Probabilities don't appear to reflect posterior probability

[RWilliamson1994]

Hi,

Currently have this history.

I have noticed that you don't appear to be updating my probability of having a small spike, despite the fact that my inputs are very specifically for a large spike or a decreasing pattern. I don't know the underlying distribution, but assuming that the probability is uniform over the small spikes range, the posterior probability of me being in a small spike pattern given my history should be very low (applying Bayes theorem).

I don't know how the system works though, so this may be incorrect if it is very likely to see this pattern for a small spike.

Cheers

[artem6]

Hi, thanks for the feedback. You are correct, I should adjust the probability of the small spike pattern down in your case by about 89%. I was considering adding more precise probabilities as an enhancement but reprioritized since it was a fairly small improvement and a lot of work. For the Monday AM price, the decreasing/large spike patterns predict 77-82 bells (given your purchase price). Small spike is a wider range of 36-82 bells that morning. Every subsequent day, the price falls by 3-5 bells on all three patterns, so only the Monday AM price should technically affect the probability. Monday AM prices are uniformly distributed within the patterns, so getting a small spike price in the range of 77-82 bells is fairly unlikely: 5/46 or about 11%. I think this quick calculation is accurate, but let me know if you see a mistake in my logic. Thanks again for pointing out the discrepancy.