While looking at prime numbers between 1 to 1000 I noticed that the number of non-primes between prime numbers are almost always also prime or 1. In other words, if we take the prime gap, $$ g_n=p_{n+1}-p_n, $$ for prime numbers larger than 2, and subtract 1 from it, we almost always get a prime number or 1. In the cases (between 1 to 1000) where $g_n-1$ is not equal to a prime number or 1 it is instead equal to 9. Below is a figure I made to better explain:

enter image description here

What is the reason that the "jump length" (i.e. numbers in red or blue above) between primes are either 1, 9 or also prime when looking at 1 to 1000?

  • 1
    $\begingroup$ Because the gap between consecutive primes is always an even number, except for $3-2=1$ $\endgroup$ – rtybase Sep 25 '18 at 21:21
  • 1
    $\begingroup$ It does, $p_{282}=1831, p_{283}=1847$ and then again almost immediately, $p_{295}=1933,p_{296}=1949$. $\endgroup$ – lulu Sep 25 '18 at 21:34
  • 13
    $\begingroup$ General note: it's tempting, but risky, to read too much into numerical computations for such small numbers. In this case, there are many small primes, so any random-y list of small odd numbers is going to contain lots of primes. To study anything like this, I'd want to look at much larger primes. $\endgroup$ – lulu Sep 25 '18 at 21:38
  • 2
    $\begingroup$ You can assume Cramer's conjecture: $g_n \approx O\left((\log{p_n})^2\right)$ and draw one or two conclusions, like $\left((\log(1000))^2\right)\approx 48$, consider only even gaps, that's $24$ and there are $14$ primes (or "jumps") between $1$ and $48$ (excluding $2$), thus the chance is $\frac{14}{24}=\frac{7}{12}>\frac{1}{2}$. $\endgroup$ – rtybase Sep 25 '18 at 21:58
  • 1
    $\begingroup$ Absolutely agree with @lulu on this one. "Law of small numbers." A cursory look at oeis.org/A046933 seems to confirm your hypothesis. But look at the "b-file": it's still mostly primes, but 9, 15, 25 make more appearances the further along you go. $\endgroup$ – Robert Soupe Sep 26 '18 at 4:13

This phenomenon was studied by Odlyzko, Rubinstein and Wolf (with whom I have written a paper on primes).

Definition: An integer $d$ is called a jumping champion for a given $x$ if d is the most common gap between consecutive primes up to $x$.

They postulated that most common gap between primes or the jumping champions are the product of prime numbers i.e. the first jumping champion is $2$. The after some time, $2*3 = 6$ takes over as the next jumping champion. At around $1.74*10^{35}$, the number $2*3*5 = 30$ takes over as the next jumping champion. At around $10^{425}$, the number $2*3*5*7 = 210$ takes over as the next jumping champion and so on.

This means that per your definitions, if we subtract $1$ from the gap between primes then the most common results would be $1$ initially. This would be overtaken by $5$ which would in turn be overtaken by $29$ and this would be overtaken by $209$ as you go higher up the number line.

So it is a co-incidence that $5, 29$ are prime but $209 = 11*19$ is not. This means if you looked large enough to reach the range of $209$, you would actually have the opposite conclusion based on experimental evidence that gap between prime minus $1$ is often a non prime.

Equivalent question: With the above arguments, your question is equivalent to

How often is the primorial number $\cdots p - 1$ a prime?


  1. Odlyzko, A.; Rubinstein, M.; and Wolf, M. "Jumping Champions." Experiment. Math. 8, 107-118, 1999.
| cite | improve this answer | |
  • 2
    $\begingroup$ Nice answer! Do you know whether any result of this form (classifying jumping champions) has been proven, or if it's all just conjecture so far? $\endgroup$ – Carl Schildkraut Sep 26 '18 at 6:47
  • 2
    $\begingroup$ As with most problems with primes, we have way more conjectures on jumping champions than theorems. However you can find a few theorems on them by Goldston in this paper. arxiv.org/pdf/1102.4879.pdf $\endgroup$ – Nilotpal Kanti Sinha Sep 26 '18 at 6:50

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.