BBO Discussion Forums: State of the Art 3 - BBO Discussion Forums

Jump to content

  • 2 Pages +
  • 1
  • 2
  • You cannot start a new topic
  • This topic is locked

State of the Art 3 Combining chances - I would bet no robot will do this

#1 User is offline   Scarabin 

  • PipPipPipPip
  • Group: Full Members
  • Posts: 382
  • Joined: 2010-December-30
  • Gender:Male
  • Interests:All types of games especially bridge & war games.
    old bidding systems & computer simulation programming.

Posted 2012-January-31, 00:55



Same bidding for all robots.

As background this is a hand where we should try everything else before taking a spade finesse and since everything else is favourable we don't need the fatal spade finesse. However all the other plays have lower individual probabilities than the finesse.

There is nothing much to the play except to see if any robot came close to making the contract. Incidently I sat West and East did not commit any blunders.

In all cases declarer won the diamond lead with the Q.

Shark Bridge then led the 7 to the Q, the 7 to the K, and finessed the J.

Wbridge5 cashed the K and A and finessed the Q.

Jack Bridge also cashed the K & A and finessed the Q.

For a moment I thought GIB was going to disprove my prediction; he cashed the A,K, & Q before crossing to the A and finessing the J.

Why did all four robots miss a fairly basic play and when they cashed various numbers of club tricks were they working toward the right play?

Maybe Bridgify 104 can provide an answer: _ After declarer has won the first trick Bridgify shows his hand like this:
JT75AK434AK7
--8888--8877---8---888

the subscript figures are the number of tricks made if the card is led, so that there is no indication he should not lead a spade. As long as he does not lead a spade this pattern persists: that he may lead any card but the 3 or 4 of hearts.

If the lead is in dummy then Bridgify shows dummy must not lead the Q or 2 of spades and if declarer does lead a spade from his hand then Bridgify shows he must play the A.

Remembering that a new simulation is performed for every play I think the play of the clubs and hearts is purely random and the simulations keep showing the finesse is the best percentage chance. _ And yet GIB came so close and never even cashed the long club!
0

#2 User is offline   inquiry 

  • PipPipPipPipPipPipPipPipPipPip
  • Group: Admin
  • Posts: 14,563
  • Joined: 2003-February-13
  • Gender:Male
  • Location:Amelia Island, FL
  • Interests:Bridge, what else?

Posted 2012-January-31, 22:04

The problem seems to be the robots play bridge differently than you do, with different goals.

IF you try the two hearts before the spade finesse (assuming the lucky Qx) is doesn't happen, you will go down several extra tricks (lose diamonds, and a few hearts). I also think you overlooked the possibility that diamonds might be divided 4-4 (since you see the ew hands, you don't consider this possibility perhaps). The robots figure net gain (average number of points) for one line versus the other. Cashing the heart king increases, on average, the number of undertricks while (at 50 or 100 points per) while not cashing it stops some of the bleeding. In addition, after a successful spade finesse, you rack up a couple of spade tricks, plus if the queen drops that is just extra. In addition, cashing the heart ak leads to disaster if the heart queen does not fall and diamonds are 4-4.

So let;s imagine this from a small sample of 100 deals. On these deals, the spade finesse will work 50% of the time. The heart queen will fall doubleton 16 times (8 times when the finesse wins, eight times when it does not). Further, we will ignore things like blocked diamond suit, but we will include the possibility of 4-4 diamonds, and 5-3 and 6-2 diamonds. We know diamonds are not 8-0, and lets assume we know they are not 7-1, so diamonds will split 4-4, 45 times, where you always make with the spade finesse (they win 3 diamonds and the spade) and split 5-3 or 6-2 the other 55 times. (the numbers below may clearly be wrong estimates, I am doing this for illustration purposes, the big difference is the robots don't know about 5-3 diamonds, so they take into account possibility of 4-4 where the finesse is not instant death).

So what has the robot's simulations done for us?
BY PLAYing the spade finesse without the cashing heart AK, they make 9 tricks on half the 55 times the diamonds are not 44, the other 1/2 the time, they make at least 10 tricks (3+2+1+4), with chances for additional tricks if the queen falls, or if spades split nicely. Let's estimate that in the robot simulations, the spade suit divided 3-3 or the queen dropped doubleton about 1/3 of the time the spade finesse won, this gives the robots line of success...

WHEN spade finesse wins, robot will 10 tricks (50 * 2/3 x 430 = 33 x 430 = 14,190 points)
when spade finesse wins, and heart queen falls or spades friendly (17 * 460 = 7,820 points)
WHEN spade finesse failed but diamonds were 4-4 (50 x .45% = x 400 = 9,000 points)
WHEN spade finesse failed, and diamonds 5-3 (50 * .48% x -50 = 24.5 x -50 = -2450 points)
when spade finesse failed, and diamonds 6-2 (1.5 -100 = -150 points)
33 hands at = 14,190
17 hands at = 7,820
22.5 hands at + 9,000
26 hands at -1225
1.5 hands at - 150
100 hands = 33930 points = 339.3 points/hand on average (note the big bump for 4-4 diamond split you will not see below, and less negative results when heart queen does not fall.

Cashing the heart AK before taking the spade finesse will gain 9 tricks right way 16 times, I assume for this comparison that the robots will forgo the finesse in this case, but this might be wrong, but for the calculations lets say yes.... so cashing the hearts

Cashing hearts 16 x 400 = 6400 points (the 16 out of 100 times the heart queen is doubleton, we are working on heart queen not singleton, so odds might be slightly higher now that the queen is doubleton, since void and singleton heart queen are no longer possible.
half the time cashing heart fails, the spade finesse will win (100-16 = 84/2 = 42 * 430 = 18,060), we could throw in a tiny bit more for lucky spades, but heart queen falling has already failed, so I will add a bonus 42*30%*30 for an extra trick, raising this by 378 points to 18,438
when the spade finesse fails, you always go down, at least two (3D, 1h, 1S), and sometimes 3 or more (couple hearts, 4 or more diamonds).
Lets say down just two when diamonds are 4-4 and down three other times (can be more, a 5th diamond, or a 2nd heart).
So, of the 42 times the finesse loses, you are down two 45% of the time (diamonds 4-4, or 42*.45 * -100 = 19 * -100 = -1900 points
The other 23 time you are down three (23 x -150 = -3450 points)
Total for the 100 hands,
16 at = 6,400
42 at =18.438
19 at - 1,900
23 at -150
100 = 19,488 = +195 points/hand for play heart queen then hook....

One would think you gained some chances cashing the heart AK before taking the spade finesse on you line play. HEART Ace-king drops the queen doubleton 16% of the time removing need for finesse, and then finesse wins 1/2 of the remaining time (for .5 x 84% = 42%) for a total of 58%. The robot looks at it differently. if the spade finesse fails, you still make if diamonds are 3-3 (35% of the time), as long as you don't cash the AK of hearts setting up the setting trick for the defenders. So they see 50% (with overtricks) + 1/2 of the 4-4 diamond splits (and since 8=0 and 0=8 are no longer possible, they figure the odds of 4=4 around 40%, i figured them a little higher, above since 7-1 you make anyway). So 50% + 1/2 40% = 70%.


So the robot, over the course of a very long match (infinite) would gain 144.3 or 4 imps on average to your play. Who is right, and who is wrong. NOT sure. In a short match, I surely don't want to go down in this game. So combining all the chances to go plus makes a lot of sense. On the other hand, you can see the 5-3 diamond split, so you know you are down, at the table, I, like the robot, can not see how diamonds split. And even if they are 5-3, is it worth the risk for an extra down one? Cashing the top three clubs and the heart AK before taking the spade finesse risk down two when diamonds are 5-3 instead of down one a lot more frequently than one might expect (you need both qx of hearts and 3-3 clubs, that is 16% x 35.5% = about 5%. So your risk -100 on the 5-3 split (not to mention going down when diamonds are 4-4) 95% of the time. Even if we ignore the 44 diamond split, your line gains 10 imps 5% of the time, but loses 2 imps 95% of the time. In the long run, even assuming no 4-4 diamond split, you line seems to be too costly (net gain on 100 hands, 50 imps making five times, minus 190 not making 95 times (costing 2 imps each time down). So put me down for thinking the robots played this one correctly (I would not cash akq of clubs like gib did, so gib actually messed it up I think).
--Ben--

#3 User is offline   Scarabin 

  • PipPipPipPip
  • Group: Full Members
  • Posts: 382
  • Joined: 2010-December-30
  • Gender:Male
  • Interests:All types of games especially bridge & war games.
    old bidding systems & computer simulation programming.

Posted 2012-January-31, 22:53

Thank you for that prompt and detailed reply.

Your analysis is masterly and I would not seek to refute it except to make the minor point that the opening lead was the 6, which I failed to specify, and that a 5-3 split is more likely than 4-4. I must confess that because my analysis is basic I seek out deals which have been the subject of analysis and I ignore match-point play.

But, and this is the real point, your analysis is that of a bridge master and has nothing in common with a dumb robot or a monte carlo simulation? _ I am convinced that the last 4 paragraphs of the original post accurately describe the operation of the simulations, and I would be very happy to have you or Barmar confirm or deny this.

I think the side plays are really random plays of equally permissible cards which we unfortunately endow with human motives and insights. - You will recall that you previously spoke of Jack and GIB making unblocking plays, and I wondered how Wbridge5 could discard a key K? I have also credited Jack with entry-killing plays, probably because I can be careless about entries when I am not alert.

In replying to a post "Why not lead A or K from AK", johnu observed that GIB discards at random without reference to rank. - I fear this is literally true and is the real reason for plays we try to endow with intelligence?

Of course I realise that a program that is pure monte carlo simulation would not work: it has to be overlaid with some sort of pragmatic interface but given this, would you agree with my contention?
0

#4 User is offline   cloa513 

  • PipPipPipPipPipPip
  • Group: Advanced Members
  • Posts: 1,446
  • Joined: 2008-December-02

Posted 2012-February-01, 01:14

View Postinquiry, on 2012-January-31, 22:04, said:

The problem seems to be the robots play bridge differently than you do, with different goals.

IF you try the two hearts before the spade finesse (assuming the lucky Qx) is doesn't happen, you will go down several extra tricks (lose diamonds, and a few hearts). I also think you overlooked the possibility that diamonds might be divided 4-4 (since you see the ew hands, you don't consider this possibility perhaps). The robots figure net gain (average number of points) for one line versus the other. Cashing the heart king increases, on average, the number of undertricks while (at 50 or 100 points per) while not cashing it stops some of the bleeding. In addition, after a successful spade finesse, you rack up a couple of spade tricks, plus if the queen drops that is just extra. In addition, cashing the heart ak leads to disaster if the heart queen does not fall and diamonds are 4-4.

So let;s imagine this from a small sample of 100 deals. On these deals, the spade finesse will work 50% of the time. The heart queen will fall doubleton 16 times (8 times when the finesse wins, eight times when it does not). Further, we will ignore things like blocked diamond suit, but we will include the possibility of 4-4 diamonds, and 5-3 and 6-2 diamonds. We know diamonds are not 8-0, and lets assume we know they are not 7-1, so diamonds will split 4-4, 45 times, where you always make with the spade finesse (they win 3 diamonds and the spade) and split 5-3 or 6-2 the other 55 times. (the numbers below may clearly be wrong estimates, I am doing this for illustration purposes, the big difference is the robots don't know about 5-3 diamonds, so they take into account possibility of 4-4 where the finesse is not instant death).

So what has the robot's simulations done for us?
BY PLAYing the spade finesse without the cashing heart AK, they make 9 tricks on half the 55 times the diamonds are not 44, the other 1/2 the time, they make at least 10 tricks (3+2+1+4), with chances for additional tricks if the queen falls, or if spades split nicely. Let's estimate that in the robot simulations, the spade suit divided 3-3 or the queen dropped doubleton about 1/3 of the time the spade finesse won, this gives the robots line of success...

WHEN spade finesse wins, robot will 10 tricks (50 * 2/3 x 430 = 33 x 430 = 14,190 points)
when spade finesse wins, and heart queen falls or spades friendly (17 * 460 = 7,820 points)
WHEN spade finesse failed but diamonds were 4-4 (50 x .45% = x 400 = 9,000 points)
WHEN spade finesse failed, and diamonds 5-3 (50 * .48% x -50 = 24.5 x -50 = -2450 points)
when spade finesse failed, and diamonds 6-2 (1.5 -100 = -150 points)
33 hands at = 14,190
17 hands at = 7,820
22.5 hands at + 9,000
26 hands at -1225
1.5 hands at - 150
100 hands = 33930 points = 339.3 points/hand on average (note the big bump for 4-4 diamond split you will not see below, and less negative results when heart queen does not fall.

Cashing the heart AK before taking the spade finesse will gain 9 tricks right way 16 times, I assume for this comparison that the robots will forgo the finesse in this case, but this might be wrong, but for the calculations lets say yes.... so cashing the hearts

Cashing hearts 16 x 400 = 6400 points (the 16 out of 100 times the heart queen is doubleton, we are working on heart queen not singleton, so odds might be slightly higher now that the queen is doubleton, since void and singleton heart queen are no longer possible.
half the time cashing heart fails, the spade finesse will win (100-16 = 84/2 = 42 * 430 = 18,060), we could throw in a tiny bit more for lucky spades, but heart queen falling has already failed, so I will add a bonus 42*30%*30 for an extra trick, raising this by 378 points to 18,438
when the spade finesse fails, you always go down, at least two (3D, 1h, 1S), and sometimes 3 or more (couple hearts, 4 or more diamonds).
Lets say down just two when diamonds are 4-4 and down three other times (can be more, a 5th diamond, or a 2nd heart).
So, of the 42 times the finesse loses, you are down two 45% of the time (diamonds 4-4, or 42*.45 * -100 = 19 * -100 = -1900 points
The other 23 time you are down three (23 x -150 = -3450 points)
Total for the 100 hands,
16 at = 6,400
42 at =18.438
19 at - 1,900
23 at -150
100 = 19,488 = +195 points/hand for play heart queen then hook....

One would think you gained some chances cashing the heart AK before taking the spade finesse on you line play. HEART Ace-king drops the queen doubleton 16% of the time removing need for finesse, and then finesse wins 1/2 of the remaining time (for .5 x 84% = 42%) for a total of 58%. The robot looks at it differently. if the spade finesse fails, you still make if diamonds are 3-3 (35% of the time), as long as you don't cash the AK of hearts setting up the setting trick for the defenders. So they see 50% (with overtricks) + 1/2 of the 4-4 diamond splits (and since 8=0 and 0=8 are no longer possible, they figure the odds of 4=4 around 40%, i figured them a little higher, above since 7-1 you make anyway). So 50% + 1/2 40% = 70%.


So the robot, over the course of a very long match (infinite) would gain 144.3 or 4 imps on average to your play. Who is right, and who is wrong. NOT sure. In a short match, I surely don't want to go down in this game. So combining all the chances to go plus makes a lot of sense. On the other hand, you can see the 5-3 diamond split, so you know you are down, at the table, I, like the robot, can not see how diamonds split. And even if they are 5-3, is it worth the risk for an extra down one? Cashing the top three clubs and the heart AK before taking the spade finesse risk down two when diamonds are 5-3 instead of down one a lot more frequently than one might expect (you need both qx of hearts and 3-3 clubs, that is 16% x 35.5% = about 5%. So your risk -100 on the 5-3 split (not to mention going down when diamonds are 4-4) 95% of the time. Even if we ignore the 44 diamond split, your line gains 10 imps 5% of the time, but loses 2 imps 95% of the time. In the long run, even assuming no 4-4 diamond split, you line seems to be too costly (net gain on 100 hands, 50 imps making five times, minus 190 not making 95 times (costing 2 imps each time down). So put me down for thinking the robots played this one correctly (I would not cash akq of clubs like gib did, so gib actually messed it up I think).


Your analysis is one of a statistician not a bridgeplayer that's why you miss that elementary deduction from the lead (assume opponents are trying to get your contract down actively)- it is so likely to be from AKxxx or xxx (a worthless hand with no chance of getting the lead again) with that terrible lead you have to play for good breaks without finessing as much as possible- the computer programmes aren't bridge playing orientated enough- most do better than GIB but still miss a lot.
0

#5 User is offline   Scarabin 

  • PipPipPipPip
  • Group: Full Members
  • Posts: 382
  • Joined: 2010-December-30
  • Gender:Male
  • Interests:All types of games especially bridge & war games.
    old bidding systems & computer simulation programming.

Posted 2012-February-01, 03:09

View Postcloa513, on 2012-February-01, 01:14, said:

Your analysis is one of a statistician not a bridgeplayer that's why you miss that elementary deduction from the lead (assume opponents are trying to get your contract down actively)- it is so likely to be from AKxxx or xxx (a worthless hand with no chance of getting the lead again) with that terrible lead you have to play for good breaks without finessing as much as possible- the computer programmes aren't bridge playing orientated enough- most do better than GIB but still miss a lot.
Forgive me for intervening but I feel I may have lured Inquiry into an ambush and I think your comments may be intended for me rather than Inquiry. Certainly I fit the description: weak bridge analyst with some competence as a statistician whereas Inquiry is a pretty good analyst and bridge player.

Inquiry seeks to convince me that the robot plays are consistent with intelligent analysis and I seek to prove they're not. To put it bluntly I am accusing him of anthromorphicism (that is not a word I use often)and confessing to this as well.

The joke is we are probably both right in our analyses and I don't know how we can establish absolute truth but perhaps we may reach agreement. We are all reasonable men.
0

#6 User is offline   Antrax 

  • PipPipPipPipPipPip
  • Group: Advanced Members
  • Posts: 2,458
  • Joined: 2011-March-15
  • Gender:Male

Posted 2012-February-01, 04:53

I'm not sure that's a fruitful line of discussion, though. Programs don't have to do things the way we think we do things to excel at them. The most obvious example is chess, where programs compensate for their lack of pattern recognition with brute force.
What I think people are saying is not that a GIB-style approach can't work, but that it might be a good idea to shore it up with some human-style heuristics. I've been reading this forum for a while, and it seems that what people basically want are some rules like "don't ever pass a bid that you know is not natural", "all things being equal, return my suit / don't discard top honors even if it doesn't seem to matter", etc.
I do think inquiry did a good job explaining why this time, such a heuristic ("secure the contract even if it risks extra undertricks"?) would cause GIB to play worse.
0

#7 User is offline   Scarabin 

  • PipPipPipPip
  • Group: Full Members
  • Posts: 382
  • Joined: 2010-December-30
  • Gender:Male
  • Interests:All types of games especially bridge & war games.
    old bidding systems & computer simulation programming.

Posted 2012-February-02, 23:21

View PostAntrax, on 2012-February-01, 04:53, said:

I'm not sure that's a fruitful line of discussion, though. Programs don't have to do things the way we think we do things to excel at them. The most obvious example is chess, where programs compensate for their lack of pattern recognition with brute force.
What I think people are saying is not that a GIB-style approach can't work, but that it might be a good idea to shore it up with some human-style heuristics. I've been reading this forum for a while, and it seems that what people basically want are some rules like "don't ever pass a bid that you know is not natural", "all things being equal, return my suit / don't discard top honors even if it doesn't seem to matter", etc.
I do think inquiry did a good job explaining why this time, such a heuristic ("secure the contract even if it risks extra undertricks"?) would cause GIB to play worse.


I think your comments correct and perceptive and I would only take issue with the relevance of your last sentence. I agree Inquiry did a very good job of analysing GIB's play from a human standpoint but ask yourself this: does GIB contain any heuristics (other than those necessarily contained in a double-dummy program and in an unseen hand sample simulation program)and does GIB just play to make the maximum number of tricks irrespective of contract target?

I think you are correct that my present line of approach (of building a gradual case by examples) is proving unfruitful and I propose to scrap this and go for broke in SOA 4.
0

#8 User is offline   barmar 

  • PipPipPipPipPipPipPipPipPipPipPip
  • Group: Admin
  • Posts: 16,450
  • Joined: 2004-August-21
  • Gender:Male

Posted 2012-February-03, 11:01

View PostScarabin, on 2012-February-02, 23:21, said:

I think your comments correct and perceptive and I would only take issue with the relevance of your last sentence. I agree Inquiry did a very good job of analysing GIB's play from a human standpoint but ask yourself this: does GIB contain any heuristics (other than those necessarily contained in a double-dummy program and in an unseen hand sample simulation program)and does GIB just play to make the maximum number of tricks irrespective of contract target?

I believe GIB plays to maximize the expected value of the score, not necessarily the maximum number of tricks. This takes into account the form of scoring: MP, IMP, or total points. However, for MP this requires estimating what "the field" will do, and I'm not confident in its ability to do this well. We've seen it go down in thin, but cold, slams, because it took a risk for an overtrick, presumably because it assumed everyone would be in the same contract.

#9 User is offline   Scarabin 

  • PipPipPipPip
  • Group: Full Members
  • Posts: 382
  • Joined: 2010-December-30
  • Gender:Male
  • Interests:All types of games especially bridge & war games.
    old bidding systems & computer simulation programming.

Posted 2012-February-03, 20:48

View Postbarmar, on 2012-February-03, 11:01, said:

I believe GIB plays to maximize the expected value of the score, not necessarily the maximum number of tricks. This takes into account the form of scoring: MP, IMP, or total points. However, for MP this requires estimating what "the field" will do, and I'm not confident in its ability to do this well. We've seen it go down in thin, but cold, slams, because it took a risk for an overtrick, presumably because it assumed everyone would be in the same contract.
_ Surely this must require some addition of pragmatic reasoning to the basic random simulation? I have difficulty in believing that my copy of GIB can do this. Perhaps the BBO version has been developed far beyond the Ginsberg version?
0

#10 User is offline   barmar 

  • PipPipPipPipPipPipPipPipPipPipPip
  • Group: Admin
  • Posts: 16,450
  • Joined: 2004-August-21
  • Gender:Male

Posted 2012-February-04, 02:06

I think it's basically how GIB has worked for many years -- I don't think there have been any major changes to the design of the program since Ginsberg sold it to BBO (most of the work has been on the bidding database, only minor tweaks have been made to the program itself). If you use the CLI version, it has the following command line options:

-I IMP scoring
-J matchpoint scoring
-K num num total-point scoring; nums are N/S and E/W part scores
-L par scoring (trust some signals, increase time for declarer play only)

These options are used in converting the trick score into an actual score.

Have you read Ginsberg's paper describing the original design of GIB?

http://www.jair.org/...0-1957-jair.pdf

Section 5 describes some refinements to solve the problem of simple Monte Carlo simulation; I think this is what became the GIBson algorithm in GIB.

I got a good laugh out of this quote:

Quote

In 1997, Martel, a computer scientist himself, suggested that he expected gib to be the best bridge player in the world in approximately 2003. The work appears to be roughly on schedule.

Chip is in good company -- many great computer scientists have underestimated the difficulty of AI. In the 60's, the prediction of a computer like HAL 9000 by 2001 was not considered outlandish. We're now a decade later, and the best we have are Siri and Watson.

#11 User is offline   Scarabin 

  • PipPipPipPip
  • Group: Full Members
  • Posts: 382
  • Joined: 2010-December-30
  • Gender:Male
  • Interests:All types of games especially bridge & war games.
    old bidding systems & computer simulation programming.

Posted 2012-February-05, 23:58

View Postbarmar, on 2012-February-04, 02:06, said:

I think it's basically how GIB has worked for many years -- I don't think there have been any major changes to the design of the program since Ginsberg sold it to BBO (most of the work has been on the bidding database, only minor tweaks have been made to the program itself). If you use the CLI version, it has the following command line options:

-I IMP scoring
-J matchpoint scoring
-K num num total-point scoring; nums are N/S and E/W part scores
-L par scoring (trust some signals, increase time for declarer play only)

These options are used in converting the trick score into an actual score.

Have you read Ginsberg's paper describing the original design of GIB?

http://www.jair.org/...0-1957-jair.pdf

Section 5 describes some refinements to solve the problem of simple Monte Carlo simulation; I think this is what became the GIBson algorithm in GIB.

I got a good laugh out of this quote:

Chip is in good company -- many great computer scientists have underestimated the difficulty of AI. In the 60's, the prediction of a computer like HAL 9000 by 2001 was not considered outlandish. We're now a decade later, and the best we have are Siri and Watson.


Thanks for that very helpful info. I have downloaded Ginsberg's paper which is new to me, and I am poring over it in detail.
On a semi-related point,I discovered recently that Google can access our posts. It may be that I am the only person using BBO who did not realize this but if not should we warn users not to disclose personal details they would not wish to see on the internet?
0

#12 User is offline   Antrax 

  • PipPipPipPipPipPip
  • Group: Advanced Members
  • Posts: 2,458
  • Joined: 2011-March-15
  • Gender:Male

Posted 2012-February-06, 00:02

Assume anything you write anywhere on the internet is public.
2

#13 User is offline   blackshoe 

  • PipPipPipPipPipPipPipPipPipPipPip
  • Group: Advanced Members
  • Posts: 15,799
  • Joined: 2006-April-17
  • Location:Rochester, NY

Posted 2012-February-06, 05:37

"You are being watched. The government has a secret system. A machine. It spies on you every hour of every day. I know, because… I built it." -- "Mr. Finch", in "Person of Interest".
--------------------
As for tv, screw it. You aren't missing anything. -- Ken Berg
I have come to realise it is futile to expect or hope a regular club game will be run in accordance with the laws. -- Jillybean
0

#14 User is offline   barmar 

  • PipPipPipPipPipPipPipPipPipPipPip
  • Group: Admin
  • Posts: 16,450
  • Joined: 2004-August-21
  • Gender:Male

Posted 2012-February-06, 09:43

I'm not sure why you'd be surprised that Google can read our posts. You only need a password to post, not to read here.

#15 User is offline   MrAce 

  • VIP Member
  • PipPipPipPipPipPipPipPip
  • Group: Advanced Members
  • Posts: 6,659
  • Joined: 2009-November-14
  • Gender:Male
  • Location:Houston, TX

Posted 2012-February-09, 21:03

View PostScarabin, on 2012-February-01, 03:09, said:


Inquiry seeks to convince me that the robot plays are consistent with intelligent analysis and I seek to prove they're not. To put it bluntly I am accusing him of anthromorphicism (that is not a word I use often)and confessing to this as well.


You are correct, bot plays are not consistent with intelligent analysis. First of all they can not play single dummy hands. You can either set them to GODLIKE mode which they play double dummy, or u can disable them from playing dd and then they make things that can not be explained by any logic. The example hand you choosed was not a good one.

Here is one that i played last night. Look at what GIB did at trick 8 :) (you can click next button to see the play trick by trick)




"Genius has its own limitations, however stupidity has no such boundaries!"
"It's only when a mosquito lands on your testicles that you realize there is always a way to solve problems without using violence!"





0

#16 User is offline   barmar 

  • PipPipPipPipPipPipPipPipPipPipPip
  • Group: Admin
  • Posts: 16,450
  • Joined: 2004-August-21
  • Gender:Male

Posted 2012-February-10, 12:51

Not that it should matter, but was it IMPs or MP?

#17 User is offline   MrAce 

  • VIP Member
  • PipPipPipPipPipPipPipPip
  • Group: Advanced Members
  • Posts: 6,659
  • Joined: 2009-November-14
  • Gender:Male
  • Location:Houston, TX

Posted 2012-February-10, 16:49

View Postbarmar, on 2012-February-10, 12:51, said:

Not that it should matter, but was it IMPs or MP?



IMPs but as you said it doesnt matter, should not matter for that type of play. Even the worst programmed bot should be able to not to block an established suit willingly and drop a trick or 2.

I actually had a lot of plays like that made by GIB, saved with the intention to post, but for some reason lost my motivation to do so.
"Genius has its own limitations, however stupidity has no such boundaries!"
"It's only when a mosquito lands on your testicles that you realize there is always a way to solve problems without using violence!"





0

#18 User is offline   barmar 

  • PipPipPipPipPipPipPipPipPipPipPip
  • Group: Admin
  • Posts: 16,450
  • Joined: 2004-August-21
  • Gender:Male

Posted 2012-February-10, 23:45

Was it a basic or advanced bot? The basic bots can do some pretty dumb things, since they don't use the GIBson algorithm. The advanced bots also do dumb things, but stuff like this is pretty rare.

#19 User is offline   MrAce 

  • VIP Member
  • PipPipPipPipPipPipPipPip
  • Group: Advanced Members
  • Posts: 6,659
  • Joined: 2009-November-14
  • Gender:Male
  • Location:Houston, TX

Posted 2012-February-11, 04:17

View Postbarmar, on 2012-February-10, 23:45, said:

Was it a basic or advanced bot? The basic bots can do some pretty dumb things, since they don't use the GIBson algorithm. The advanced bots also do dumb things, but stuff like this is pretty rare.


I dunno what u mean but i dont think it is relevant. I tried both advanced (the one that we can use in web version only) and basic there is no difference between each other when it comes to making dumb things. Here is my proof why it doesnt matter which bot because 14 scores and all 14 of the GIBS at different tables opened 1 and rebid 2 with 20 hcp making 2+5 or +4.





The reason of my frustration is this BS (pardon my language) seems to happen much more often after the "so called improvements". Imo while they are trying to improve it they seriously screwed up some other stuff somewhere in the programe.

Here it is , which happened today, GIB my pd holds Kx Jxxx Kxxx AJx opens 1 opps silent

1--1
1NT--2 (and alerts my 2 as retreat from NT weak hand 6+)
3!!!!!!!! But bids 3 with 3 card lol

Look i will be honest, we all know what GIB is capable of and what not by now. Gib will never lead pd's suit when pd overcall, Gib will never let you play 5 minor if you bid game by a jump, both of us coming from pass suddenly GIB will balance and drag you all the way to 5 or 6 level etc etc. But this type of incidents that i posted never occured before the "so called improvements" .

I mean the card play error that i showed in my previous post for example, it really should not matter, basic GIB or advanced, it should be able to at least play a suit without blocking it. I am not a programmer nor i have any knowledge about it. I am about to believe that the issues with GIB is not just some bugs. The bids are not even remotely related to the explenation of itself. GIB alerts 1 and then 2 bid as "Opener rebids his C -- 3-S;11+ HCP rebidable C; 12-16 total points" Yet it has 20 hcp with no wastage and prime cards. I dont think all 14 GIBs were basic.

Please dont take my post as a rant, eventhough it sounds like one. If ranting was my intention i would make a new thread and complain there. I am just trying to tell the OP that i agree with him about thinking the action of GIB is not consistent with intelligent analysis. Maybe in the future another programe, but not this one, not the current GIB imo. Not sure if improvement patches or whatever they do is helping much, perhaps it needs to be redone from the scrtach.
"Genius has its own limitations, however stupidity has no such boundaries!"
"It's only when a mosquito lands on your testicles that you realize there is always a way to solve problems without using violence!"





0

#20 User is offline   barmar 

  • PipPipPipPipPipPipPipPipPipPipPip
  • Group: Admin
  • Posts: 16,450
  • Joined: 2004-August-21
  • Gender:Male

Posted 2012-February-11, 22:20

View PostMrAce, on 2012-February-11, 04:17, said:

I dunno what u mean but i dont think it is relevant.

Basic = $1/week rental
Advanced = $1/day rental and robot tourneys

For that 20 point hand, it's not clear what the right bid is. It has the points to reverse into , but LHO bid them so it can't. It can't bid 3, that requires a 6-card suit. It can't bid 3, that requires a 4-card suit. It can't bid 2NT because of the singleton . Every bid is flawed. Be glad it didn't pass!

In card play, GIB doesn't use logic or think about card combinations, everything is pretty much based on simulations. Sometimes it will block a suit because in the majority of the hands it dealt it didn't matter. This is especially possible with basic bots, because they don't use many hands in their simulations, so they can easily miss some layouts where it matters.

It's true that sometimes the bid explanations don't match the hand. There are two reasons for this. 1) The bid comes from a simulation, which causes it to make a non-system bid because it thinks it's likely to work out better than the system bid for its hand; 2) many low-priority entries in the bidding database don't specify what they show, or they specify something a little different from the criteria required to make the bid, because we expect these entries to be used rarely -- we cater to the more likely descriptions to avoid confusing the bots. The classic example of this is raising responder with only 3-card support -- this is generally only done if there are no other decent options, and it's not included in the description because we want the robot to think it's in an 8-card fit and bid game. Unfortunately, the description language doesn't allow for OR, so we can't say "4+ S OR (3+ S AND (1- H OR 1- D))", which is closer to the actual meaning of 1 1 2.

Share this topic:


  • 2 Pages +
  • 1
  • 2
  • You cannot start a new topic
  • This topic is locked

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