my method to show off your method. How many bugs did you found in the past at this forum?
Rarely !



0

#23 User is offline   gwnn 

  • Csaba the Hutt
  • PipPipPipPipPipPipPipPipPipPip
  • Group: Advanced Members
  • Posts: 13,027
  • Joined: 2006-June-16
  • Gender:Male
  • Location:Göttingen, Germany
  • Interests:bye

Posted 2016-April-23, 03:52

The data is the hand in the opening post. Everything else you raise is irrelevant. It is irrelevant how many bugs I found and how many I did not find. It is also irrelevant what GIB does in other sequences if it has a bug in this one. My method is: look at the hand in question and see if GIB has a bug. Yours is: look at other sequences and prove that GIB is not 100% buggy (which nobody claimed). It is not "rediculous" to compare a "method" to another "method". You are on my ignore list from now. Good bye. I apologize for people who had to listen to this. In case anyone is wondering how to do it (to block me, or lycier, or anyone else), click on your name in the top right corner of the website, and click on the last item "Manage ignored users."
... and I can prove it with my usual, flawless logic.
      George Carlin
0

#24 User is offline   Stephen Tu 

  • PipPipPipPipPipPipPip
  • Group: Advanced Members
  • Posts: 4,070
  • Joined: 2003-May-14

Posted 2016-April-23, 04:08

View Postlycier, on 2016-April-23, 03:10, said:

Actually about a bug, it is a very complex issue, I believe that many issue is how to choose bids instead of Bug since I have some the data to support my point, how about your data?


We're saying that one data point is perfectly sufficient to report a bug. You bringing up a bunch of other different auctions is totally irrelevant. Computer programming is not like say a drug efficacy trial where you need a certain number of statistically significant results to conclude a drug was effective. One instance is enough to report! A programmer can take a single report of one hand where GIB did something weird/bad, and generally be able to reproduce it, and find a bug in the rule database that it is causing it to do that. Computers are supposed to get it right every time, not just most of the time. We don't say "well it didn't screw up on every deal, therefore there is no bug". Just because it passed 2H some of the time (which is also wrong IMO, too conservative, it missed game), doesn't mean that when it bids 2S that isn't a mistake, isn't a bug. We have different versions of GIB floating around here, and some actions are RNG random number generator dependent, so there is some variance in response. But generally we want GIB to bid the same thing every time given the same auction, so even the fact that sometimes it bid 2s and sometimes it passed 2h is something worth reporting, and investigating. Now maybe they investigate and say one set of GIBs is say the download windows app, old version, and they ignore the report, but so what? You clearly don't have any expertise in computers to put you in position to criticize what other people think is a bug as not a bug. I agree with gwnn, you don't seem to understand the definition.

As for the hands where the opening hand took a second call, those are even more totally irrelevant as that gives the computer a different decision than over 2H - pass.

We are generally forgiving of GIB's errors because bridge is a very complex game with a practically uncountable number of possible bidding sequences. So it's understandable that it gets some bids wrong, it's a near impossible programming job to cover all the auctions. But it doesn't mean that the bid wasn't wrong, that there isn't a bug there, that GIB cannot be improved on a particular sequence. You seem to have invented your own unique definition of what a bug is, with little to no background in computer science.
1

#25 User is online   steve2005 

  • PipPipPipPipPipPipPip
  • Group: Advanced Members
  • Posts: 3,148
  • Joined: 2010-April-22
  • Gender:Male
  • Location:Hamilton, Canada
  • Interests:Bridge duh!

Posted 2016-April-23, 08:47

View PostStephen Tu, on 2016-April-22, 20:31, said:

It's apparent that this hand was likely played by different versions of GIB. Some of them choose a poor 2 call and continue to rebid spades, strangely, perhaps because simulations say that it shouldn't give up after 2 (agree), but likely the definition of heart rebids is poor and preventing it from choosing that, so it finds 2 as a poor 2nd choice
[/size][/color]

How could it be using different version of Gib is in a regular tournament not an instant.
It's clear when there is interference it gets it right.
With no interference it is making the 2nd best call as you said, perhaps to conserve bidding space, but that is a bug. When you have a hand that clearly describes your hand 3 you make it you don't conserve bidding space.
I have to wonder if the description of the bid is different then the definition Gib is using, it must be hard to keep this consistent when changes are made
Sarcasm is a state of mind
0

#26 User is offline   gwnn 

  • Csaba the Hutt
  • PipPipPipPipPipPipPipPipPipPip
  • Group: Advanced Members
  • Posts: 13,027
  • Joined: 2006-June-16
  • Gender:Male
  • Location:Göttingen, Germany
  • Interests:bye

Posted 2016-April-23, 08:57

Stephen Tu: you are still misreading the stats. GIB always bid 2 over 2. Only after:

1NT-2-p-2 (2=majors)
p-?

Did it pass 2. So all lycier's point really only was that GIB does not produce bugs on every single auction.
... and I can prove it with my usual, flawless logic.
      George Carlin
0

#27 User is online   steve2005 

  • PipPipPipPipPipPipPip
  • Group: Advanced Members
  • Posts: 3,148
  • Joined: 2010-April-22
  • Gender:Male
  • Location:Hamilton, Canada
  • Interests:Bridge duh!

Posted 2016-April-23, 10:11

View Postgwnn, on 2016-April-23, 08:57, said:

Stephen Tu: you are still misreading the stats. GIB always bid 2 over 2. Only after:

1NT-2-p-2 (2=majors)
p-?

Did it pass 2. So all lycier's point really only was that GIB does not produce bugs on every single auction.

Most people didn't open 1N on this hand I did cause i had 16 pts and didn't think i could describe my strength otherwise so 1N. So probably scared Gib out of game.
One person opened 1 and same sort of Michaels auction occurred
Where the auction differed was when Human player south rebids 3 or 3. That 3 is now the cheapest bid may be a coincidence but in any case now Gib always bids 3
Sarcasm is a state of mind
0

#28 User is offline   Stephen Tu 

  • PipPipPipPipPipPipPip
  • Group: Advanced Members
  • Posts: 4,070
  • Joined: 2003-May-14

Posted 2016-April-23, 11:27

View Poststeve2005, on 2016-April-23, 08:47, said:

How could it be using different version of Gib is in a regular tournament not an instant.


View Postgwnn, on 2016-April-23, 08:57, said:

Stephen Tu: you are still misreading the stats. GIB always bid 2 over 2. Only after:
1NT-2-p-2 (2=majors)
p-?
Did it pass 2. So all lycier's point really only was that GIB does not produce bugs on every single auction.


My bad!!!! I didn't carefully read lycier's wall of examples and notice the auction was different since I've mostly been in the habit of ignoring his walls of examples since they are also usually irrelevant in the other posts of this forum which are not always tournaments with identical GIBs.

So here it's instant tournament, GIB's are all the same, and lycier was only pointing out GIB did something different when the humans made the auction different by opening 1nt super off shape or bidding again the 2nd time or something. That just makes lycier's arguments even more totally ridiculous. "GIB doesn't display bug on another sequence, therefore it's not a bug on this sequence??" Lycier, basically you are arguing that if GIB doesn't screw up every single sequence, nothing it does can ever be considered a bug. Don't you see how insane you are being?
1

#29 User is offline   gwnn 

  • Csaba the Hutt
  • PipPipPipPipPipPipPipPipPipPip
  • Group: Advanced Members
  • Posts: 13,027
  • Joined: 2006-June-16
  • Gender:Male
  • Location:Göttingen, Germany
  • Interests:bye

Posted 2016-April-23, 13:02

No he doesn't see it. Well, maybe he's more open to understand if you are the one who says it as he thinks you are the only person on this subforum worth listening to.
... and I can prove it with my usual, flawless logic.
      George Carlin
1

#30 User is offline   Stefan_O 

  • PipPipPipPip
  • Group: Full Members
  • Posts: 469
  • Joined: 2016-April-01

Posted 2016-April-24, 18:37

View Postlycier, on 2016-April-22, 07:07, said:

- The worst things Gibs did is that Gibs shouldn't rebid 2♠ after responding 2♥, there are 6 hands in total.


Yes, THAT one certainly IS an error/bug/defect in the bidding-logic, EVEN if ALL the other sequences are OK.

Such obvious errors should be CORRECTED.

Somehow, I start to believe you don't know what the word "bug" means....
It does NOT mean that EVERYTHING is wrong.
It CAN mean there is an error ONLY in a very specific situation or detail.

"Bug" simply means "SOMETHING that needs to be CORRECTED".
The robot should NOT bid 2 over 2 here -- that's all it means.

The simple reason is that 2 here DEVIATES from what the EVERY 2/1 player expect for that bid.
0

#31 User is offline   Stefan_O 

  • PipPipPipPip
  • Group: Full Members
  • Posts: 469
  • Joined: 2016-April-01

Posted 2016-April-24, 18:55

View Postlycier, on 2016-April-23, 03:10, said:

Actually about a bug, it is a very complex issue


NO, a bug in the bidding does NOT HAVE to be a "very complex issue".

If the robot bids something that NO 2/1 players would agree with,
and which most likely ends up in a WORSE result, it certainly IS a BUG,
simply because 2/1 is the system we have AGREED upon.

In BIDDING, it is as SIMPLE as that.


Card-play issues, on the other hand, MIGHT be very complex, because the robot uses VERY different methods from humans,
and you may have to use a lot of mathematics to tell what is right or wrong and ALSO understand the robots algorithms and especially their LIMITATIONS.

But OBVIOUS bidding errors are always VERY EASY to spot.
0

  • 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