What's up with this computer analysis??

Well, c8=B gives you the Bishop Pair :)

c8=B is more fun (and mildly obnoxious) haha

No matter whether you promote to a B or Q, black must take, so you will lose the just-promoted piece. Clearly, it is better to lose a bishop rather than a queen .

Seriously, though, I suppose the moves were evaluated at different depth when the analysis time limit per move kicked in.

I always have the same reaction to Halloween Gambit games:

"Cool.  I should learn this opening sometime." And then I go right back to hoping nobody plays main line Petroff with me.  Where did you pick up the Halloween?

rooperi is right - the bishop pair in open positions is easily more valuable than the queen pair =P Nice find - it seems that the program gives some weighting to a faulty assessment of the value of lost promoted material.  I wonder if this affects other evaluations...

well the black rook isnt doing....anything, where as, your pawn is, so why allow black the opportunity to capture a queen with his Useless rook, instead you promote to a bishop. this way black might not consider the immediate -2value of capturing the bishop as a good idea.  makes sense to me.

What makes this even more curious is that the alternate line offered after 11. ... Ke6 has 13. c8=Q+ in it.

I picked up the halloween gambit after entering multiple tournaments with that as the theme.  i've done pretty well with it.  of 21 games as white, i've won 12, lost 8, and drawn one.  as black, i've played 13 games, of which i've won 9 and lost 4.  i'm currently in the second round of this tournament (http://www.chess.com/tournament/96-halloween-gambiteers), soon to move onto the third.  it's a lot of fun to play, and i've got a good opening book for it.