haha, there's an analysis on the web, funnily enough from randi's website -- but with a dead link.
the only other thing is a formula on flickr:
http://www.flickr.com/photos/chrstphre/6277432976/The formula is a P in the A to calculate, as it has a few n! calcs in it, as in most probability calcs -- and not sure if it includes 'card counting' as a factor, which is not necessary or desirable in a blinded test.
Just put the formula as supplied into a spreadsheet. According to the formula, the chances (by chance, rounded) of getting exactly
1: 1 in 42
5: 1 in 5
8: 1 in 16
9: 1 in 34
10: 1 in 85
11: 1 in 249
12: 1 in 854
13: 1 in 3,416
14: 1 in 15,941
15: 1 in 86,949
20: 1 in 5,477,852,523
25: 1 in 3 E 17 or 300 quadrillion (as suggested above)
in any single trial of 25 cards and guesses, of 5x5 different types.
Although the most common result of guessing would be 5 correct, the probability of any one trial delivering exactly 5 matches is 1 in 5, as noted. The distribution behaves as you would expect, i.e. the chances of guessing less than 5 by chance are lower than guessing 5 correct. One might think repeatedly guessing 10 or more correctly might be taken as considerably more than chance.
Interestingly, there are a couple of Flash apps that deliver a Zener test, which strikes me as a bit useless, as it's a bit hard for someone with mind-reading ability trying to tune in on what someone is thinking when that someone is a computer server somewhere in the world -- some people may have evolved being able to read other humans' minds for simple thoughts or images -- reading an 'image' off silicon RAM is a slightly different kettle of fish, and fundamentally invalid I would think. maybe a psychic computer can read another computer's mind!!!?? And do androids dream of electric sheep?