[jcalex2]

I understand there are two types of robots we can buy, basic versus advanced. What type of robot is used in the robot duplicate tournaments?

Additionally, why do the robots bid so badly? I have seen novice bridge players with less than one week experience bid better...

[hrothgar]

 jcalex2, on 2020-June-17, 14:54, said:

Additionally, why do the robots bid so badly?

Sunk cost fallacy

[jcalex2]

 hrothgar, on 2020-June-17, 15:57, said:

Sunk cost fallacy

Sorry not an economist, you are saying, cost for previous robot development already incurred, cost for future robot improvement/development provides no additional profit?

[pilowsky]

You cannot buy robots. Slavery was abolished by Abraham Lincoln. The final decree was on June 19, 1865, I'm told. You can rent their services. I would opt for the East robot. On a good day, his bidding and declarer play are both excellent. I would be very grateful if you could put me in touch with the Novices that you speak of. I've been playing for more than one week and I take lessons yet even the North robot can beat me.
Please provide some examples of so-called "bad bidding". In my experience, GIB is extremely accurate and reasonable.

[johnu]

 pilowsky, on 2020-June-18, 04:03, said:

Please provide some examples of so-called "bad bidding". In my experience, GIB is extremely accurate and reasonable.

LOL

[pilowsky]

Hilarious. I suggest you practice the GIB system

[johnu]

 pilowsky, on 2020-June-18, 16:51, said:

Hilarious. I suggest you practice the GIB system

There a saying that even a blind possum finds a nut once in a while. RTFF(orum) and even you might find something that GIB does that doesn't seem right to you

[sfi]

 jcalex2, on 2020-June-17, 22:02, said:

Sorry not an economist, you are saying, cost for previous robot development already incurred, cost for future robot improvement/development provides no additional profit?

My interpretation is Hrothgar is saying that BBO has invested a lot of time and money in the robots, and doesn't want to throw that away by replacing them with something else (either starting from scratch or with another product). His statement suggests that future investment required with these robots may be more than if they did start over, and that they are making an incorrect economic assessment in sticking with the current ones.

[pilowsky]

Quote

"The BBO robot is poorly designed. Its bidding is terrible. First, it does not follow SAYC conventions so right off the bat, most people are confused. Then it is designed for esoteric conventions which are screwy to say the least. The stupid robot competes with its partner rather than work with its partner. It will keep bidding even with an inferior hand until it reaches a contract in its suit. I come to BBO because of its excellent graphics but I am fed up."

OK, perhaps I am not being clear, let's try again.

How many times do I see commentaries similar to this one on the Forum and elsewhere? You may be interested to know that in a review published by Rahwan et al., in Nature (2019) It was proposed in the title that there is such a thing as "Machine behaviour". Well, I have some bad news for Dr Rahwan and his friends at MIT. They are wrong, there isn't. You can call it Ginsberg Intelligent BridgePlayer or GIB, but it isn't the algorithm that is intelligent it's the human. This fallacy is known as anthropomorphisation: assigning human qualities to non-human things. It is commonly done as a means of satire. I do it frequently on the Forum when I mock my own poor play by ascribing my inadequacies to the incompetence of my friend the North robot (see what I mean…).

The Rahwan review starts with the following sentence.

"In his landmark 1969 book, Sciences of the Artificial Nobel Laureate Herbert Simon wrote: Natural science is knowledge about natural objects and phenomena." Before I continue, pause to understand the use of the words 'landmark' and 'Nobel Laureate' here: The idea behind this kind of writing is How can you argue with me now? After all my argument is backed by a Nobel laureate. The authors then go on to develop the idea that if Humans have a quality that we refer to as 'intelligence' then so do machines.

You might think "well, it's published in Nature, the worlds best scientific Journal, it must be true". Think of the still unretracted paper by Benveniste 1988 showing that water can 'remember' the presence of antibodies. In 2019 alone 14 papers were retracted from another Nature group publication "Scientific Reports" which has an impact factor of more than 4.

All sorts of ridiculous ideas flow from this concept of artificial intelligence: when in fact it is nothing more than a perpetual motion machine. Pit four robots against each other and guess who wins? Nobody - that's right nobody is playing.

In fact, GIB is not artificial intelligence at all. Only people that believe in Santa Claus, the Tooth Fairy, or have not passed the stage of object permanence could think otherwise. Here I am thinking of people that believe that COVID-19 will magically go away if you stop testing, or that your mother vanishes when she leaves the room.

Here is how it really works:

When you play in a robot tournament it is exactly the same as playing IRL (in real life) or FTF (face to face). Except that it is better. There is no possibility of cheating. Your opponents cannot kick each other under the table, blink, wink, fart, move their fingers, or do anything else. Even better, they make a pretty good attempt at describing their bid. They do not arrive at the table and say things like we play Acol. Or we play Myxi 2's or some other stuff. They do not behave like a lesion looking for a space to occupy. They never gloat and they never complain.

The most important thing that people seem to be unaware of - in the Kruger/Dunning sense of the term - is that when you play in a game against robots, the result that you get is not a result that is compared with robots.

You are comparing yourself with actual people. Yes, PEOPLE.

So, next time somebody complains about bad bidding or bad play from the robots remember that what they are really saying is that someone (or more likely several other someones) found a better way to bid and play the same hand. Alternatively, the advanced robots played it differently to the basic robots, or two humans played it differently, but again, is this any different to what happens IRL or FTF? It still has nothing to do with 'intelligence' on the part of the robot.

In order for GIB to be intelligent, it would need to be capable of several things that humans can do that the program cannot. It would need to be able to learn new stuff. It can't. It would need to be able to improvise. It doesn't. It would need to be able to write long whiny aposiopesical posts on the Forum about...

[smerriman]

 sfi, on 2020-June-18, 22:32, said:

My interpretation is Hrothgar is saying that BBO has invested a lot of time and money in the robots, and doesn't want to throw that away by replacing them with something else (either starting from scratch or with another product). His statement suggests that future investment required with these robots may be more than if they did start over, and that they are making an incorrect economic assessment in sticking with the current ones.

I think it's the other way around, to be honest. Making changes to GIB appears so complicated that they can't afford to spend any time on it, which is why nobody works on it anymore. And that there probably is a long term plan to switch to Argine, in which case there is no reason at all to spend anything more on GIB.

[zhasbeen]

 pilowsky, on 2020-June-18, 16:51, said:

Hilarious. I suggest you practice the GIB system

Not that GIB doesn't make some good calls at times, probably more than some us give it credit for. However, it is poor overall, often making some of the worst calls you'll ever see. BBO sorely needs a new robot, or find a way to fix Gib, which apparently is impossible.

[pilowsky]

 zhasbeen, on 2020-June-19, 09:34, said:

Not that GIB doesn't make some good calls at times, probably more than some us give it credit for. However, it is poor overall, often making some of the worst calls you'll ever see. BBO sorely needs a new robot, or find a way to fix Gib, which apparently is impossible.

Try to follow the reasoning presented above. You are not actually competing against the computer. You are competing against another human who is competing against the computer. There are not actually three robots and one human, there is one human - you, and one computer GIB! What is so difficult to understand. It doesn't matter what GIB bids. It only matters what you bid compared to what the other human bids That's why it's called human intelligence! Otherwise, you would use it to heat bread and make toast.
FYI the original 'bug' was an actual moth on the circuit board of a mainframe - how about that. from JCR Licklider and the men who made computing personal.

[Stephen Tu]

All of us understand that the playing field vs other humans is level, in that the robots bid equally atrociously with everyone. What you don't seem to understand is that the multitude of bugs significantly decrease enjoyment of the contest for most of us. Because having a bad partner limits a good player's ability to win by making better bids, bids that would work well with a good human partner, but are a crapshoot with the current bot. You make a good judgment to sacrifice over the opp's game, with a good partner you would win IMPs/MPs. But the robot fails to take the normal preference of the longest trump suit, and thus your sacrifice goes for too much when playing in the wrong strain. You judge well to bid to 5m, which would be a top, but the computer raises to 6 for no good reason, which becomes a bottom. You make a splinter bid which should lead to a good slam, but robot partner can't evaluate and does something ridiculous, you no longer gain against people who don't know about splinters. You want to make a penalty double which would get a good board if left in, but are afraid to make it because you think the bot will do something stupid. Too many bugs means what are normally good bids (when playing with a good adv/exp human) get randomly punished instead of rewarded as they should be.


Imagine a hypothetical, say the bots had a bug such that every single hand, your robot partner bid 7nt and redoubled. Now, this would be a level contest for everyone, sure, but it wouldn't be very enjoyable, and doesn't resemble at all bridge played by normal humans. Now GIB obviously isn't this bad, and is actually better than the average human on the site, but it is substantially worse than more advanced humans and other bots that haven't been abandoned by their original creators for 18+ years and have been continually improved at a faster rate. The question BBO has to ask is what the ROI is in investing money into licensing a different bot (and altering it to function in their distributed server environment), or increasing resources into GIB. It's not clear, because there's not much competition with other sites, and people generally aren't quitting playing robot tourneys out of disgust. They are cheap, and the robots are "good enough", for most, in that on most hands it generally takes reasonable actions. It's only maybe 1/10 of the time it does something egregious, it's just frustrating because some of the bugs are well known, repetitive, and have been there for years without being addressed. BBO is basically doing the McDonald's model here, cheap, good enough for average customer, no real incentive to improve without significant competition or sales plummeting.


If you want examples of bugs, read through hundreds (thousands?) of bug reports on this forum below, many of which remain unfixed, many for years.

[Vampyr]

The sunk cost fallacy is when you keep a product because you invested time, money and other resources in it, You don’t want those resources “wasted” by scrapping the project. This is a fallacy because those resources are gone (and are already wasted if the product is no good) and you are never getting them back, so it might be time to move on and use your new/ongoing resources to better effect. I am not sure what the BBO employees are doing with themselves, since every time a request is made to improve playability or functionality, or fix one of the bot’s many bugs, they say either “we can’t do that” or “we don’t want to do that”. There does seem to be some realisation that improving the robots is sending good money after bad, so maybe they are working on transitioning to this other robot that someone mentioned. Or maybe they will do nothing. The influx of lockdown players don’t play with robots, so if BBO’s traditional customer base consists of weaker players who are even worse than the robots, why invest in a robot that good players will want to play with/against?

[pilowsky]

Unfortunately, Stephen, your arguments are not supported by facts. They are sweeping generalisations. You claim that "all of us understand" and then in the same breath say that "I don't understand"! Immediately afterwards you launch into what seems to be a clear demonstration of what is known as "external locus of control". You can read about it here (https://en.wikipedia...ocus_of_control). The computer is just a computer. It doesn't care about what you want. Neither does BBO. If you do not enjoy it then go elsewhere. I do enjoy it. I find it to be a stimulating mental exercise that is unsullied by distractions that allows me to meet people from all over the world. It is far and away superior to anything I can experience in the Club in Sydney.
It is nothing to do with the computer. If you want to play face to face bridge then just get three other people and off you go. I do this on Stepbridge and other computer platforms. I will not be doing it at the face to face "Club" because it does not cater for all people, and because it is expensive and because I have seen and experienced a large amount of bullying racism antisemitism and misogyny.
As for so-called errors, I have seen the computer make choices that I would prefer it not to make. These choices are never better or worse than choices that a human partner would make based on bids that I have made on the trust-your-partner principle. What many (not all) people do not seem to understand is that the computer only plays one system. (like the terminator) It never deviates. It does not understand irony and if you indicate that you have a certain holding then it will simulate based on that estimation. Sometimes this will result in a contract that humans characterise as "ridiculous" - bad luck - no-ones dying.

[Vampyr]

Just read a few of the bug reports

Anyway, we get it. You think the robot plays well, and no one else here agrees.To be perfectly honest, I would rather have a robot as a partner then a random unknown (to me) player. I choose to do neither. I can always watch an opera instead.

[Stephen Tu]

 pilowsky, on 2020-June-19, 19:25, said:

Unfortunately, Stephen, your arguments are not supported by facts. They are sweeping generalisations. You claim that "all of us understand" and then in the same breath say that "I don't understand"!

Those statements were referring to different concepts. I was saying that we understand your argument, that it is a level playing field and that the competition is against other humans. What I am claiming what you appear not to understand, from your statements, is that a significant number of us would enjoy it a lot more (and perhaps play more often) if the substantial number of bugs in the software were reduced, either by increased investment in bug-fixing development hours, or by moving to another software entirely.

Quote

Immediately afterwards you launch into what seems to be a clear demonstration of what is known as "external locus of control". You can read about it here (https://en.wikipedia...ocus_of_control).

Wtf nonsense is this? I know who controls the computer. I am under no delusions. I worked as a software engineer for two decades, I know how they operate. The computer executes code based on human composed instructions. Unfortunately, humans are far from perfect, and can provide the computer faulty instructions, and incomplete instructions (i.e. database entries/rules telling how the computer to interpret bids, and rules on how to prioritize choosing particular bids). BBO's software has had minimal development other than fixing some of the most egregious bidding database bugs/gaps for about 18 years. It has fallen significantly behind other competitors in terms of the bot play.

Quote

The computer is just a computer. It doesn't care about what you want. Neither does BBO.

Of course I know a computer doesn't "care", I don't expect it to. But BBO should care about customer desires for improved software. Now maybe they don't think the ROI is there to justify spending the money to improve the bots (as opposed to other areas of their operation, like their recent need to increase general capacity to handle an increased number of players due to COVID-19). But if they choose to not improve their bots, or continue their current snail's pace of fixes, some of the customers are going to post complaints about it, and I don't think it's really your place to say that we shouldn't complain.

Quote

As for so-called errors, I have seen the computer make choices that I would prefer it not to make. These choices are never better or worse than choices that a human partner would make based on bids that I have made on the trust-your-partner principle. What many (not all) people do not seem to understand is that the computer only plays one system. (like the terminator) It never deviates. It does not understand irony and if you indicate that you have a certain holding then it will simulate based on that estimation. Sometimes this will result in a contract that humans characterise as "ridiculous" - bad luck - no-ones dying.

We understand this. We also understand that software can be improved if the programmers work on it. People have the right to complain about software bugs. You think Microsoft/Google/AWS/every other commercial software operation don't get complaints when stuff doesn't work or doesn't work as well as it could? You think people should just ignore any and all bugs and never ask for fixes, just live with it? Your position is insane IMO.

[pilowsky]

And there's the problem. You say "I know who controls the computer". I have worked with computers since I was programming in Fortran in 1976. I have used a slide rule to calculate politicians broken promises and publish the result in National newspapers at the age of 14. Everyone here is pretty clever. But if you really think that someone is actually "controlling the computer..." then I think it's time to look up external locus of control. Entirely up to you.

[Stephen Tu]

A computer program doesn't control what sequence of instructions a computer executes? It doesn't control what a bridge playing robot given a particular hand, position, previous bidding etc. chooses to bid on its turn? A programmer or team of programmers that maintain/develop/fix that program doesn't control how it behaves? We as end users can't request those programmers to alter the software to not make the same bidding mistakes in certain situations?

WTF are you talking about, external locus of control? You are making a total nonsense argument.

[zhasbeen]

 pilowsky, on 2020-June-19, 14:58, said:

Try to follow the reasoning presented above. You are not actually competing against the computer. You are competing against another human who is competing against the computer. There are not actually three robots and one human, there is one human - you, and one computer GIB! What is so difficult to understand. It doesn't matter what GIB bids. It only matters what you bid compared to what the other human bids That's why it's called human intelligence! Otherwise, you would use it to heat bread and make toast.
FYI the original 'bug' was an actual moth on the circuit board of a mainframe - how about that. from JCR Licklider and the men who made computing personal.

"robots and one human, there is one human - you, and one computer GIB! What is so difficult to understand."

Yes, this absolutely true, and I have stated this to someone I know who is interested as recently as yesterday. In my own words:
"We all have same robot partner and 2 robot opponents, while playing the same convention card. We don't have to alert or remember what all the bids mean. All we have to do is run our mouse cursor over any bid in the auction to see what the robots' bids mean, or what the robots expect us to hold for our bids."
I go on to say: "While you are trying to outplay the robots while at the "table", your real opponents are the other humans that held your cards. At the end of the game your scores are match pointed and listed in the results.
It all comes down to playing your cards and making decisions based on the info that's available to you and all the other players."

So, I agree that Gib's antics probably don't have much impact on matchpoint scores. "It's the same for everyone", as the saying goes. However, it can still have a negative impact on the enjoyment that many players would experience otherwise, and sometimes it borders on painful. There is so much else I like about these tournaments, that I and many others put up with Gib. If only BBO would do something about this big elephant in the living room!

And while I'm at it, for the past 2 years I've kept complete results of every ACBL Daylong Individual tournament at matchpoints since January 1, 2019. At the very least I can tell you number of games, average matchpoint score, and how any 2 players have fared relative to each other when in the same field. You are in there!
I also have more detailed info such as masterpoints, power ratings, etc for those who include real name in profile and have average scores ranking in the upper 20%. I include my real name in player profile.

So, with all the being said, you or anyone else who reads this can drop me a private email if your are curious, and I can tell you number of tournaments played, average score, and how your record compares with any other player(s) when you've been in same field. Pilowsky; are you curious?