|
Sorry, I've only got a minute before I'm back working, so I need to keep this brief.
I had similar concerns earlier (months ago) concerning the fact that a DIPS system would not affect the stats that were produced but rather just 'how' they were generated. Thus, I wondered whether it was worth risking statistical accuracy for the few extreme examples, in exchange for a more realistic way of generating those stats.
In the end, I decided that Markus would almost definetly program around the outliers, and have some way of helping extreme examples (perhaps a hidden option on one in every 100) compensating (for who random chance doesn't alone). I slept a lot better after that!
Apologies for the brevity. Your post deserves more - and I agree that these issues must be raised!
|