3 ms·
Agreed. Sites not loading for me, but I've looked into this nonsense from time to time over the last 15 years. A couple of issues I remember: 1. They only pub
by hackcasual 12y ago
Agreed. Sites not loading for me, but I've looked into this nonsense from time to time over the last 15 years.
A couple of issues I remember:
1. They only publish results for positive events. They only use pseudorandom data, not sampling other time frames of their existing data
2. Their random source showed consistent bias (I believe it was something like 51% chance of 1 vs 49% of 0. This is after Von Neumann whitening which means they have some additional biases in their random sampling
- ArneBab 12y ago1. They also publish for negative events. Look at the really low Z-scores: http://noosphere.princeton.edu/results.html http://noosphere.princeton.edu/results.html Also you can get all their data (downloading it day by day). 2. Possibly, but unavoidable to get rid of a systematic error which would pollute everything else. Details are here: http://noosphere.princeton.edu/gcpdata.html http://noosphere.princeton.edu/gcpdata.html
- hackcasual 12y ago1. They publish events that don't have high scores, but they don't publish non-events that do have high scores. 2. I was actually wrong on the VN whitening, they use XOR, either 10 for some devices, or a 560 bit mask for others(!). This is just silly. the XOR is used to exclude an important class of potential spurious effects — biases that might arise from temperature changes, component aging, etc. It's not excluding spurious effects, it's hiding it.