[nathan2008]

hrothgar, on 2012-April-02, 06:39, said:

I came in second in the last BBO Forums tournament playing with GIB as a partner

Now I usually play Robot Reward, Robot Duplicate, ACBL speedball, ACBL robot duplicate. I find BBO robots are very good. At first, they are always trying to improve GIB version. Everyone wants to make things better, so why do we still criticize on it? Secondly, many players think they are so "good", (actullay they are not.) After their stupid mistakes, they also want to argue with me to find stupid reasons. This will make me mad lol. At least robot will always keep silent. I only play with the people who don't argue too much with me now lol. Finally there are a lot of rude players on BBO. I am tired of that. Like our famous <removed>, really stupid. Never will I play with/against him in my whole life. I will definitely prefer playing with robots than him. Until now i don't understand why people play with stupid <removed>.

This post has been edited by barmar: 2012-April-02, 09:26
Reason for edit: Removed personal attack

[barmar]

Vampyr, on 2012-April-02, 05:48, said:

It could, though, couldn't it?

Anything is possible, if we're willing to redesign GIB. Your suggestion is more than just a minor tweak, so we're less likely to attempt it.

Quote

But perhaps knowledge that only, say, 20% of the 1NT opening bids are in the stated range and distribution will prevent GIB from being too wedded to the description.

That seems to be a pretty extreme supposition. Unless there are lots of players bidding like this, it doesn't seem like there's much value in redesigning GIB to learn about them.

It seems like you want us to redesign GIB just because Leo has learned how to get good results against them. Instead of having the computer make things harder for him, why don't you just learn from him and beat him at his own game? If he can get good results by opening out-of-range 1NT, so can you.

[pigpenz]

barmar, on 2012-April-02, 08:35, said:

It seems like you want us to redesign GIB just because Leo has learned how to get good results against them. Instead of having the computer make things harder for him, why don't you just learn from him and beat him at his own game? If he can get good results by opening out-of-range 1NT, so can you.

yes, and some have learned alot from that...but you still have to have good declarer play.
I dont even think some players spend that much time looking at other results, the only way to find out
if its a problem is if people complain about bidding or results.

[mgoetze]

Vampyr, on 2012-April-02, 05:59, said:

Partnership agreements may be implicit (40B1b)and may be based on partnership experience (40B6a). What you and others seem to be saying is that when playing with robots, the partnership's "agreements" are a fiction, and the player is under no obligation to actually abide by them most of the time. This cannot be correct, and cannot be bridge.

Really, the robot has no idea that you are doing anything other than what is on its card. It has no partnership experience at all, when you sit down with GIB it's as if you're playing the first hand together every single hand.

If you played every single hand of your bridge career with a new partner it would be legal to psyche on every single one of them. In reality, humans are not interested in sitting down with you for only one hand.

[barmar]

If we were able to do something like this, I assume you would consider it fair if we also taught partner-GIB how to look in the database of past actions, so it could adjust its bidding accordingly -- e.g. if you frequently open 1NT with 13HCP, it won't always raise to 3 with 10.

But really, this is a very hard AI problem, called "classification".

http://en.wikipedia...._classification

You don't want to keep the records in too much detail, because then each one will have a very small percentage: 13 HCP 7222 1%, 14 HCP 7222 1%, 13 HCP 4443 1%, etc. This fragmentation will make it very difficult for a sufficient number of hands to show up in the simulations. To solve this, you need to group related hand types together -- e.g. all 13-14 HCP hands, all 5431 hands, etc. This is a difficult problem in general, although maybe the limited context of bridge bidding simplifies it a bit -- we might be able to give it hints about how to group things.

[Vampyr]

barmar, on 2012-April-02, 08:35, said:

Anything is possible, if we're willing to redesign GIB. Your suggestion is more than just a minor tweak, so we're less likely to attempt it.

That seems be a pretty extreme supposition. Unless there are lots of players bidding like this, it doesn't seem like there's much value in redesigning GIB to learn about them.

Maybe it would make sense for GIB to keep some statistics to find out just how widespread the phenomenon is.

Quote

It seems like you want us to redesign GIB just because Leo has learned how to get good results against them. Instead of having the computer make things harder for him, why don't you just learn from him and beat him at his own game? If he can get good results by opening out-of-range 1NT, so can you.

I'm not interested in playing robot tournaments, though I'm sure there are some people trying to play their normal game whose results are negatively affected; still, I don't think that who exactly wins these tournaments is particularly relevant.

What matters to me is that the ACBL are awarding masterpoints for a game that breaks several laws plus its own regulations. It's easy to say "who cares?", but a lot of people must be interested in these points, as the ACBL (fee-charging) Robot Games are more popular than similar free games. Recently a poster expressed dismay at the fact that the points that he spent years travelling around to tournaments to accumulate can now be had by sitting in front of the computer for a few months. Hell, probably many people don't even bother with that -- they have programmed their own bots to play with the GIBs and earn their masterpoints.

So where is the incentive to earn points "the old-fashioned way"? And should I care? Well, I do. For me, bridge is a card game played by four people sitting at a table, preferably surrounded by a lot of other tables with four people at each also playing cards. If nothin would be very sad if the real game lost out to the online form, let alone to an online form that didn't involve any actual human interactions. And while I don't play in the ACBL myself, I fear that trends set in the NBO with the largest and most numerous tournaments will eventually be followed elsewhere.

Anyway, so wishing that the ACBL would award its masterpoints to a game that was a reasonable facsimile of the one played with people and cards is a very modest wish for me, and would perhaps only satisfy a principle. I do think, somehow, that it is important.

If nothing else, it would make it a lot easier for robot players to transition into live bridge if it were roughly the same game.

[Vampyr]

mgoetze, on 2012-April-02, 09:56, said:

Really, the robot has no idea that you are doing anything other than what is on its card. It has no partnership experience at all, when you sit down with GIB it's as if you're playing the first hand together every single hand.

If you played every single hand of your bridge career with a new partner it would be legal to psyche on every single one of them. In reality, humans are not interested in sitting down with you for only one hand.

Is playing every hand as if it were with a brand-new partner a reasonable approach to a partnership game?

barmar, on 2012-April-02, 11:23, said:

If we were able to do something like this, I assume you would consider it fair if we also taught partner-GIB how to look in the database of past actions, so it could adjust its bidding accordingly -- e.g. if you frequently open 1NT with 13HCP, it won't always raise to 3 with 10.

Maybe. But just as in real life, it is not considered as serious a problem when you mislead partner.

Quote

But really, this is a very hard AI problem, called "classification".

http://en.wikipedia...._classification

You don't want to keep the records in too much detail, because then each one will have a very small percentage: 13 HCP 7222 1%, 14 HCP 7222 1%, 13 HCP 4443 1%, etc. This fragmentation will make it very difficult for a sufficient number of hands to show up in the simulations. To solve this, you need to group related hand types together -- e.g. all 13-14 HCP hands, all 5431 hands, etc. This is a difficult problem in general, although maybe the limited context of bridge bidding simplifies it a bit -- we might be able to give it hints about how to group things.

I guess all this depends on how desirable it is considered for robot games to be "like bridge". I don't suppose BBO or the ACBL have much of an incentive to make any changes in their cash cow.

[barmar]

Vampyr, on 2012-April-02, 11:49, said:

guess all this depends on how desirable it is considered for robot games to be "like bridge". I don't suppose BBO or the ACBL have much of an incentive to make any changes in their cash cow.

it doesn't matter how desirable it is, if we don't have the AI technology to do it.

[Vampyr]

barmar, on 2012-April-02, 12:02, said:

it doesn't matter how desirable it is, if we don't have the AI technology to do it.

Right. Sorry.

[fred]

Enough is enough, Vampyr. In a previous thread you claimed not to have an agenda, but you have now made it very clear that you are not a fan of online bridge especially online bridge played against robots (even though you have never played in and apparently refuse to play in the specific types of games that you somehow think you are qualified to criticize).

Here is some news for you: this is an online bridge site and you are a guest here.

What you rudely refer to as our "cash cow" in no small way helps to finance this operation (and pays the salaries of dozens of hard-working bridge players, and allows us to keep vugraph free, and allows hundreds of thousands of BBO members to enjoy our site without ever giving us any money, and allows us to contribute in various ways to getting more young people interested in bridge, and...).

Meanwhile you continuously exaggerate and offer misleading non-facts in your posts. I could provide several examples, but this vicious little piece of ignorance wins the prize:

Quote

Hell, probably many people don't even bother with that -- they have programmed their own bots to play with the GIBs and earn their masterpoints.

So I will now ask you to please stop posting on these matters on this site. If you insist on continuing your rant then please take it elsewhere. As far as posting on these matters are concerned, you are no longer welcome here.

Fred Gitelman
Bridge Base Inc.
www.bridgebase.com

[barmar]

I wish he'd made the comment about people programming their own bots a few days ago. Yesterday would have been a great day to announce that Leo LaSota is actually a bot, and that's how he's been able to play so much online bridge.

[Vampyr]

fred,

Obviously it's your way or the highway, but I would like to point out that not being a fan of something is not at all the same as having an "agenda". Gosh, if it were I'd be positively swamped with all the agendae(?) I have. I don't mind disclosing, in this space, that I am not a fan of Brussels sprouts, Peter Gabriel, Chelsea tractors, tofu, criminals, tanning beds, Newt Gingrich, football or the Daily Mail. If it helps, I promise not to discuss any of these things on BBO either.

[fuburules3]

Perhaps the admins would prefer this thread just die, but I am wondering if there are any available statistics on what matchpoint percentages the top robot players average (and whether there actually is a gain to strange bidding)? I occasionally have looked at Leo's averages (as a benchmark on how well you can do in these tourneys) and he seems to average a bit over 60% (61% over the last 30 days). I'd be curious to know the matchpoint average of a top expert playing robot tourney's bidding by the book (I believe Joshua Donn suggested on another site that he played a lot of robot tourneys bidding by the book, so perhaps with his permission someone could dig up his stats for one point of comparison).

My own experience playing straightforward bridge is that my personal average (playing mostly the $.25 games as opposed to the possibly more difficult $1 ACBL ones) seems to be somewhere in the 55%-58% range. If we accept these numbers, then Leo has a 3-6% edge over me. Since Leo has won an NABC and I know from my live bridge results that I am far from a top expert, some of this edge surely comes from superior card play, which suggests that any edge gained from "creative" bidding is fairly minimal.

[fred]

fuburules3, on 2012-April-02, 17:10, said:

Perhaps the admins would prefer this thread just die, but I am wondering if there are any available statistics on what matchpoint percentages the top robot players average (and whether there actually is a gain to strange bidding)? I occasionally have looked at Leo's averages (as a benchmark on how well you can do in these tourneys) and he seems to average a bit over 60% (61% over the last 30 days). I'd be curious to know the matchpoint average of a top expert playing robot tourney's bidding by the book (I believe Joshua Donn suggested on another site that he played a lot of robot tourneys bidding by the book, so perhaps with his permission someone could dig up his stats for one point of comparison).

My own experience playing straightforward bridge is that my personal average (playing mostly the $.25 games as opposed to the possibly more difficult $1 ACBL ones) seems to be somewhere in the 55%-58% range. If we accept these numbers, then Leo has a 3-6% edge over me. Since Leo has won an NABC and I know from my live bridge results that I am far from a top expert, some of this edge surely comes from superior card play, which suggests that any edge gained from "creative" bidding is fairly minimal.

This particular admin doesn't especially care if this thread continues (or not) - it does contain some interesting food for thought. My objection is to people using our site to make posts with the apparent intention of damaging our site. I also don't like it when I see posts that are abusive or insulting to members of our staff (not an issue in this thread).

In general we are not afraid of constructive criticism. In fact we welcome it - we have a lot of smart members and their feedback can be very useful in terms of helping us to improve our software and service. But when such criticism comes from a person who has never even tried what he/she is criticizing, is rude, sarcastic, or (intentionally or not) contains misleading or downright false information, then that's another story. In such cases I think that self-defense is more than justified.

As for your request for statistics, sorry but I don't think it is realistic to hope that we will be able to compile them.

For whatever it's worth, Leo himself has repeatedly claimed that he believes his results are at least as good when he plays "down the middle". Several other highly skilled and highly experienced robot players (including me) agree with this and have expressed that they prefer to play "normally" in our robot games.

Fred Gitelman
Bridge Base Inc.
www.bridgebase.com

[Bbradley62]

Vampyr, on 2012-April-02, 11:40, said:

... the ACBL (fee-charging) Robot Games are more popular than similar free games.

I don't see how you're making this comparison. The free games are limited to 10 players, and those 10 spots fill almost immediately when they become available. I'm sure the free games would have 40-60 players each if they were allowed to. (I'm not suggesting they should be allowed to, since this is a business decision based on real money for BBO; just noting the invalidity of the comparison.)

[Vampyr]

Bbradley62, on 2012-April-02, 17:48, said:

I don't see how you're making this comparison. The free games are limited to 10 players, and those 10 spots fill almost immediately when they become available. I'm sure the free games would have 40-60 players each if they were allowed to. (I'm not suggesting they should be allowed to, since this is a business decision based on real money for BBO; just noting the invalidity of the comparison.)

Oh, I didn't realise. I was told in another thread that the ACBL games are much more popular, but not the reason.

[pigpenz]

I always thought it would be a good idea if the people who rent the bots could make free games available just for the other
people who rent the bots.

and whoever is TD ing could have random hands or best hand south

[mgoetze]

Vampyr, on 2012-April-02, 17:56, said:

Oh, I didn't realise. I was told in another thread that the ACBL games are much more popular, but not the reason.

I believe the claim was they are more popular than the non-ACBL pay robot tournaments (which are what I play).

[Bbradley62]

mgoetze, on 2012-April-02, 22:54, said:

I believe the claim was they are more popular than the non-ACBL pay robot tournaments (which are what I play).

That would also be difficult (though not impossible) to measure. I'm sure the size of ACBL games varies largely based on the time in the US... non-ACBL games run more frequently... non-ACBL games have fewer boards and are speedball, which might be contributing factors having nothing to do with ACBL masterpoints.

[nathan2008]

fuburules3, on 2012-April-02, 17:10, said:

Perhaps the admins would prefer this thread just die, but I am wondering if there are any available statistics on what matchpoint percentages the top robot players average (and whether there actually is a gain to strange bidding)? I occasionally have looked at Leo's averages (as a benchmark on how well you can do in these tourneys) and he seems to average a bit over 60% (61% over the last 30 days). I'd be curious to know the matchpoint average of a top expert playing robot tourney's bidding by the book (I believe Joshua Donn suggested on another site that he played a lot of robot tourneys bidding by the book, so perhaps with his permission someone could dig up his stats for one point of comparison).

My own experience playing straightforward bridge is that my personal average (playing mostly the $.25 games as opposed to the possibly more difficult $1 ACBL ones) seems to be somewhere in the 55%-58% range. If we accept these numbers, then Leo has a 3-6% edge over me. Since Leo has won an NABC and I know from my live bridge results that I am far from a top expert, some of this edge surely comes from superior card play, which suggests that any edge gained from "creative" bidding is fairly minimal.

Very easy, arrange JEC match for leo Lasota and his pd. You will see how strong he is in the real game. From my experience, i believe that a player who knows how to play with robots very well is indeed very good. Because in robot game, you also need endplay, squeeze, safe play etc.