I have two separate implementations of Hull-White ? one of them is based on a paper titled ?Using Hull-White Interest-Rate Trees? and the second based on a paper they wrote some time later, called ?The General Hull-White Model and Super Calibration?. I am certain the implementations are correct, given that the trees they generate are the same as the examples in the papers, but when I use the same example in both models, I end up with different trees. Both trees seem to be calibrated correctly, but the shape of the trees, the values at the nodes, and the probabilities are different. I?m not sure I understand how this would be possible, given that the underlying process should be the same. Has anyone seen this? Is there a possible explanation beyond errors in the code? I?ve run several examples based on examples in books (most of them based on the first paper), and I seem to be getting the expected results. Still both implementations yields different results when the same example is run on each.

Yes, I've seen tiny differences in specifications on how they match the moments of the distribution

Hi, LBDo you still have the implementation for both paper?Thank you.Best regards,John.

I would like to bring this topic to the top and raise the question and see if anybody would kindly help to explain.In the two paper, there are basically two different branching logic. In 1996, it first find the max and minimum nodes and choose a different branching probability calculation on the boundaries. In 2000, they use a new index k to make sure in each node, probability will be in the range of 0 to 1.I was just wondering are these two version consistent? Looks like the 1996 version is a special version of 2000. It assumes constant vol and time interval. Can I get rid of the 0.184/M part when I implement 2000 methodology? Appreciate all the help and opinions in advance.