[shyams]

This hand comes from a daylong IMP tournament of 7 October. The robot's play simply allowed declarer to make an extra overtrick. Yet, I found it strange because I thought the position is like an open book and it should be automatic for the East robot to find a duck at trick 9.

The link to this hand: https://tinyurl.com/yywoklas

After 7 tricks, the position was:

---

North now exits with a low diamond. East puts up the Queen, and West correctly overtakes to win with the King. West now switches to ♥7 and when I play ♥Q, East actually wins this trick!!! If East ducks as is normal, I lose two heart tricks and cannot make my overtrick.

The hand was played more than 150 times and it appears that the overtrick was conceded in some 45 instances and not conceded in some 100+ instances.

If these are advanced robots, shouldn't they (a) find the correct defence every time? (b) play identically at all tables?

[pilowsky]

Are you saying that this hand was bid and played in the same way in the same tournament by more than 150 players?
And then they all got up to trick 7 after playing it exactly the same way...

This seems incredibly unlikely, can we have a link to the traveller, please.

[johnu]

Faulty analysis by GIB standards. West could have had ♥K and North could have had ♦A. You might ask if North had ♦A then why wasn't it played on the 1st diamond trick. Fortunately, GIB never asks those kind of questions.

[shyams]

pilowsky, on 2020-October-09, 03:30, said:

Are you saying that this hand was bid and played in the same way in the same tournament by more than 150 players?
And then they all got up to trick 7 after playing it exactly the same way...

This seems incredibly unlikely, can we have a link to the traveller, please.

Let me clarify/correct my previous post. I did not check whether the hand was bid and played in an identical fashion. I realise that the bidding/play so far are inputs for the bot's decision matrix.

I was merely noting the times that the hand made an overtrick vs. the times it did not. To me, the number of times the OT was allowed was indicative of some recurring misplay by the bot.

[pilowsky]

shyams, on 2020-October-09, 03:43, said:

Let me clarify/correct my previous post. I did not check whether the hand was bid and played in an identical fashion. I realise that the bidding/play so far are inputs for the bot's decision matrix.

I was merely noting the times that the hand made an overtrick vs. the times it did not. To me, the number of times the OT was allowed was indicative of some recurring misplay by the bot.

I was also trying to think of an IMPS Daylong where >150 players would bid a hand the same way.

[shyams]

A very similar situation occurred yesterday at the BBO Forums Daylong. This was board 6 (link)

After 7 tricks, with dummy on lead, the situation was as below:

---

When I lead a club and play the ♣K, West can & should duck. This saves it from the obvious endplay and ensures two tricks in clubs. Yet, West immediately won and played back a club thereby gifting me the overtrick and a 77.5% on the board. If these are truly advanced robots in play, this should not be happening.

I haven't played Robot Rebate 55% for a few weeks now; I do recall that when similar positions arose in Robot Rebate 55%, the bots would always correctly find the duck.

[pilowsky]

Well done on your 77.53% - achieved by 210 others and beating 233 of us who made only 4H= with 29.89%.

I'm not mentioning anyone in particular, but someone has anger management issues with 0.00% making 7NSxx-13 for -7000.

In any event, you played the QUEEN not the KING fwiw. Could this affect the response?

[smerriman]

shyams, on 2020-October-12, 10:47, said:

When I lead a club and play the ♣K, West can & should duck. This saves it from the obvious endplay and ensures two tricks in clubs. Yet, West immediately won and played back a club thereby gifting me the overtrick and a 77.5% on the board. If these are truly advanced robots in play, this should not be happening.

Just to play the usual devil's advocate, from GIB's perspective, isn't it possible that you hold Qx of clubs? In which case ducking costs a trick.

[smerriman]

And likewise, in the first case, isn't it possible North has Qx of hearts and a second diamond, where ducking costs a trick?