| 0 | 1 | 2 | 3 | 4 | 5 | 6 | 7 | 8 | 9 | A | B | C | D | E | F | |
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| 0 | 24 | 23 | 24 | 18 | 17 | 22 | 15 | 18 | 22 | 24 | 18 | 21 | 23 | 15 | 24 | 14 |
| 1 | 20 | 23 | 15 | 19 | 21 | 17 | 20 | 17 | 21 | 30 | 20 | 22 | 18 | 24 | 24 | 22 |
| 2 | 18 | 28 | 22 | 24 | 21 | 17 | 19 | 11 | 23 | 25 | 23 | 18 | 9 | 18 | 28 | 22 |
| 3 | 16 | 18 | 25 | 19 | 20 | 16 | 23 | 23 | 20 | 17 | 18 | 20 | 19 | 17 | 25 | 20 |
| 4 | 22 | 22 | 21 | 15 | 16 | 24 | 20 | 22 | 17 | 15 | 19 | 16 | 18 | 12 | 16 | 15 |
| 5 | 24 | 20 | 25 | 18 | 22 | 24 | 22 | 26 | 22 | 16 | 16 | 16 | 19 | 21 | 23 | 20 |
| 6 | 31 | 20 | 23 | 18 | 16 | 20 | 24 | 14 | 17 | 21 | 22 | 20 | 28 | 8 | 22 | 22 |
| 7 | 19 | 24 | 20 | 21 | 19 | 20 | 14 | 21 | 13 | 24 | 16 | 20 | 13 | 15 | 20 | 16 |
| 8 | 16 | 24 | 16 | 17 | 23 | 19 | 17 | 17 | 22 | 19 | 16 | 20 | 15 | 18 | 23 | 23 |
| 9 | 14 | 20 | 17 | 20 | 25 | 19 | 25 | 16 | 19 | 25 | 21 | 29 | 22 | 14 | 17 | 21 |
| A | 23 | 24 | 15 | 26 | 17 | 26 | 17 | 19 | 13 | 19 | 18 | 20 | 16 | 27 | 19 | 21 |
| B | 17 | 17 | 20 | 26 | 14 | 31 | 27 | 17 | 22 | 23 | 21 | 15 | 19 | 20 | 11 | 16 |
| C | 17 | 16 | 18 | 21 | 16 | 22 | 22 | 24 | 20 | 18 | 22 | 12 | 21 | 10 | 13 | 16 |
| D | 20 | 15 | 21 | 17 | 9 | 19 | 20 | 11 | 13 | 19 | 19 | 23 | 14 | 14 | 13 | 23 |
| E | 22 | 18 | 18 | 15 | 22 | 19 | 20 | 19 | 24 | 19 | 27 | 18 | 20 | 17 | 22 | 22 |
| F | 19 | 21 | 26 | 22 | 12 | 19 | 19 | 20 | 17 | 10 | 24 | 26 | 14 | 20 | 24 | 18 |
This shows the first 5000 hexadecimal digits of pi. The rows represent values of d(n), the columns represent values of d(n+1). Each entry in the table shows the frequency of events which turn a given d(n) to a given d(n+1). This convention is adopted throughout these webpages.
The c++ program, pifreq.cpp (6K) can be used to generate phase spaces for other regions of pi.
Pseudo-phase spaces are used to search for patterns within number streams. At the moment, I doubt whether the above phase space could be distinguished from the phase space of a random number generator. I intend to include a Kolmogorov-Smirnoff test at some point. As a rough estimate, I think it would need at least 16M digits to produce any useable data.
{number of digits N=16M,
random walk error E = sqr(N) = 4K,
expected number per interval EXP(n) = N/256 = 64K}
Calculating the digits takes time, an obvious next step would be to save the results so that new digits could be added to the phase space.
That's enough for now.