BBO Robot Hands

[Bbradley62]

I am curious, do any of you actually play Robot Bridge?

It is easy enough to confirm (through myhands) that Jack, hrothgar, barmar and I all play lots of robot bridge.

[Bbradley62]

I know you (close enough) duplicated some of the numbers I posted earlier, so we seemed to be doing the same counting, so I wanted to duplicate yours...

“No finesses” includes 2 way finesses or finesses that do not make any difference (ex: one side hold A with QJ in separate hands so two tricks no matter where K is located). The first letters are AKQJ (Ace, King, Queen, Jack, the last of the letter combinations is one of CDHS (clubs, diamonds, hearts or spades). Then there is a gap with a single letter, NSE or W for north, south, east or west.

Tournament 3250 Feb 2

Hand 1: KD S in front AQD W: Loss Stiff, no finesse, AH W in front KQH N: Win

Hand 2: AKS S in front QS W: Loss, KD W in front AQD S: Win, KC E in front QC S: Wintoss-up due to missing both KJ

Hand 3: AQC S in front KC W: Loss, QD E in front AKD S: Win, AH N in front KQH E: LossNo finesse, you're winning one H and losing one H

Hand 4: QH W in front AJH N (KH S): Win. QS E in front AKJS S: Win. KD S in front AD W: LossIf anything it's a win due to Q and T, so you get 2 of the first 3 diamonds, but I'll call it irrelevant

Hand 5: KJS S in front AQ10S W: 2 Losses, QH N in front KH E: Loss, KDE in front AJD S: WinMissing KQ it's a 75% proposition and shouldn't be considered with the 50% ones

Hand 6: AQD S in front KD W: LossSee your explanation about split QJ

Hand 7: AKJC S in front QC W: Loss, KJD S between AD E and QD W: Loss and Win

Hand 8: AQS E in front KS S: Win, KQC S in front AJC W: Loss

Hand 9: No finesses

Hand 10: KC S in front AJC W: Lossirrelevant non-finesse

Hand 11: No finesses

Hand 12: AQC S in front KC W: LossAgain, see comment re: split QJ

Totals: 9 7 Wins, 14 8 Losses

Not nearly as bad as you made it out to be.

[barmar]

Antrax: Please excuse me for using your message to post, but I cannot get the code style to work in Fast Reply

Click on Use Full Editor to get out of the fast reply and into the regular post editor.

Well, your programmer is one cocky little SOB.

I promise you, cross my heart and hope to die, that we have not touched the dealing code in years. The last change we made to it was when we implemented Best Hand robot games, and that was just to implement the part that swaps the human's hand with the best hand.

 

Now please will you let this go. You've been told over and over that your understanding of statistics is flawed. I'm not an expert on statistics, either, that's why I defer to the expertise of people like hrothgar and Helene.

[1eyedjack]

Now please will you let this go.

It is a common enough accusation (and for that reason if no other I suspect shared by a significant silent minority) that if it can be put to bed then some effort to that end is worthwhile.

 

Provided, of course, that someone is prepared to put in an equal effort into testing the hypothesis by their peers as is put into promoting it. Which is no small task to ask someone freely to undertake just for the price of some goodwill.

 

For all that he seems to have some strange notions, tx10s has made the bravest attempt yet at supporting his allegation.

 

For as long as these rumours persist, I personally think it would be worth the programmers doing something to keep count automatically, and periodically publish some statistics about the randomness of deals, which could include onside honours.

[hrothgar]

It is a common enough accusation (and for that reason if no other I suspect shared by a significant silent minority) that if it can be put to bed then some effort to that end is worthwhile.

 

Provided, of course, that someone is prepared to put in an equal effort into testing the hypothesis by their peers as is put into promoting it. Which is no small task to ask someone freely to undertake just for the price of some goodwill.

 

For all that he seems to have some strange notions, tx10s has made the bravest attempt yet at supporting his allegation.

 

For as long as these rumours persist, I personally think it would be worth the programmers doing something to keep count automatically, and periodically publish some statistics about the randomness of deals, which could include onside honours.

 

Here's a rub...

 

Every whackjob has their own quack theory about how the hand records are biased. tx10s thinks that the finesses are being manipulated, however, we've also seen accusations that there are too many extreme shapes, that there are too many voids, that the trump suit doesn't split right, yada yada yada.

 

Trying to prove that the dealing program is unbiased with respect to all of these diverse claims takes an enormous amount of work.

Sadly, I'm not aware of any general purpose routine designed to test that bridge deals are unbiased.

There are some tests like diehard design to test the quality of a random number generation, but testing bridge hands is would require a lot of special cases.

 

(I suppose, in theory, one could use Thomas Andrews library of hands and see whether the dealer was drawing from them in a uniform fashion, however, given the accusation that BBO is deliberately skewing the deals this probably wouldn't be a suitable test)

[1eyedjack]

Of passing academic interest to me is whether there are genuine psychological reasons why an individual might form a perception of bias that departs (significantly) from reality (regardless of whether or not reality conformed with statistically random predictions).

 

Put simply, is the brain hard-wired to remember a failing success more than a successful finesse?

 

Of possible relevance, I can well imagine that a finesse taken (whether it worked or not) might be more memorable than a finesse avoided in favour of some more promising play. That in itself would not be very helpful, unless there were a non-random relationship between the success rate of finesses where forced into the position of taking them contrasted with the alternative.

 

Among "finesses not taken" I would lump occasions such as

................North

...................xxx

West.........................East

xxxx..........................Kxx

................South

................AQx

 

West leads the suit to the King and Ace. So the finesse was working. Did you mentally "log" it as a finesse? You were effectively a spectator to proceedings.

If it were trick 1, West (at least a GIB West) would be more likely to lead the suit absent the King than holding it.

So if he does not lead the suit, and instead you lead low to the Q, you should expect it to lose to the King more frequently than 50%. Would that contribute to, or reinforce, a preconceived perception of bias?

[hrothgar]

Of passing academic interest to me is whether there are genuine psychological reasons why an individual might form a perception of bias that departs (significantly) from reality (regardless of whether or not reality conformed with statistically random predictions).

 

Our brains are hardwired to find patterns.

We're so good at it that we frequently find patterns that don't actually exist.

 

Equally significant, people are very poor at groking random noise.

People perceive random data as having patterns and think that structured information is random.

[barmar]

We're also psychologically biased to remember negatives more strongly than positives. This has obvious survival benefits: it's important to remember dangerous things so you can avoid them, and the creatures that weren't good at this got eaten more and didn't pass on their genes. This has implications everywhere from unconscious pessimistic biases to PTSD.

 

FYI, a few weeks ago Hans van Staveren, the author of Big Deal, performed a statistical analysis of our deals. He was just looking at distributions, not locations of high cards, but he concluded that our dealing is OK.

[Bbradley62]

We're also psychologically biased to remember negatives more strongly than positives.

I haven't played an ACBL tournament in almost 20 years, but the hands that I distinctly remember are almost all good scores.

[1eyedjack]

Also, no disrespect to posters in this thread, and speaking in generalities, it s my observation that some players seek to explain away their shortcomings in skill at the game by concluding that the game is rigged.

 

You don't tend to get top class or successful players making this complaint. Can it be that they have noticed that it is rigged and used their skill to turn that knowledge to their advantage, keeping silent lest competitors catch on? That is a bit far-fetched.

[helene_t]

We're also psychologically biased to remember negatives more strongly than positives.

Yes, if you train rats by giving them electric shock for doing the wrong thing, the memory tends to stick better than if you give them rewards for doing the right thing.

 

It may also be that fineses actually tend to lose when playing against good defenders because if the finese was working the defenders might have tried to prevent you from taking the finese, either by removing an entry to the table or by offering you an alternative. One of the great problems in Marten's "Virtual Europen Championship" was a hand where you had to play for a singleton king offside because you know the finese doesn't work because the defenders didn't remove your entry to the table.

[barmar]

It may also be that fineses actually tend to lose when playing against good defenders because if the finese was working the defenders might have tried to prevent you from taking the finese, either by removing an entry to the table or by offering you an alternative. One of the great problems in Marten's "Virtual Europen Championship" was a hand where you had to play for a singleton king offside because you know the finese doesn't work because the defenders didn't remove your entry to the table.

The defenders are robots. I don't think anyone would consider them particularly great defenders.

[Bbradley62]

Robots drop meaningful cards that they shouldn't drop, but they are good at cutting off your communication when appropriate.

Edit: But, I don't know whether they're sneaky enough to leave your entry to allow you to take the losing finesse.

[fred]

Robots drop meaningful cards that they shouldn't drop

Within the last few days I saw a robot be very "smart" about dropping a meaningful card. Declarer had the following suit combination at notrump:

 

32

 

AK1087

 

Rightly or wrongly, declarer started by cashing a top honor. LHGIB dropped the Queen.

 

Now it seemed best for declarer to cross to the dummy and finesse the 8, but LHGIB had played the Queen from Queen-9 doubleton!

 

I suspect the robot didn't "know" what it was doing, but it was still a pretty play (as would be the play of the 9 from 9x).

 

This is not meant to prove anything. I just thought it was an interested story and BBradley's post reminded me of it.

 

Fred Gitelman

Bridge Base Inc.

www.bridgebase.com

[Bbradley62]

Yes, GIB makes some Rueful Rabbit plays, but (for the benefit of those who don't read the GIB forum) recently we've been admiring this beautiful unblock:

[hv=bbo=y&lin=pn|vito,~~M35546,~~M35544,~~M35545|st%7C%7Cmd%7C1S69H8D23QKC45JQKA%2CS7JH679KD49TJC289%2CS28KH45TJQD578AC7%2C%7Crh%7C%7Cah%7CBoard%2011%7Csv%7Co%7Cmb%7C1C%7Can%7CMinor%20suit%20opening%20--%203%2B%20%21C%3B%2011-21%20HCP%3B%20%7Cmb%7Cp%7Cmb%7C1H%7Can%7COne%20over%20one%20--%204%2B%20%21H%3B%206%2B%20total%20points%20%7Cmb%7C1S%7Can%7C5%2B%20%21S%3B%207%2B%20HCP%3B%208-18%20total%20points%20%7Cmb%7C2C%7Can%7COpener%20rebids%20his%20C%20--%203-%20%21H%3B%203-%20%21S%3B%2011-%7Cmb%7Cp%7Cmb%7C2D%7Can%7CNew%20suit%20--%203%2B%20%21D%3B%204%2B%20%21H%3B%2012%2B%20total%20points%3B%20forcing%20to%203N%20%7Cmb%7C2S%7Can%7C6%2B%20%21S%3B%2011-18%20total%20points%20%7Cmb%7C3S%7Can%7C3-%20%21H%3B%203-%20%21S%3B%2012-15%20HCP%3B%20twice%20rebiddabl%7Cmb%7Cp%7Cmb%7C4D%7Can%7C5%2B%20%21D%3B%205%2B%20%21H%3B%2012%2B%20total%20points%20%7Cmb%7Cp%7Cmb%7C5D%7Can%7C3%2B%20%21D%3B%203-%20%21H%3B%203-%20%21S%3B%2012-15%20HCP%3B%20twice%20re%7Cmb%7Cp%7Cmb%7Cp%7Cmb%7Cp%7Cpc%7CSA%7Cpc%7CS6%7Cpc%7CS7%7Cpc%7CS2%7Cpc%7CHA%7Cpc%7CH8%7Cpc%7CH9%7Cpc%7CH4%7Cpc%7CSQ%7Cpc%7CS9%7Cpc%7CSJ%7Cpc%7CSK%7Cpc%7CD5%7Cpc%7CD6%7Cpc%7CDK%7Cpc%7CDJ%7Cpc%7CD2%7Cpc%7CD9%7Cpc%7CDA%7Cpc%7CST%7Cpc%7CC7%7Cpc%7CC3%7Cpc%7CCA%7Cpc%7CC2%7Cpc%7CCK%7Cpc%7CC8%7Cpc%7CS8%7Cpc%7CCT%7Cpc%7CCQ%7Cpc%7CC9%7Cpc%7CH5%7Cpc%7CC6%7Cpc%7CDQ%7Cpc%7CDT%7Cpc%7CD7%7Cpc%7CS5%7Cpc%7CCJ%7Cpc%7CH7%7Cpc%7CHT%7Cpc%7CH2%7Cpc%7CC5%7Cpc%7CH6%7Cpc%7CHJ%7Cpc%7CS4%7Cpc%7CC4%7Cpc%7CD4%7Cpc%7CD8%7Cpc%7CH3%7Cpc%7CHQ%7Cpc%7CS3%7Cpc%7CD3%7Cpc%7CHK%7C]360|270[/hv]

[mgoetze]

We're also psychologically biased to remember negatives more strongly than positives.

 

Yes, if you train rats by giving them electric shock for doing the wrong thing, the memory tends to stick better than if you give them rewards for doing the right thing.

On the other hand, parents clearly tend to remember all the good things about being a parent and forget about all the minor annoyances.

[barmar]

On the other hand, parents clearly tend to remember all the good things about being a parent and forget about all the minor annoyances.

If our ancestors allowed the pains of childbirth and child rearing to bother them too much, the species would have stopped reproducing and we wouldn't be here.

[ArtK78]

If our ancestors allowed the pains of childbirth and child rearing to bother them too much, the species would have stopped reproducing and we wouldn't be here.

Fortunately, the joys of sex more than counter the pains of childbirth and child rearing. The species was never in much danger.

[tx10s]

Of passing academic interest to me is whether there are genuine psychological reasons why an individual might form a perception of bias that departs (significantly) from reality (regardless of whether or not reality conformed with statistically random predictions).

 

Put simply, is the brain hard-wired to remember a failing success more than a successful finesse?

 

Of possible relevance, I can well imagine that a finesse taken (whether it worked or not) might be more memorable than a finesse avoided in favour of some more promising play. That in itself would not be very helpful, unless there were a non-random relationship between the success rate of finesses where forced into the position of taking them contrasted with the alternative.

 

Among "finesses not taken" I would lump occasions such as

................North

...................xxx

West.........................East

xxxx..........................Kxx

................South

................AQx

 

West leads the suit to the King and Ace. So the finesse was working. Did you mentally "log" it as a finesse? You were effectively a spectator to proceedings.

If it were trick 1, West (at least a GIB West) would be more likely to lead the suit absent the King than holding it.

So if he does not lead the suit, and instead you lead low to the Q, you should expect it to lose to the King more frequently than 50%. Would that contribute to, or reinforce, a preconceived perception of bias?

Did you even read my post? I looked at potential finesses, not ones actually taken. Please dispute the data rather then play psychologist.

[tx10s]

It took 5 days for me to receive the message that the files I sent last Saturday were too bid for the rt@tr.bridgebase.com mailbox. I tried sending one of the files to the support@bridgebase.com site, but I am no holding my breath with expectations.

[tx10s]

Also, no disrespect to posters in this thread, and speaking in generalities, it s my observation that some players seek to explain away their shortcomings in skill at the game by concluding that the game is rigged.

 

You don't tend to get top class or successful players making this complaint. Can it be that they have noticed that it is rigged and used their skill to turn that knowledge to their advantage, keeping silent lest competitors catch on? That is a bit far-fetched.

I scored in 15 of the 20 tournaments that I posted about. That my be a shortcoming to an expert like you, but I was quite pleased with those results.

[1eyedjack]

Don't take it so personal. My interest in the psycho aspect has nothing to do with you.

[tx10s]

Don't take it so personal. My interest in the psycho aspect has nothing to do with you.

Then I apologize, sorry for my misinterpretation.

[Bbradley62]

I don't know whether they're sneaky enough to leave your entry to allow you to take the losing finesse.

Basic GIB just did this to me! After cashing AK of his suit, he gave me a ruff/sluff, which allowed me to get to the board when I couldn't otherwise. Having started with xxx opposite AKJxx of trump, I then lost a finesse to the offside Qx. Next time, I'll know better!

[barmar]

On the other hand, don't forget all the times that GIB leads his kingleton, often into declarer's AQ. Left to his own devices, declarer probably would have lost that finesse (most of us are not rabbis).

 

This comes from GIB's reliance on double dummy simulations for defense. Since it uses DD analysis, it assumes declarer will drop the king if possible instead of finessing, so it's not giving away anything.