[Thranduil]

I just had this game in today's Free Daylong Tournament:

Spoiler

How can that be a bottom? My hand is a clear 1♥ opener, we are always going to reach 4♥ and as far as I know robots always play the same in the same given situation, meaing they'll always find the ruffs for three down. So, bugs aside, how is that score even possible?

[smerriman]

They play the same if you play the same. I don't know what the lead was or what you played, but you have different cards to play to their tricks.

[Thranduil]

smerriman, on 2024-August-22, 02:16, said:

They play the same if you play the same. I don't know what the lead was or what you played, but you have different cards to play to their tricks.

Lead was the ♠K (J from dummy, small from hand) and then a switch to diamonds. I fail to see how my play here affects what the robot does? As soon as it switches to diamonds and its parter ruffs, it knows the exact distribution of the diamond suit and how to defend to bring the contract down three times, regardless of which diamonds I toss on the ruffs.

From my perspective I did nothing wrong, 4♥-3 should be a roughly average score considering the 4♥ are cold (if it wasn't for the ruffs) and the bidding is about as simple as it can get. The only explanation for that score being a bottom from my perspective is either a bug (much more likely) or that everyone but me has misbid this board and did not get to 4♥... What adds to the strangeness of my score being a bottom is that E-W actually have a grand in spades, meaning if the auction goes wrong and they end up playing, they will certainly score more than 150.

[smerriman]

Thranduil, on 2024-August-22, 02:26, said:

The only explanation for that score being a bottom from my perspective is either a bug (much more likely) or that everyone but me has misbid this board and did not get to 4♥...

Neither of the above; once the tournament concludes and you can see the other tables, you will virtually certainly discover that the robots took a different line at other tables after identical auctions. And it took a different line because you played different cards to earlier tricks.

The robots base their plays on random numbers. If you play identical cards to other tables, the random numbers will be identical, and the robots will make the same plays. If you differ slightly, your table will have a different set of random numbers.

Perhaps 99% of the time the robot will take the line they did against you, and take the contract down 3. 1% of the time, it gets a weird sample which tells it to cash a second spade at trick 2, or something else. If there was no synchronisation of random numbers, 1% of tables would get lucky and only be taken down 2 for a huge MP score, while everyone else would score a tick under 50%.

But when that 1% chance hits, it hits for *everyone* who played the same way. If this happens to occur when you make the 'boring' play of the lowest card to each trick, which most other tables would have done, they will all get lucky and score just above 50%, while the people who followed with less common cards will get a very low score, as here.

[Thranduil]

smerriman, on 2024-August-22, 02:53, said:

Neither of the above; once the tournament concludes and you can see the other tables, you will virtually certainly discover that the robots took a different line at other tables after identical auctions. And it took a different line because you played different cards to earlier tricks.

The robots base their plays on random numbers. If you play identical cards to other tables, the random numbers will be identical, and the robots will make the same plays. If you differ slightly, your table will have a different set of random numbers.

Perhaps 99% of the time the robot will take the line they did against you, and take the contract down 3. 1% of the time, it gets a weird sample which tells it to cash a second spade at trick 2, or something else. If there was no synchronisation of random numbers, 1% of tables would get lucky and only be taken down 2 for a huge MP score, while everyone else would score a tick under 50%.

But when that 1% chance hits, it hits for *everyone* who played the same way. If this happens to occur when you make the 'boring' play of the lowest card to each trick, which most other tables would have done, they will all get lucky and score just above 50%, while the people who followed with less common cards will get a very low score, as here.

I am certainly not the only one who played the ♠J in the first trick, knowing that there is no way to prevent the loss of two tricks in that suit. It literally does not matter which card is tossed on a spade lead.That is emphasized even further by the leader having AKQ in that suit, meaning they already know where the Q is (otherwise discarding the J could give away that declarer is short in spades without the Q).

Or does my play in earlier boards affect the RNG seed too? So I am doomed because I am among the only ones who found the winning play in board 2 to make the 3NT and thus played differend cards? If that is the case, the system needs to be fixed. In a competition like this, everyone should play to the same conditions, including the robots not randomly playing one for a bottom score just because that player played a very rare (winning) line in a previous board - therefore the RNG seed needs to be reset at the beginning of every board. Otherwise, why are you even playing with robots if you cannot rely on them doing the same in the same situation? What's the point of robot tournaments then?

[smerriman]

No, the random numbers are linked solely to the current board.

The choice of cards you play don't have any influence whatsoever on the best line. But they influence the random numbers, and that's what GIB uses to choose what to play.

I just gave this to an older version of GIB, asking it to perform simulations of 30 hands (which is an estimate of what they use in free tournaments), then asked it to play this hand 50 times. On 5 of those 50 occasions, it will switch to the spade Ace at trick 2, and the best it can do then is take you down 2. The other 45 times, it switched to a diamond for -3.

So:

90% * 90% of daylongs, people that had different random numbers by playing different cards would all end up down -3, and everyone would score 50%.
90% * 10% of daylongs, you would score -3 while everyone else scores -2, and you score 2%, like here.
10% * 90% of daylongs, you would score -2 while everyone else scores -3, and you score 98%.
10% * 10% of daylongs, everyone would get -2, and you'd score 50%.

You can confirm this after the daylong ends, but I can guarantee that's what happened.

[Thranduil]

smerriman, on 2024-August-22, 03:09, said:

No, the random numbers are linked solely to the current board.

The choice of cards you play don't have any influence whatsoever on the best line. But they influence the random numbers, and that's what GIB uses to choose what to play.

I just gave this to an older version of GIB, asking it to perform simulations of 30 hands (which is an estimate of what they use in free tournaments), then asked it to play this hand 50 times. On 5 of those 50 occasions, it will switch to the spade Ace at trick 2, and the best it can do then is take you down 2. The other 45 times, it switched to a diamond for -3.

So:

90% * 90% of daylongs, people that had different random numbers by playing different cards would all end up down -3, and everyone would score 50%.
90% * 10% of daylongs, you would score -3 while everyone else scores -2, and you score 2%, like here.
10% * 90% of daylongs, you would score -2 while everyone else scores -3, and you score 98%.
10% * 10% of daylongs, everyone would get -2, and you'd score 50%.

You can confirm this after the daylong ends, but I can guarantee that's what happened.

So you are saying that me carelessly discarding the ♠J doomed me here? because that is the only card played before the diamonds switch. It does not make sense to me that I am the only one who discards the ♠J on the first trick, therefore getting the diamonds switch immediately for three down.

[smerriman]

The combination of the two spades you played to trick 1 seems easily the most likely, though it's possible the way you played diamonds caused a difference later too. You'll be able to tell where the difference is when you see the other tables. But most people follow low; the fact you scored 2% meant you weren't the only one who played as you did, but that your cards were rare compared to the majority.

[cencio]

tomorrow look the results of other tables

[Thranduil]

Thanks for the explanation!

[smerriman]

Confirmed - everyone that played low on the first spade (i.e. almost everyone) received a spade continuation for -2.