On the opening roll, Snowie calculates your luck on the
assumption that you were destined to win the opening
roll, disregarding the fact that you were probably lucky
to get the opening roll. It also assumes you could
have rolled doubles.
For example, if you start with a 31, Snowie 1-ply (which
it appears the luck calculation is always based on) says
you are worth +.154. Since your equity was 0 before the
game started, the luck calculation should be based on
this .154. Snowie instead compares your potential equity
with the 31 (+.154) to the equity with you on roll in the
opening position -- including doubles -- which Snowie
estimates as +.072.
David Montgomery
mo...@cs.umd.edu
monty on FIBS
>
>There is a bug in Snowie's luck calculation.
>
>On the opening roll, Snowie calculates your luck on the
>assumption that you were destined to win the opening
>roll, disregarding the fact that you were probably lucky
>to get the opening roll. It also assumes you could
>have rolled doubles.
I think there is a second bug, related to the import function. The
last move of a game is truncated in the output from FIBS logs,
although a clear message is sent about who wins the game. It would be
easy for Snowie to calculate an equity change for this last move, but
instead it makes the luck calculation as if the last shake never
happened.
The last shake can be extremely volatile. Say you have two remaining
on the ace point and your opponent has three there and is on shake.
You win 5/6 for a money play equity of +0.833. If your opponent tosses
a set your equity swing on that shake is -1.833. I'd argue that this
move should be included in the luck calculation.
I sent an e-mail to Oasya, but like so many other comments, this one
did not even get an acknowlegement, let alone a response.
Stuart
Oops. If you win 5/6 your equity is 4/6 or +0.667. So the antijoker
only costs you -1.667! Pretty bad luck in my opinion.
Stuart