Open ThePokerDude opened 8 months ago
Yes, this looks like a Lebensohl error
Lebensohl is difficult. What bothers me more is that N even though it thinks X +185 and pass +40 does not double. I'd suppose the expected score is the criterion which bid to choose. The choice between a +185 and a +40 bid seems obvious but somehow BEN chooses the +40 choice.
Candidates: Bid: P, Expected score: 40, NN Score: 0.871, Adjusted: 436 Bid: X, Expected score: 185, NN Score: 0.126, Adjusted: 63
Score is the neural network, and it adds a bonus of 436 giving 476, that is greater than 185+63 Might be wrong here, and can be adjusted, how much trust there should be in neural network compared to samples (They are just samples)
ok thanks. Now I understand. Probably adjustment is needed. But fine tuning will come later :)
1N 2S 3H - is forcing, with a weak hand with hearts you bid 2N lebensohl 1S 1N 2S 3H - maybe here 3H is also forcing, so 3H is a mistake?
But after 3N this is the evaluation by N:
Bid: PASS by Simulation Candidates: Bid: P, Expected score: 40, NN Score: 0.871, Adjusted: 436 Bid: X, Expected score: 185, NN Score: 0.126, Adjusted: 63
Wy does it pas when it expects X to be 145 points better?