Re-trumping a trumped single-single-pair throw with two pairs matches the 2nd pair incorrectly. about shengji

[rbtying]

ooh, thanks for the bug report! will try to repro and then fix

I think your guess is probably right, but we'll check.

[rbtying]

So, it turns out what's actually happening here is:

The mapping of a given set of cards, (e.g. 6♣6♣2♣2♣) to a given trick-format (e.g. 2/1/1) is treated as fixed (attempt_format_match returns only one possible mapping, even though there may be more than one), as we can see here

It so happens that, in your example, the one it chose is the weaker mapping! but the point in the code where it does this logic isn't quite the place where it evaluates which trick wins...