[jardaholy]

In Covid-19 time, I started playing daylongs. I have played around 10 000 boards approaximately, mainly in BBO MP1, MP2, IMP1 and JD MP1. I also participated in BIC.
I reached 1000 BBO points recently. I like daylongs and play them often.

I have several general, but important questions to the daylongs:
Question 1: How are the boards for daylongs generated? Is it purely random generation or there is some selection of generated boards?
It seems to me - and not to me only - that the boards are selected from much broader population. In comparison with real bridge world in club, these boards are
much more entertaining, difficult and didactic. In other words - the frequency of interesting boards separating average, good and excellent players is much higher than
in real bridge (sometimes, it seems to me that any almost board represents interesting bridge problem). Is it just my feeling or the boards are selected such a way?
Question 2: I understand that a kind of principle of broader pool of boards is applied in any daylong to prevent cheating. My question is: how are concrete boards assigned to
concrete players? Is it based on completely random selection or there is some additional selection on the base of characteristics of the players? Do players
with more BBO points (BBO masters) get more difficult, more unbalanced, specific in some manner etc. boards? (I would not expect that, but would like to know that for sure).
Question 3: How are the PLAYERS divided into the groups playing the same boards? Is it completely random or are the experts put in the groups with more experts than the
average players (or something like that)
Question 4: Is the development of robots´ bidding and cardplay continuing (Up to now, I collected more than 120 boards with serious robots´ mistakes of various kind)?
Question 5: Is there some development of the explanation of bidding system used by the robots, which is available during bidding? There are plenty very strange explanations -
like the distributions 5-5-4-0 (5-5 in major suits and 4+ cards in diamonds in addition) or like the situations, there are 48 points (HPC) around the table etc. Some bids are explained as
determined for the boards with 25-32 HPC and some even for 32+ HPC.
Question 5: (a very important one) Is there some development of GIB manual? It seems to me that it would deserve more details in some areas of bidding, but in particular
in carding. As a risk engineer, I completely understand the principles, robots´ defensive play is based on (Monte Carlo simulation), but I am reluctant to believe that
it could be such big problem to add a bit more signals to the robots play (o/e, for example), what would immensely improve the play of both bots and human players.

[smerriman]

To be succinct:

1,2,3) Purely random.
4,5,6) No.

OK, BBO haven't denied the latter, but:

- the last update was two years ago (bidding only; I believe the only cardplay change in the last 10 years was about 7 years ago which was a staggering complex change of.. forcing opening leader to cash an ace against 7NT).
- the BBO representative who used to respond to bug reports on the forum said he'd start doing so more regularly, then suddenly stopped and hasn't posted in 3 years.
- my first bug report on this forum was a simple bidding sequence back in June 2016 (RHO opens 2S; I overcall 2N, GIB transfers to 3H and then passes with a 1624 13 count). The founder of BBO himself replied and said he'd make sure the programmers looked at it. It is still 100% replicable today.

So, if that isn't a no, I don't know what is.

[pescetom]

 smerriman, on 2021-February-14, 18:39, said:

- the BBO representative who used to respond to bug reports on the forum said he'd start doing so more regularly, then suddenly stopped and hasn't posted in 3 years.

He did however show up briefly last year, to chastise me that enabling directors to impose penalties as foreseen by the laws of bridge (which boils down to read a small number and subtract it from the total of MP or IMPs) was not a simple programming task.

[jardaholy]

 smerriman, on 2021-February-14, 18:39, said:

To be succinct:

1,2,3) Purely random.
4,5,6) No.

OK, BBO haven't denied the latter, but:

- the last update was two years ago (bidding only; I believe the only cardplay change in the last 10 years was about 7 years ago which was a staggering complex change of.. forcing opening leader to cash an ace against 7NT).
- the BBO representative who used to respond to bug reports on the forum said he'd start doing so more regularly, then suddenly stopped and hasn't posted in 3 years.
- my first bug report on this forum was a simple bidding sequence back in June 2016 (RHO opens 2S; I overcall 2N, GIB transfers to 3H and then passes with a 1624 13 count). The founder of BBO himself replied and said he'd make sure the programmers looked at it. It is still 100% replicable today.

So, if that isn't a no, I don't know what is.

Thank you very much for the answers!
Just one recent example from my collection of boards (played 28.1.2021, BBO IMP1 daylong).

---

I was thinking about 2C opening, but since I have got very bad experience with reaction of the robot on the opposite side having good values (jumps to 6NT or 7NT in next rounds of bidding), I decided to be conservative.
As you can see, it did not help. 4D was cuebid and strong nice hand so that nothing could stop me from bidding slam after Blackwood response promising two aces from five. I expected most of the missing points in the diamond
suit so that we really have to be in a good slam. Actually, the first part of my assumption was completely true, because all diamond points were located at the opponents, but the second part was unfortunately wrong...

[mythdoc]

 jardaholy, on 2021-February-14, 17:04, said:

How are the boards for daylongs generated? Is it purely random generation or there is some selection of generated boards?
It seems to me - and not to me only - that the boards are selected from much broader population. In comparison with real bridge world in club, these boards are
much more entertaining, difficult and didactic. In other words - the frequency of interesting boards separating average, good and excellent players is much higher than
in real bridge (sometimes, it seems to me that any almost board represents interesting bridge problem). Is it just my feeling or the boards are selected such a way?

I enjoy robot bridge very much and play almost daily. I have long felt the same suspicions that you posted. I have even developed some “rules” for bidding the daylongs that are grounded in the observations quoted above. A kind of contrarian approach for avoiding what is likely to go wrong based on seemingly straightforward bidding sequences.

I also play a lot of challenges of all the different types of scoring. Recently I played a series of “just declare,” IMPs challenges with friends and found that the hands seemed a lot less “entertaining, difficult, and didactic,” as you so eloquently put, compared to “just declare,” MP challenges. Of course, one would have to generate so many hands to find any statistical evidence. Not my specialty. But they did seem strikingly simpler and uninteresting.

When all is said and done, I am going to play the hands whether they are random or cooked. May the best player win!

[smerriman]

 jardaholy, on 2021-February-15, 15:35, said:

As you can see, it did not help. 4D was cuebid and strong nice hand so that nothing could stop me from bidding slam after Blackwood response promising two aces from five.

Yeah, the robots are notoriously bad at slam bidding. With a human partner, it's trivial to avoid slam after 4♥ - 5♣ - 5♥, but the robots simply don't know how to control bid properly and no doubt would launch into blackwood after the normal 4♥ cue.

I've done far too much analysis of daylong hands and any sense of non-randomness is purely in your mind.

My impression is that their biggest flaw is the fact you get the usual number of boring hands, which give you unavoidable 50% scores and drop you down the table.

[pigpenz]

does just bidding 5♠ help at all, it used to be that shows 2 quick losers in that suit, but
GIB has a hard time with normal things