tomkron

In a tournament this weekend, I hit the "explain" button to get more detail about the alerted bid. The new explanation didn't answer my actual question, so I hit "explain" again, but the newer explanation still didn't answer my actual question. This gets frustrating. It would be nice if the "explain" button had an entry field where a specific question could be asked of the bidder who alerted.

Another variant would be to rotate the hands such that the human or his partner has the best hand, choosing which at random. Now even if you have a good hand your partner could have a better one.

Actually, it's a terrible random number generator, even if you are using excellent seeding. (How are you seeding it?) The version of rand() on my Linux box has a period of at most 34,359,738,352 (~2**35) per the man page which means it can produce only an infinitesimal fraction of the 53,644,737,765,488,792,839,237,440,000 (~2**96) possible bridge deals. That tiny fraction might be representative of deals in general, but maybe the people doing the complaining are in fact observing patterns which weren't checked for by your statistical analysis. Why can't you use a better random number generator, perhaps Mersenne Twister? It has a period of roughly 2**19937. Or get 96 bits of entropy from a hardware RNG or /dev/random or a cryptographically-secure RNG and use them to select a deal from the space of all possible deals? Several web sites (I believe Thomas Andrew's and separately Richard Pavlicek's) contain descriptions of how to do this.

It would interest me, at least, to get statistics about auctions from a large number of hands by many players all using the same system and same type of scoring. Frequency of various contracts, longest auction, some metric of similarity of different auctions if anyone could think of how to define and measure such a thing (i.e, 1H-4H and 1S-4S are different, but very similar in some sense), etc. That might be possible here since the robot individuals are all using the GiB 2-over-1 system. A bunch of matchpoint hands could be looked at, and/or a bunch of imp hands. Of course 3/4 of the bids would be GiB rather than Human, but the analysis would be interesting to me anyway.

When I click on the button for the preview version and get the username and password boxes, Firefox complains that the password box is not secure and shouldn't be used. It's Firefox 62.0 64-bit on fully-patched Windows 7, if that matters. Firefox's complaint

There's some logic: 50% is in some sense the generic expectation for the typical player, so the dampening is making the assumption that future boards are more likely to regress to the mean than to continue being extreme. It might be interesting to look at a large dataset and see how effective early boards are at predicting session results, vs. using the prediction that non-average results will regress toward the mean. That suggests, I suppose, that using a player's long-term average rather than 50% would be a better damper.

Also, it mitigates the emotional impact of the barometer after a great first board, where your second board is pretty good but the barometer reading drops, then the third board is also pretty good but the barometer drops.... It's consistent with the 50% shown before the first board is played, too.

OK, thanks!

That bridgemate gives distorted barometer readings in face-to-face bridge is not really relevant to whether BBO's barometer could be improved in online bridge. I know that an extreme result on an early board isn't the end of the game; indeed my objection to the current barometer is that it emphasizes the early results disproportionately making them seem artificially like they are the end of the game. That was my motivation for thinking to request an option to hide the barometer. Then I don't have to use post-its to hide it. But, on reflection, I thought the barometer could be modified to give each board equal apparent weight. This table shows the barometer readings I got on an actual average session, compared with what the hypothetical dampened barometer's readings would be after each board. Board Actual Actual Dampened Result Barometer Barometer 1 7.1% 7.1% 46.4% 2 50.0% 28.6% 46.4% 3 53.6% 36.9% 46.7% 4 7.1% 29.5% 43.2% 5 14.3% 26.4% 40.2% 6 85.7% 36.3% 43.2% 7 67.9% 40.8% 44.6% 8 89.3% 46.9% 47.9% 9 57.1% 48.0% 48.5% 10 71.4% 50.4% 50.3% 11 82.1% 53.2% 53.0% 12 57.1% 53.6% 53.6% Maybe people would like it better, or maybe not; but I think I'd like it better.

That's pretty arcane. Reminds me of sendmail format. Thanks for the reply, but I was actually asking about the CLI format relative to understanding what would be involved in integrating a different bot. Does each call to GIB pass it the entire history of the deal so far? Does the GIB program stay running for the duration of the deal and handshake back-and-forth as new info becomes available? That sort of thing. Thank-you for any insight you are willing to share.

I find I wish I could hide the barometer like I can hide My Results during matchpoint play, but on reflection I think it's because in showing the average of the played boards it gives the appearance that the early boards are more significant than the later boards. This makes an early bad board seem overwhelming or makes you watch as an early great board declines board after board (even after good boards). Would it make sense to dampen the barometer so that early and later boards had more equivalent effect on what it showed? I'm thinking of something like populating a dozen (or however long the event is) boards all at 50% then as each board is played, replace that board with the actual percent. This would converge to the correct result at the end, but earlier boards wouldn't oscillate so wildly. Thanks.

Is there documentation anywhere on GIB's input/output formats? I would be curious to know how simple or complex this format was.

Putting firefox back to the original default privacy settings did not make any difference.

I have uBlock Origin and HTTPS Everywhere, but disabling them makes no difference. I also have all the private settings turned up, e.g. it doesn't remember any history.

The "www." may be the key factor. If I start at www.bridgebase.com and use the login screen, or if I use a bookmark to www.bridgebase.com/client/client.php, then I am successful. If I start at bridgebase.com and use the login screen, or if I use a bookmark to bridgebase.com/client/client.php, then it hangs on the blue screen instead of presenting the login dialog. It seems strange to me that they behave differently, but the difference is consistent on my machine. Now I'm curious whether anyone else sees this difference.