BBO Discussion Forums: Gib play conventions full disclosure - BBO Discussion Forums

Jump to content

Page 1 of 1
  • You cannot start a new topic
  • You cannot reply to this topic

Gib play conventions full disclosure

#1 User is offline   kgr 

  • PipPipPipPipPipPipPip
  • Group: Advanced Members
  • Posts: 3,423
  • Joined: 2003-April-11

Posted 2021-October-29, 08:15

Hello,
I wonder what 'agreements' Gib has on play and how it plays:
- How does it start? I thought 4th best, but now it started against 3NT from QJ954 the 4. (...how does it decide to start the 4 here?)
- Will it encourage when partner leads; or does it give count (normal count).
- Will it discard from length? Will it give count when discarding?
And is advanced GIB doing this differently?
0

#2 User is offline   smerriman 

  • PipPipPipPipPipPipPip
  • Group: Advanced Members
  • Posts: 3,720
  • Joined: 2014-March-15
  • Gender:Male

Posted 2021-October-29, 15:18

The standard lead from QJ954 would actually be Q rather than 5. GIB normally follow standard leads but sometimes leads randomly and nobody knows why.

It doesn't provide any useful signals ever. According to the system notes it's meant to signal in some situations, but does this so inconsistently that it's impossible to get any meaningful information from this.

See here for some analysis of both situations.

Advanced GIB only differs in the fact it simulates more hands and occasionally simulates during the bidding as well to deviate from the basic GIB bid.
0

#3 User is offline   johnu 

  • PipPipPipPipPipPipPip
  • Group: Advanced Members
  • Posts: 4,833
  • Joined: 2008-September-10
  • Gender:Male

Posted 2021-October-29, 17:23

View Postsmerriman, on 2021-October-29, 15:18, said:

It doesn't provide any useful signals ever. According to the system notes it's meant to signal in some situations, but does this so inconsistently that it's impossible to get any meaningful information from this.


Even if GIB only gets the signals right 50% of the time, that's still 50% of the time giving the right signal. That's as good as a successful finesse.
1

#4 User is offline   kgr 

  • PipPipPipPipPipPipPip
  • Group: Advanced Members
  • Posts: 3,423
  • Joined: 2003-April-11

Posted 2021-October-30, 01:54

View Postsmerriman, on 2021-October-29, 15:18, said:

It doesn't provide any useful signals ever. According to the system notes it's meant to signal in some situations, but does this so inconsistently that it's impossible to get any meaningful information from this.
I had the idea it often gives parity, but I can be wrong

View Postjohnu, on 2021-October-29, 17:23, said:

Even if GIB only gets the signals right 50% of the time, that's still 50% of the time giving the right signal. That's as good as a successful finesse.
correct...Is GIB using this knowledge that his partner signals sometimes in some situations? If not, then it is better for GIB to never signal and why is it signalling sometimes?
0

#5 User is offline   kgr 

  • PipPipPipPipPipPipPip
  • Group: Advanced Members
  • Posts: 3,423
  • Joined: 2003-April-11

Posted 2021-October-30, 04:26

View Postsmerriman, on 2021-October-29, 15:18, said:

See here for some analysis of both situations.

That is an interesting post :)
It seems GIB has some rules on how it leads and also on how to signal on the lead of partner's Ace. And GIB deviates sometimes randomly from these rules (less on the leads).
I have some questions:
- Is the original developer of GIB not longer around and nobody really knows how it works?
- The convention card says that primary signal to partner's lead is Count:
- - For partner's lead of an Ace this seams to be attitude (showing K or Q). (I almost never lead an Ace, so I'm surprised that it has a rule only in that case)
- - Is it count otherwise or is convention card wrong?
- If it uses another signal then I would expect standard count in some situations. E.g. when following to a suit lead by opps; and when count is very important to block a long suit in dummy without other entrees; when partner leads another honor than the Ace; and when discarding in the majority of cases. => that would be another interesting simulation (I have no idea how much work this is for you).
0

#6 User is offline   smerriman 

  • PipPipPipPipPipPipPip
  • Group: Advanced Members
  • Posts: 3,720
  • Joined: 2014-March-15
  • Gender:Male

Posted 2021-October-30, 13:57

Here's a classic example:


When declarer plays a diamond at trick 2, West should obviously give count. I redealt this hand many times, varying the spot cards, and West played its lowest diamond every time (and East held up to the third round every time).

But don't think this means the robots play upside down count.



I had to force GIB to lead a spade here since it preferred a diamond, and South to continue with a diamond as the order of play sometimes differs, but regardless, when declarer plays diamonds, West almost* always plays low again, and East holds up until the third round.

*On one occasion West GIB held 987, and played the 8. But followed up on the next round with the 9, lol.

And if I sit in West's seat and play the 9 in the first example, East still holds up to the third round.

So West doesn't signal, and East doesn't interpret the signal.

View Postkgr, on 2021-October-30, 04:26, said:

- Is the original developer of GIB not longer around and nobody really knows how it works?

The original developer of GIB passed it onto BBO a long time ago. But yes, it seems nobody has a clue how GIB works, or wants to spend time figuring it out. (Have offered to do this for free several times and while I was hopeful this might eventuate at one point, that hope is gone :( )

View Postkgr, on 2021-October-30, 04:26, said:

- The convention card says that primary signal to partner's lead is Count:

It says here that GIB does *not* give count on the opening lead, nor on discards. Of course, the rest of that paragraph has already been proven to be false. It does say it tries to give count in other situations, but this is also provably wrong as above.

PS - I believe johnu's post was, as per most of them, sarcastic - signalling correctly 50% of the time and wrongly 50% of the time is equivalent to playing randomly.
0

#7 User is offline   pilowsky 

  • PipPipPipPipPipPipPip
  • Group: Advanced Members
  • Posts: 3,620
  • Joined: 2019-October-04
  • Gender:Male
  • Location:Israel

Posted 2021-October-30, 14:43

The writing of the part of the GIB system notes on defence was outsourced to Python (Monty) LLC.
The description is so anodyne that it seems that the takeaway message is "GIB knows what cards have been played and can work out (by counting) what's left." Literally, count. Not UDCA or rectify or any other type of count.

It also doesn't give an attitude signal - even a bad one.
The phrase "usually give" is completely devoid of useful meaning.
It's like saying "when driving a car GIB will 'usually' stop for red lights" - not encouraging for passengers.


Computer scientists enjoy talking about artificial intelligence but if it's artificial then it isn't intelligent.
What I mean by that is that if it was intelligent (in the way that you and I are) then GIB would do things against its best interests sometimes.
GIB doesn't. It will always do the same thing in the same situation.
This is why in smerriman's example he had to 'force' GIB to play in the way he wanted.
You can take a horse to water, and randomly, it will drink.
Present GIB with the same hand and the same bidding sequence and it will always do the same thing.

JohnCleeese said:

It's difficult to describe precisely how GIB defends. It doesn't use rules and guidelines, like humans often do. It simulates hands based on the auction, using double dummy analysis to determine the average result of each defensive play, and chooses the one with the best average. Sometimes this simulation comes up with the same choice that a human would make (there's a good reason for some of the guidelines -- they actually work well), but not always (some of our rules of thumb have become popular simply because they're easy to remember and "good enough"). When it has a choice of equivalent cards, it will choose based on leading and signalling conventions.

GIB doesn't interpret your signals or make many inferences from the play, it uses simulations based on the auction. However, it's usually able to figure out that when you lead an honor, it's part of a sequence.

GIB usually leads passively against NT (read the book Winning Notrump Leads to understand why). Don't assume it's leading its longest suit. When you lead, it doesn't assume you're leading your best suit, which is why it doesn't always return the suit like a human would.

In suit contracts, GIB's opening lead is frequently a side singleton or doubleton, to try to get a ruff. When it leads a suit bid by the opponents, this is almost always the reason. Read the book Winning Suit Contract Leads for insight on the way GIB leads against suits.

If it leads an honor that's part of a sequence, it uses standard honor leads (K from AKx, A from AK doubleton). If it leads from a long suit, it leads 4th best (but see above: it doesn't always lead its long suit). When leading from 3 small, it leads low against both suit and NT contracts.

It doesn't use any signals when making discards, it just tries to make safe discards. In a suit contract it will frequently discard from a short suit while it has trumps left. Otherwise, it tends to discard from a long suit that's safe to shorten.

When it's following to partner's opening lead, it will usually give an attitude signal:
High spot card with an Ace or King
High spot card with a Queen behind dummy's Ace or King
Low in any other situation
Note that it doesn't give count in this situation, so it's hard to know when you can give it a ruff.

When it's trying to win the trick in third hand, it will play the lowest of equals. Otherwise, when following suit it usually gives standard count signals (high = even); an exception is when it's forced to play equivalent cards in a doubleton, it will randomize them because of "restricted choice".

0

#8 User is offline   smerriman 

  • PipPipPipPipPipPipPip
  • Group: Advanced Members
  • Posts: 3,720
  • Joined: 2014-March-15
  • Gender:Male

Posted 2021-October-30, 15:39

View Postpilowsky, on 2021-October-30, 14:43, said:

Present GIB with the same hand and the same bidding sequence and it will always do the same thing.

As you must be aware based on your other threads on the same topic, this is not true. Any time simulations are involved (with basic GIB during the play, or with advanced GIB during both bidding and play), GIB will make different decisions with the same hand and bidding sequence.

The exception is eg when playing in a robot tournament/challenge, when the random number generator is seeded to ensure that this won't happen across tables.

There is also some aspect of seeding at teaching tables to ensure consistency. (Though you can work around this by rotating the deal and you'll often see GIB choose a different line.)

But when playing with robots in the MBC you occasionally see variations at other tables playing with robots that go through the same sequences.
0

#9 User is offline   thepossum 

  • PipPipPipPipPipPip
  • Group: Advanced Members
  • Posts: 2,362
  • Joined: 2018-July-04
  • Gender:Male
  • Location:Australia

Posted 2021-October-30, 18:55

View Postkgr, on 2021-October-29, 08:15, said:



I wonder what 'agreements' Gib has on play and how it plays:



None with me, that's for sure

Do people realise that a clock that doesn't function at all is actually more precise a couple of times a day than most other functioning clocks

I've almost given up making sense of leads but I still often lead back once or even twice if I get a chance hoping for something

The most likely singleton lead is a trump

I used to try and show attitude from time to time but that didn't seem to make any difference either
0

#10 User is offline   kgr 

  • PipPipPipPipPipPipPip
  • Group: Advanced Members
  • Posts: 3,423
  • Joined: 2003-April-11

Posted 2021-October-31, 02:42

View Postsmerriman, on 2021-October-30, 13:57, said:

...
Thanks again for this! I'm convinced: GIB uses some logic for its initial lead; and signals K or Q on the initial lead of partners A.
I think it will also win a trick in defense with the lowest card. e.g. with KQxxx always play the Q.
And I assume that leads in later tricks also have some logic (eg. K from KQx), but a lot less logic (eg not 4th best).
You have done your analysis with 92 and 985 and it always played low. Maybe it would be more random with 42 or 642, because then simulations of playing the 4 and the 2 would more likely give the same result. Or does GIB prefer to play the lowest card, even if all DD simulations for lowest and 2nd lowest give same result. (I don't think so).

View Postsmerriman, on 2021-October-30, 13:57, said:

The original developer of GIB passed it onto BBO a long time ago. But yes, it seems nobody has a clue how GIB works, or wants to spend time figuring it out. (Have offered to do this for free several times and while I was hopeful this might eventuate at one point, that hope is gone :( )
I see here that still changes are done and new versions come out. So, somebody must know the code.
It would certainly be interesting if you were given the possibility to figure out how GIB works...and then improve it (use some signals, make it understand that opps don't play DD).
Sometimes I play on another bridge site against robots and these certainly give count. That is maybe why I was convinced that GIB also does that.

View Postsmerriman, on 2021-October-30, 13:57, said:

*On one occasion West GIB held 987, and played the 8. But followed up on the next round with the 9, lol.
And if I sit in West's seat and play the 9 in the first example, East still holds up to the third round.
So West doesn't signal, and East doesn't interpret the signal.
That is the restricted choice logic that is build in. From equal cards it specifically plays random. This one might even change if you run the same hand more than once.
0

#11 User is offline   smerriman 

  • PipPipPipPipPipPipPip
  • Group: Advanced Members
  • Posts: 3,720
  • Joined: 2014-March-15
  • Gender:Male

Posted 2021-October-31, 04:30

View Postkgr, on 2021-October-31, 02:42, said:

You have done your analysis with 92 and 985 and it always played low.

No, I was randomising the spot cards each time; the above were just two examples, but it consistently does not signal with any combinations.

View Postkgr, on 2021-October-31, 02:42, said:

I see here that still changes are done and new versions come out. So, somebody must know the code.

Past tense. GIB hasn't been updated in what's getting close to 3 years, and the only developer who used to interact with the forum left just before that.
1

#12 User is offline   kgr 

  • PipPipPipPipPipPipPip
  • Group: Advanced Members
  • Posts: 3,423
  • Joined: 2003-April-11

Posted 2021-October-31, 05:07

View Postsmerriman, on 2021-October-31, 04:30, said:

GIB hasn't been updated in what's getting close to 3 years, and the only developer who used to interact with the forum left just before that.
That is not good, considering the amount of deals that are played every day against GIB.

I verified the discards on the deals I played Today. First discard was always count (L-H odd); Second discard mostly count.
Only 22 deals, so no conclusion yet. (Low from 5-cards can be simply low and not count intended)...But could be interesting for a simulation.
1

#13 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 2021-October-31, 16:31

View Postsmerriman, on 2021-October-31, 04:30, said:

GIB hasn't been updated in what's getting close to 3 years, and the only developer who used to interact with the forum left just before that.
:( :( :(
0

#14 User is offline   pilowsky 

  • PipPipPipPipPipPipPip
  • Group: Advanced Members
  • Posts: 3,620
  • Joined: 2019-October-04
  • Gender:Male
  • Location:Israel

Posted 2021-October-31, 17:07

GIB signalling appears to be a rather blunt instrument.
I would be interested to know what experienced robot wranglers think.

The books by Bird and Anthias give useful insights into GIB behaviour; many are translatable into actual play.
0

Page 1 of 1
  • 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