Robot Daylong Scores

[m00036]

We all know how frustrating it can be to think you've done the right thing, only to find that you get a bad percentage on a bad split or poor luck. This is clearly a part of bridge that can't be controlled, and in the "long run" those who play their cards right will do better. However, I was wondering whether each daylong board could also be played by a robot (like in "Challenge a Robot"), giving you an indication (and only an indication!) of what % you should have got if you did everything right. You can then compare your score to what was "achievable" and see how/whether you should have improved.

 

I'd be interested to know what others think about this - clearly BBO robots are far from perfect although they are definitely a good guide. I thought this was particularly pertinent to daylongs though since everyone plays different boards, so it would be nice to get a more "personalised" comparison (it is for this reason that the EBU, for example, require people to play at least 70% of boards, but that's sadly not feasible in an online environment).

[barmar]

Can't you achieve the same thing by looking at how the human players who did well on the board played it? Why does it have to be a robot?

[DozyDom]

I can't imagine your play being worse than a robot's, surely?

[smerriman]

I can't imagine your play being worse than a robot's, surely?

You should check out the scores of the current forum challenge event.

[thepossum]

The best thing I have found is to use the traveller which gives the frequency of each human's scores on the hand for comparison. I certainly would not compare myself to a robot to get an indication of how good you are :)

 

PS If you want to look at theoretical scores after the hands you could use something like Bridge Solver (App and Chrome extension) which is integrated with BBO. However of course that is a full information solver. I like to use that to see where in playing a hand I went wrong, if its not obvious :) I dont know how it always compares with a top players on every hand play but it rarely misses the possible scores in my experience. You can usually see if the scores were down to bidding, declarer play or defence.

[johnu]

BBO has said that have run tests with a robot as a participant in the robot tournaments several years ago. IIRC, the robot scored in the 55% range when ranked with the human players. Obviously this is highly dependent on the quality of the BBO tournament fields.

 

I have no idea how the human fields would rank in comparison to club games, or more major tournaments. I don't know if it is due to misclicks, lack of attention, exasperation with the robots, whatever, but there are a lot more inexplicable bad results in robot tournaments that you wouldn't normally see in a real life game. Based on the barest of facts and without any confidence at all, I would guess that a 50% BBO player would score about 45% in a live game.

[barmar]

I would guess that a 50% BBO player would score about 45% in a live game.

I wouldn't be surprised. I think I generally do better in bbo robot games (except robot challenges, which I rarely win) than I do in club games and tournaments. Of course, my human partners contribute to the latter, but I don't think I'm that much better than the people I play with.

[svengolly]

I often wonder how the top winners achieve some of the insane scores I regularly see like 82%. Makes me wonder if some don't use multiple accounts to achieve them.

[barmar]

I often wonder how the top winners achieve some of the insane scores I regularly see like 82%. Makes me wonder if some don't use multiple accounts to achieve them.

You'd need to play quite a few times to make that much of a difference, because you're unlikely to play the same boards each time.

 

Maybe someone with better probability skills than me could give precise odds. Assume 25 different deals for each board, how many times would you have to play before it's likely that N deals in a single tourney have been seen before?

[johnu]

I often wonder how the top winners achieve some of the insane scores I regularly see like 82%. Makes me wonder if some don't use multiple accounts to achieve them.

 

In one of the short ACBL test tournaments before a national robot tournament last year (?), one of the players scored something like 96%. A well known expert with a spotless record. A combination of "swingy" boards that weren't "flat" e.g. 3NT or 4 of a major with a set number of winners and no way to take more tricks, superior play and luck, bad/awful play by the robots, and very weak opponents who played the same boards.

 

Average or even pretty good players are unlikely to score 80+% except in the very rarest of situations.

[barmar]

You can find some threads at Bridge Winners where during the NABC Online where the player provided a post mortem of his pay in the event.

[smerriman]

Maybe someone with better probability skills than me could give precise odds. Assume 25 different deals for each board, how many times would you have to play before it's likely that N deals in a single tourney have been seen before?

Average number of tournaments you have to play before finding one with N deals you have seen before:

 

1: 3.3

2: 5.2

3: 7.6

4: 10.5

5: 14.0

6: 18.6

7: 24.8

8: 34.8