[pretender]

I am not sure that it plays to maximize the likelihood of making the contract (although once a contract is considered "cold" by GIB it definitely starts to do weirder things on defense).

Does it maximize the overall trick expectation? ie. It has a line that is more likely to make the hand, but down 3 if wrong, so it chooses a line that is slightly inferior, but will only be down 1 if wrong.

If it does that, does it calculate scoring into the expectation or just tricks?

[arigreen]

I believe it tries to maximize its expected score, not the number of tricks taken. Occasionally this will cause it to risk the contract for an overtrick. It will only do this when it is fairly confident that the play cannot cost, but if some of GIB's assumptions about the location of unseen cards are wrong, it can go down in a cold contract.

[helene_t]

arlgreen is right. It maximizes total points.

[pretender]

Ok. Trying again.

Let's say

line A = .6 make 600, .4 down 3 for -300

line B = .5 make 600, .5 down 1 for -100

then A=360-120=240, B=300-50=250 it would take line B

but if both were nv

A=240-60=180, B=200-25=175 it would take line A

Does that mean stretching to game vul is a suboptimal strategy with GIB vs. normal play?

My brain is dead. If something wrong with the math please chime in

[655321]

Nothing wrong with the math, but this is nothing to do with GIB either.

The normal IMP (or total point) odds for bidding game assume making or one down. If you know that when your game doesn't make it will be down plenty then you need slightly better odds, i.e. you would perhaps not bid the more marginal games.

But in any case, your own example shows that you are better off by bidding this vulnerable game at total points because you can expect to average 250 points by doing so (and by correctly playing line B). Maybe at IMPS line A would be better, but you could change the numbers to make line B better at IMPS as well.

I don't understand why you think that your numbers suggest that you should be bidding fewer games playing with GIB.

[helene_t]

Maximizing total points makes you bid vulnerable games a little more aggressively than maximizing IMPs.

But maximizing IMPs is, strictly, not well-defined, as your optimal choice depends on what happens at the other table. It doesn't matter much, though, and it probably won't be very wrong in most situations to assume that the other table is faced with the same choices as you are.