[msheald]

Hello! I've had a chance to play more robot bridge as a babysit a project with a lot of down time. I'm an intermediate player.

The game seems different than when I play against other people in BBO.

I'd been scoring over 50% consistently in the robot games, even with hands and robot bids/play that left me scratching my head, though I did not take the game scores too seriously, so I thought it would be fun to play in the 3-day robot national tournament that occurred about the time I anticipated free time during my project.

It seemed that the robots made a silly play about 1 deal in 15. Additionally, finesses seemed to work far less than 50%, and card distributions seemed more unusual than I expected from chance alone.

Robot games seem like they are different games compared to people games - robot games seem to favor the conservative player while people games seem to favor the aggressive player.

Do folks find robot games different than games against people? Or am I out in left field with my interpretation of deals and play in robot games? Best regards.

Mike

[mythdoc]

msheald, on 2021-March-18, 07:18, said:

The game seems different than when I play against other people in BBO..

Mike

Some (like me) also experience this and others say no. Hands have been analyzed for distribution and other measurables and found to be normal.

This is asked commonly on BBO. Here is the most recent iteration of the discussion

https://www.bridgeba..._1#entry1017861

[mycroft]

Wow, if I made a silly play one deal in 15, I'd be *much* better than I am now. I guess definition of "silly" is like Allan Simon's levels of "can't play", of course...

The only difference between (some) robot games and human games is that after the deal, it's either rotated or switched so that the human gets the best hand (can't remember what barmar said last time). That might skew things, I've never investigated.

But "your feelings" about finesses, shapes,... are almost certainly observer bias. Humans are great pattern matching animals, they'll find patterns anywhere.

Please note, I have a convention on my card that does not appear on partner's: APAD (always pass as dealer). This applies doubly when I play the robots in non-best-hand games. I am just as susceptible to observer bias as anyone else here.

[pilowsky]

There is a confirmation bias about finesses when playing with robots.
It turns out that the likelihood of a finesse working or failing is the same as when playing with people.
The computer does not surreptitiously move the King or Queen while you are watching the children.

How 'good' are the robots? According to the EBU they have an approximate rank of "Q" with an average of 58.6%.
If you look at the results from yesterdays Zenith Daylong which attracted 1568 competitors of wildly varying skill levels in a NOT best-hand tournament, 58.6% means that they are playing better than about 90% of entrants.
A score of 50% puts you at 780/1560.

The relationship between Rank and percentage is only linear between ~45 and ~55. Above and below these points, the curve flattens. It's very hard to score 100% or 0% - I don't think I've ever seen it.

All the same, You are still playing against 'people'. You just never meet them because they play the same hand as you against three robots - some of the robots play better than others.
I think that it might be a worthwhile amendment to designate which type of 'robot' you are playing against.
Perhaps they could be called "GoodRobot" and "BadRobot", or "Robot" and "WillRobinson".

[thepossum]

What's different between humans and robots is something called variance which in my book has some association with fun

I suspect some people also modify their behaviour against robots or people. I play my more free natural game with people and learned that against the bot your play degenerates to boring and unimaginative. Thats just my personal experience. Throwing in a bit of strange variance is often all that gives any fun in some robot tourneys. Sometimes you win, sometimes you lose

EDIT As to issues of confirmation bias over certain statistical outcomes (finesses or randomness of hands) I reckon there is plenty of evidence to suggest that both the nature of the hands and the probabilities of how things turn out is different too I am sure the bot cheats sometimes. Then again I should be more careful trying to run Turing tests against GiB. And without being able to prove it obviously the nature of the hands and the bot's plays seems to have changed since I first started. But confirmation bias is a terrible thing

[msheald]

Thank you, all! I appreciate your replies. Best regards.

Mike