[Bug-gnubg] Why strip post-Crawford MET?

[Guido Flohr]

Hi,

when GNUBG extends post-Crawford match equity tables it throws away the
last record.  At least for the most relevant case of post-Crawford data
which is type=explicit and player=both.  The rationale is explained in
this comment:

    /* Note that the post Crawford table is extended from
       n - 1 as the  post Crawford table of a n match equity table
       might not include the post Crawford equity at n-away, since
       the first "legal" post Crawford score is n-1. */

As a matter of fact, this type of table _must_ include an entry for
n-away, otherwise an assertion fails.  The question is rather academical
for a table of length 25 but significant for smaller tables.

For example dunstan.xml has only five entries.  The explicitely given
entry for 5-away gives a match winning chance of 19.07 % but GNUBG
silently replaces that with 18.3475 %.  The error is then propagated
further because the calculation is iterative.

Any objections against changing that and assuming that the author of the
table (data) knows the Crawford rule and has given valid data for
n-away?

Guido
-- 
Империя ООД                         | Imperia OOD
ул. „Княз-Борис-I“ № 86, София 1000 | ul. "Knyaz-Boris-I" № 86, Sofia
http://www.imperia.bg/