[pilowsky]

I joined the Prime Club for several reasons:
1) Unlimited access to advanced robots for practice.
2) Ability to invite friends for practise in an environment where other people wouldn't mind while ironed out bugs in our understanding.
3) Ability to try out different approaches on the teaching table using the same advanced robots that I played the hand with in the first place.

It turns out that the advertisement - the basis on which I am paying for my membership - may be inaccurate.
It seems that the teaching table robots in the Prime area behave consistently completely differently to the Advanced robots in the daylongs. Their play appears consistently worse than daylong GIB.

To add insult to injury, rub salt in the wound and really twist the knife; if I try and play and 8 board round in the Prime club, it turns out that much of the time, I might be playing with one iteration of GIB while others are playing against an entirely different iteration of GIB.

The hand shown below was played 16 times.
On the 2 tables where the bidding went 1S-4S and South was a Prime member a ♦4 was led, and the contract makes for +9.2 IMPs.
On the 12 tables where the bidding went 1S-4S and South was not a Prime member, a ♣K was led, and the contract is 4♠-1 for -0.9 IMPs.

Now, when I bring friends - even nice ones - (I know, how can I tell?) onto a Prime table for a few quick practice hands, I first have to explain to them not to worry about the scoring at all because it's meaningless.
Sure, it's a 'nice friendly place' in the sense that no-one in Prime ever (insert whatever here).
Even so, social or not, the Prime Club is still a duplicate Bridge scoring Club.
I'm sure that people who pay for the privilege would prefer it if their bidding and cardplay was compared with other people playing the same GIB.
Even if the hands are NOT always being played in the Prime area. In the example below, only Prime players get the diamond lead. All the non-Primers get a club lead.

It sometimes works to the disadvantage of the Prime players. Either way, it isn't nice, it isn't friendly, and it sucks the fun out of practice.

Can I ask:
1. Are you aware of this issue?
2. Does it matter to BBO?
3. Are there plans to change it?
Here is the advertisement: (28 October 2019)

Quote

What is BBO Prime?
BBO Prime is a monthly subscription of $5.99/month offering

• access to a private bridge club only for nice, friendly people,
• free unlimited advanced robots in the Prime Bridge Club
• A monthly tournament (on the 28th of each month) hosted by Aurora with prepared hands (but not if you live in the wrong time zone)
• weekly dedicated daylong (every Saturday)
• daily and weekly bridge articles, and…
• now digital version of BeBRIDGE magazine every two months
• a monthly prime bidders challenge conducted by Marc Smith

BBO Prime comes with its own new private club: the BBO Prime Club is a pleasant place to play social and casual bridge with nice people.
And finally, the badge! BBO Prime members will have a pretty purple badge in profile (sic).

I open 1♠ North raises to 4♠, and I get a ♦ lead and make 4S=

---

Everyone else with the same auction gets a ♣K for 4S-1

---

[Vinish]

My question is how does a diamond lead allow you to make 4S? Does declarer not still have two clubs and two heart losers? Before he can set up hearts for a pitch, it would seem that the defense should find their clubs. Do the robots continue with diamonds after winning their first heart? Do they then continue with diamonds again after winning their second heart?

G'day,

Vinish

[pilowsky]

 Vinish, on 2021-March-09, 13:53, said:

My question is how does a diamond lead allow you to make 4S? Does declarer not still have two clubs and two heart losers? Before he can set up hearts for a pitch, it would seem that the defense should find their clubs. Do the robots continue with diamonds after winning their first heart? Do they then continue with diamonds again after winning their second heart?

G'day,

Vinish

It doesn't matter, but you can click through the play by hitting the 'next' button on the diagrams and see for yourself if you are interested.

There are two types of robot on BBO. Basic and Advanced. They are programmed in different ways.
One type plays according to one set of rules compared to the other.

The example I have given above is simply for illustration. Change the bidding slightly get a different result.

That is not the issue. The issue is why is it that during one session Prime punters get one type of opponent and non-Prime punters get another?

As a side-note, the only reason I started writing on this Forum (even though I'm a beginner/intermediate) is that when I wrote to BBO (support) about a year ago they told me to write on the Forum.

On the Forum, the people that run BBO don't reply. There is no acknowledgement. But, I'm patient; I keep trying.

On the plus side, I have learned a lot more about Bridge, made some friends, met Richard Willey and Mark Sheldon and a guy called Chas. So, not a complete loss.

[smerriman]

Why do you keep posting the same thing over and over? You've already been told multiple times this is not a bug, and it is working as intended.

Here was the last time, where I even quoted barmar, in case for some reason you specifically didn't believe me and would only listen to messages from BBO management.

[pilowsky]

 smerriman, on 2021-March-09, 18:14, said:

Why do you keep posting the same thing over and over? You've already been told multiple times this is not a bug, and it is working as intended.

Here was the last time, where I even quoted barmar, in case for some reason you specifically didn't believe me and would only listen to messages from BBO management.

res ipsa loquitir

Telling me that something that is obviously bad is a feature and not a bug is not helpful.

You are flat out wrong. If you have (mis-)interpreted someone else to be saying there is no problem (even though he does actually work as a programmer for BBO) is unhelpful.

Listen carefully Stephen: I KNOW IT ISN"T A BUG. The computer is working just fine - both of them.

The error is in the implementation. BBO advertises Prime as a place where something can happen. Because of the poor way that they have implemented it, it doesn't.

It's a little bit like saying to a fencing club "look we don't have enough space for you and the cricket club, just use that flat area in the middle of the grass. You might have to duck and weave a bit more than usual on account of the people with sticks and the one throwing a ball, but it's the best we can do."

So that's why I keep making these posts. I want to use the Prime area. I like it. But I keep getting hit by bouncers and jostled by two people with sticks.

[smerriman]

 pilowsky, on 2021-March-09, 19:21, said:

The error is in the implementation. BBO advertises Prime as a place where something can happen. Because of the poor way that they have implemented it, it doesn't.

Nowhere anyone on BBO does it say that in Prime, the robots will behave exactly the same way at every table given the same input.

Instead, it says that Prime offers free advanced robots.

Advanced robots are specifically programmed to give different results every time they analyse a hand.

If you don't like that, fine, but don't blame BBO for misreading what you signed up for. Simply quit.

Back to ignoring you again..

[pilowsky]

 smerriman, on 2021-March-09, 19:26, said:

Nowhere anyone on BBO does it say that in Prime, the robots will behave exactly the same way at every table given the same input.

Instead, it says that Prime offers free advanced robots.

Advanced robots are specifically programmed to give different results every time they analyse a hand.

If you don't like that, fine, but don't blame BBO for misreading what you signed up for. Simply quit.

Back to ignoring you again..

Still not listening.
I highlighted the error in your post.
They never give different results if the bidding is the same. Never.

Or are you wilfully ignoring the point?
They give EXACTLY the same result on the same bidding but a different lead.
The ONLY difference is that Prime members get one lead and non-prime members get a different lead.

CONCLUSION: the two populations are playing different robots.

btw, one full stop is usually enough.

[barmar]

 pilowsky, on 2021-March-09, 19:38, said:

Still not listening.
I highlighted the error in your post.
They never give different results if the bidding is the same. Never.

That's only true in robot tournaments.

Much of the design of GIB involves dealing a number of random hands that are consistent with the bidding and play so far, determining the result of each potential action in all the hands, and then selecting the action with the best average across them.

In robot tournaments, we ensure that the same random hands are used at all the tables that had the same bidding and play leading up to the decision point, so the robots at all tables will behave consistently.

Outside of tournaments, we make no attempt to ensure consistency across tables. A different set of random hands can cause the robot to choose differnt actions, especially in cases where the odds are very close between two actions. And even when the odds aren't close, a couple of outliers in the random deals can sometimes skew the results significantly.

The advanced robots that you're playing with use the same bidding/play algorithm as the tourney robots.

[pilowsky]

 barmar, on 2021-March-12, 19:22, said:

That's only true in robot tournaments.

Much of the design of GIB involves dealing a number of random hands that are consistent with the bidding and play so far, determining the result of each potential action in all the hands, and then selecting the action with the best average across them.

In robot tournaments, we ensure that the same random hands are used at all the tables that had the same bidding and play leading up to the decision point, so the robots at all tables will behave consistently.

Outside of tournaments, we make no attempt to ensure consistency across tables. A different set of random hands can cause the robot to choose differnt actions, especially in cases where the odds are very close between two actions. And even when the odds aren't close, a couple of outliers in the random deals can sometimes skew the results significantly.

The advanced robots that you're playing with use the same bidding/play algorithm as the tourney robots.

Thank you, Barry.
That's very helpful.
So, what you are saying is that in tournament situations (say a Daylong), " we ensure that the same random hands are used at all the tables that had the same bidding and play leading up to the decision point".
But not in tables in the Prime Club.
So long as I know that the Prime Club is simulating multi-skilled opposition, that's fine. I don't like it, but it is what it is. It feels like the tournament players are getting Prime cuts and the Prime players are getting salami - so to speak. In any case, I'll just leave it as something that I would prefer to be different.

OTOH
In another post, I provide an example (https://www.bridgeba...age__mode__show) which was Board number 4 from DL (MP) 1 - 2021-03-10.
Here, after the same bidding has occurred, 3 players get one lead, but 10 others get a completely different lead for exactly the same auction.

• Me and 2 others (pilowsky; 3NE+1; (3) -630; 2.94%; Your Movie). ♦4 is led to me (passive).
• The winner and nine others (Other Human; 3NE-1 (10) +100; 86.76%; Sample Movie). ♥5 led to others (4th down from AK653).

Does this mean that this was an error? Or am I missing something? Again?

[smerriman]

 pilowsky, on 2021-March-12, 20:03, said:

So, what you are saying is that in tournament situations (say a Daylong), " we ensure that the same random hands are used at all the tables that had the same bidding and play leading up to the decision point".
But not in tables in the Prime Club.

Yes. Just like I kept telling you:

 smerriman, on 2020-September-05, 03:18, said:

In a tournament situation, different tables will match the random seeds and ensure the robot runs the same simulations and comes up with the same bid. Anywhere else, like the situation your described, the simulated hands will differ, and therefore the outcome can differ every time the hand is played.

In your example you were the one that made the different lead, not the robot.. when the robot simulated hands at your table, it discarded ones where you don't have the 4 of diamonds, while at other tables it discarded ones where you don't have the 5 of hearts. Thus causing completely different results from that point forward.

[pilowsky]

 smerriman, on 2021-March-12, 20:30, said:

Yes. Just like I kept telling you:


In your example you were the one that made the different lead, not the robot.. when the robot simulated hands at your table, it discarded ones where you don't have the 4 of diamonds, while at other tables it discarded ones where you don't have the 5 of hearts. Thus causing completely different results from that point forward.

Oops! Thank you - I thought I was missing something.
But the something that I missed was that I made the correct lead to beat the contract.
Now it sort of makes sense.
Make the right lead, you fail,
Make the wrong lead; you succeed.
Obviously, in this case only.

Edit: I was conflating two posts. this one (3NT - I was on lead), and 4S - robots on lead. Both questions now understood.



FWIW, Stephen, sometimes people (viz., me) do have difficulty realising 'obvious' things.
Even if this never happens to you.
But thanks for the reminder.

This post has been edited by pilowsky: 2021-March-12, 21:10