Our testing · September 23, 2026

NIST Randomness validation

We used the NIST Statistical Test Suite to check one million bits from our Random Numbers API for bias and patterns. Here’s what we found.

Back to Random Numbers

14 test groups passed every check.

1 million

Output bits tested

14 / 15

Test groups passed in full

184 / 188

Individual checks passed

For everyday choices, dice rolls and simulations, these results are encouraging. The sample passed every check in 14 of the 15 test groups, including checks for balanced bits, repetition and unusual structure.

The remaining group looks for specific bit patterns. Four of its 148 checks were flagged because those patterns appeared more or less often than expected. The other 144 passed.

Flags can occur by chance in random data. This run does not establish the cause of these four flags; the complete results are included below.

Explore the results by test

NIST STS results by test type
What we checked	Passed	Flagged
Are zeros and ones reasonably balanced?Frequency (monobit)	1	0
Does that balance hold across smaller sections?Block frequency	1	0
Does the sequence drift too far toward zeros or ones?Cumulative sums	2	0
Does it switch between zeros and ones as often as expected?Runs	1	0
Are long runs of ones within the expected range?Longest run of ones	1	0
Does arranging the bits into grids reveal too much structure?Binary matrix rank	1	0
Are there unusually strong repeating cycles?Discrete Fourier transform	1	0
Do specific bit patterns occur too often or too rarely?Non-overlapping templates	144	4
Do overlapping patterns occur as often as expected?Overlapping templates	1	0
Is the sequence unusually easy to compress?Maurer's universal	1	0
Do short patterns repeat more than expected?Approximate entropy	1	0
Does a walk through the sequence revisit positions as expected?Random excursions	8	0
Are the total visits to those positions within the expected range?Random excursions variant	18	0
Do different short bit patterns appear in balanced proportions?Serial	2	0
Can a simple linear rule describe the sequence too easily?Linear complexity	1	0

How we tested

We collected 31,250 numbers through the same API available to customers, then ran all 15 test types in NIST’s Statistical Test Suite. Together, they examine different characteristics of a random sequence.

We used the first complete sample and kept every result. You can download the sample, read the full results or reproduce the checks yourself.

Detailed results (JSON)Test sample (125,000 bytes)Original test logs (ZIP)NIST suite and documentation

Technical method and sample fingerprint

One sequence of 1,000,000 bits, collected in 32 API jobs across the full signed 32-bit range. We used NIST STS 2.1.2 with default parameters and a significance level of 0.01. A check passes when its p-value meets that threshold.

We preserved the original order, added 2,147,483,648 to each integer and encoded it as unsigned 32-bit big-endian bytes, read most-significant bit first. No values were filtered, substituted or padded. Individual tests use their standard block handling.

The suite was checked against NIST’s reference sample before testing our output. This run evaluates one sequence; the original logs’ multi-sequence summary fields are not used in our results.

Sample SHA-256:

7c8b5ba8632c97a0af747a51e3940b76ab21edbd84ff9d98933ef22e7a264611

The full results

Open any test to see its individual checks and p-values, using the same numbering and precision as the original logs.

Frequency (monobit) · 1 passed

Are zeros and ones reasonably balanced?

0: 0.874457 · Passed

Block frequency · 1 passed

Does that balance hold across smaller sections?

0: 0.186022 · Passed

Cumulative sums · 2 passed

Does the sequence drift too far toward zeros or ones?

0: 0.753323 · Passed
1: 0.890593 · Passed

Runs · 1 passed

Does it switch between zeros and ones as often as expected?

0: 0.108720 · Passed

Longest run of ones · 1 passed

Are long runs of ones within the expected range?

0: 0.462613 · Passed

Binary matrix rank · 1 passed

Does arranging the bits into grids reveal too much structure?

0: 0.884661 · Passed

Discrete Fourier transform · 1 passed

Are there unusually strong repeating cycles?

0: 0.620220 · Passed

Non-overlapping templates · 144 passed, 4 flagged

Do specific bit patterns occur too often or too rarely?

0: 0.488543 · Passed
1: 0.024912 · Passed
2: 0.956441 · Passed
3: 0.064138 · Passed
4: 0.762904 · Passed
5: 0.645609 · Passed
6: 0.877426 · Passed
7: 0.062192 · Passed
8: 0.657117 · Passed
9: 0.599571 · Passed
10: 0.831365 · Passed
11: 0.994429 · Passed
12: 0.685766 · Passed
13: 0.839431 · Passed
14: 0.140180 · Passed
15: 0.581135 · Passed
16: 0.260878 · Passed
17: 0.793349 · Passed
18: 0.170406 · Passed
19: 0.957863 · Passed
20: 0.669801 · Passed
21: 0.720717 · Passed
22: 0.385986 · Passed
23: 0.558315 · Passed
24: 0.284879 · Passed
25: 0.723856 · Passed
26: 0.608299 · Passed
27: 0.049763 · Passed
28: 0.255052 · Passed
29: 0.475974 · Passed
30: 0.152873 · Passed
31: 0.318727 · Passed
32: 0.194078 · Passed
33: 0.419182 · Passed
34: 0.573074 · Passed
35: 0.862910 · Passed
36: 0.678086 · Passed
37: 0.905982 · Passed
38: 0.053715 · Passed
39: 0.738536 · Passed
40: 0.114697 · Passed
41: 0.137630 · Passed
42: 0.603107 · Passed
43: 0.798974 · Passed
44: 0.493835 · Passed
45: 0.158049 · Passed
46: 0.543328 · Passed
47: 0.644779 · Passed
48: 0.773656 · Passed
49: 0.020469 · Passed
50: 0.056791 · Passed
51: 0.164637 · Passed
52: 0.063038 · Passed
53: 0.161315 · Passed
54: 0.067242 · Passed
55: 0.362401 · Passed
56: 0.187426 · Passed
57: 0.241129 · Passed
58: 0.668261 · Passed
59: 0.372514 · Passed
60: 0.647982 · Passed
61: 0.862440 · Passed
62: 0.676430 · Passed
63: 0.101853 · Passed
64: 0.777323 · Passed
65: 0.699900 · Passed
66: 0.924132 · Passed
67: 0.498153 · Passed
68: 0.335881 · Passed
69: 0.502265 · Passed
70: 0.211426 · Passed
71: 0.777323 · Passed
72: 0.115895 · Passed
73: 0.004054 · Flagged
74: 0.488543 · Passed
75: 0.858749 · Passed
76: 0.950257 · Passed
77: 0.756805 · Passed
78: 0.284017 · Passed
79: 0.006315 · Flagged
80: 0.703070 · Passed
81: 0.291605 · Passed
82: 0.123073 · Passed
83: 0.424378 · Passed
84: 0.456302 · Passed
85: 0.592041 · Passed
86: 0.657829 · Passed
87: 0.535657 · Passed
88: 0.436850 · Passed
89: 0.118131 · Passed
90: 0.063630 · Passed
91: 0.452269 · Passed
92: 0.020461 · Passed
93: 0.283626 · Passed
94: 0.831772 · Passed
95: 0.226946 · Passed
96: 0.421215 · Passed
97: 0.943310 · Passed
98: 0.625687 · Passed
99: 0.903340 · Passed
100: 0.118975 · Passed
101: 0.704595 · Passed
102: 0.147470 · Passed
103: 0.159131 · Passed
104: 0.541378 · Passed
105: 0.977977 · Passed
106: 0.273822 · Passed
107: 0.441122 · Passed
108: 0.161415 · Passed
109: 0.185512 · Passed
110: 0.454389 · Passed
111: 0.185736 · Passed
112: 0.047072 · Passed
113: 0.204537 · Passed
114: 0.398180 · Passed
115: 0.250605 · Passed
116: 0.199489 · Passed
117: 0.421419 · Passed
118: 0.491737 · Passed
119: 0.623555 · Passed
120: 0.112006 · Passed
121: 0.992162 · Passed
122: 0.590044 · Passed
123: 0.047407 · Passed
124: 0.960275 · Passed
125: 0.434671 · Passed
126: 0.205814 · Passed
127: 0.195127 · Passed
128: 0.224189 · Passed
129: 0.817444 · Passed
130: 0.138592 · Passed
131: 0.483826 · Passed
132: 0.004557 · Flagged
133: 0.076549 · Passed
134: 0.963382 · Passed
135: 0.380791 · Passed
136: 0.786912 · Passed
137: 0.665773 · Passed
138: 0.629005 · Passed
139: 0.060489 · Passed
140: 0.284800 · Passed
141: 0.195829 · Passed
142: 0.295132 · Passed
143: 0.686002 · Passed
144: 0.261319 · Passed
145: 0.669327 · Passed
146: 0.848525 · Passed
147: 0.004054 · Flagged

Overlapping templates · 1 passed

Do overlapping patterns occur as often as expected?

0: 0.828687 · Passed

Maurer's universal · 1 passed

Is the sequence unusually easy to compress?

0: 0.614901 · Passed

Approximate entropy · 1 passed

Do short patterns repeat more than expected?

0: 0.519719 · Passed

Random excursions · 8 passed

Does a walk through the sequence revisit positions as expected?

0: 0.219325 · Passed
1: 0.376025 · Passed
2: 0.302738 · Passed
3: 0.997984 · Passed
4: 0.638494 · Passed
5: 0.621859 · Passed
6: 0.962349 · Passed
7: 0.358395 · Passed

Random excursions variant · 18 passed

Are the total visits to those positions within the expected range?

0: 0.680865 · Passed
1: 0.570971 · Passed
2: 0.502285 · Passed
3: 0.527650 · Passed
4: 0.532993 · Passed
5: 0.558961 · Passed
6: 0.754835 · Passed
7: 0.988513 · Passed
8: 0.841870 · Passed
9: 0.600493 · Passed
10: 0.954074 · Passed
11: 0.928907 · Passed
12: 0.954901 · Passed
13: 0.822415 · Passed
14: 0.641088 · Passed
15: 0.538180 · Passed
16: 0.536487 · Passed
17: 0.390419 · Passed

Serial · 2 passed

Do different short bit patterns appear in balanced proportions?

0: 0.385271 · Passed
1: 0.046110 · Passed

Linear complexity · 1 passed

Can a simple linear rule describe the sequence too easily?

0: 0.663200 · Passed