How can Gib's simulation go wrong?

[smerriman]

There are two stages to Gib's defense. 1) It simulates hands consistent with the bidding/play so far, 2) it runs a Monte Carlo simulation with double dummy analysis and picks the card to play that results in the best average.

 

I can't see how there can ever be a bug in 2). It's double dummy analysis is perfect, and there is really very little logic involved with running a simulation.

 

Bugs such as Gib randomly throwing a vital high card really therefore have to be due to flaws in 1) - it must be making some inaccurate assumptions based on what has happened so far that causes it to think nothing it does matters.

 

[hv=http://www.bridgebase.com/tools/handviewer.html?sn=smerriman&s=SAQ982HAQ5DAT85CA&wn=Robot&w=SKT54H93D962CT742&nn=Robot&n=SJ7HT862D4CKQJ865&en=Robot&e=S63HKJ74DKQJ73C93&d=e&v=e&b=6&a=P1S(Major%20suit%20opening%20--%205+%20%21S%3B%2011-21%20HCP%3B%2012-22%20total%20points)P1N(Forcing%20one%20notrump%20--%203-%20%21S%3B%206+%20HCP%3B%2012-%20total%20points)P3D(Strong%20hand%20--%204+%20%21D%3B%205+%20%21S%3B%2018-21%20HCP%3B%2019-22%20total%20points%3B%20forcing%20to%203N)P3N(4-%20%21H%3B%203-%20%21S%3B%207+%20HCP%3B%2012-%20total%20points%3B%20partial%20stop%20in%20%21C%3B%20partial%20stop%20in%20%21H)PPP&p=DKD5D2D4S6S2S5S7H2H4HQH3CAC4C5C9S8SKSJS3C7CJC3H5&c=11]400|300[/hv]

 

Fastforward to West on lead at trick 6.

 

From a daylong tourney, so basic Gib. This means less simulations than advanced, but still some reasonable number.

 

How could a simulation possibly result in a club lead being equal or better than any other lead? (Clearly not better, which means every simulation had it equal.. but surely that couldn't be.)

[iandayre]

Deleted