How do robots "decide" on close defensive situations?

[mamos]

This was a hand from a robot BBO MP duplicate, with 30+ tables in play and human players sitting South

 

[hv=pc=n&s=sjt2h4daj83cakt54&w=s95hkjt932dq9cj82&n=sq86ha876dkt6c976&e=sak743hq5d7542cq3&d=s&v=e&b=3&a=1cp1h1s2cppp]399|300[/hv]

 

The auction was the same at the majority of tables and the first three tricks essentially the same. West began with ♠9 and the defence began with three rounds of that suit, with West ruffing Round 3. The only variation came from declarer's choice of which Spades to play at which trick. I began with small from dummy and played J - 10 - 2 in that order. In all cases the robot in East played ♠7 at trick 3. I assume this is a gentle Heart signal

 

At this point there was considerable variation from the robot Wests. Both the ♦Q and 9 were selected making 10 tricks for declarer pretty easy :rolleyes: . Some Wests chose a dutiful ♥K and against me a trump was continued.

 

Traveller for this board can be seen at http://www.bridgebase.com/myhands/hands.php?traveller=7277-1346503081-63437720&username=mamos

 

I drew trumps played ♥A and ruffed. Noting the ♥Q, I concluded that East was short in that suit and hence took an unsuccessful Diamond finesse. (This counting lark never seems to work!)

 

What interests me is the variation in the play of the western robot. Is it the result of the variations in declarer's play in the Spade suit or is it just that in close situations (if you think this is a close situation) the robots' actions are somewhat random?

 

Thanks if anyone can shed any light :)

[Quantumcat]

It depends on which hands the robot chooses to simulate with. At each decision, the robot generates some number of hands (50? not sure) that are consistent with the bidding and cards played thus far. It then performs double-dummy analysis on each of those hands, working out which play gives the highest average score (at IMPs) or the best likelihood of a good score (MPs) {this is explained better in other people's posts about IMPs and MPs robot play} across all the hands. Since he has the same cards and bidding in both of your situations (playing your cards in a different order), it will only be the particular simulations he chose that change what he does. {I don;t believe card order matters to the robot at all, he ignores your signals for example} I think if you choose a faster robot, the number of simulations is smaller, so you get more random results (e.g. if he simulates five hands and two of them have you with a void, he might try to give you a ruff, even if there is no real indication you have a void. If he simulates 50 hands then if five of them have a void that won't sway the resulting play as much). With slow robots they should choose the same play most of the time (since the simulations they each choose will have a lot of overlap) but with a large number of robots facing the same decision, the bell curve suggests there will be a small number of robots who choose hands to simulate that have little overlap to the rest of the crowd, causing their plays to differ.

[Bbradley62]

Quantumcat: In tournaments, all players have advanced bots. Also, the bots are set so that they will all make the same decision if they are faced with exactly the same set of circumstance (meaning if there is absolutely no variance in the actions of the humans). The order in which declarer plays his cards does affect the defending GIB's decisions.