I was originally using mlrate as a guide to implementing the most likely rating algorithm in java. However after getting it coded and working, I put some testcases in and they didn't give correct results.
I ended up dumping that solving method and using another one which gives the results I expect.
Is anyone willing to look at this with me and either explain what I'm doing wrong? Or maybe what I'm expecting to see is wrong.
Source code for mlrate is here: http://www.pem.nu/mlrate/
The main algorithm is mlrate() in mlr.c
You can contact me here or on KGS as yoyoma.