20210425, 16:30  #23 
May 2018
233 Posts 
Very neat! What does the sequence of numbers at the bottom mean?
1, 2, 4, 1, 1, 4, 0, 0, 0, 0, ... 
20210425, 19:24  #24 
Just call me Henry
"David"
Sep 2007
Cambridge (GMT/BST)
61×97 Posts 

20210426, 16:07  #25 
Dec 2008
you know...around...
677 Posts 
I've found a couple of gaps for the list.
Option 1) Submitting the full numbers with about 3420 digits each Option 2) Submitting the numbers in the shortened form m*p#/dx, but then they would still have > 1600 characters each Option 3) I post the two CRT offset numbers c and corresponding m here, where the gap offsets would then be = m*7963#+c Which option is best? 
20210427, 19:11  #26  
Mar 2021
53 Posts 
Quote:
Option 1 is easiest for me but I could probably handle any of them. What approach did you use for finding the offsets? 

20210427, 19:36  #27  
Mar 2021
110101_{2} Posts 
Quote:
Code:
Max 15/20 20 8.16E08 8.92E04 21 2.28E08 5.11E04 22 6.29E09 2.89E04 23 1.72E09 1.61E04 24 4.70E10 8.87E05 25 1.27E10 4.83E05 26 3.43E11 2.60E05 27 9.21E12 1.39E05 28 2.47E12 7.30E06 29 6.59E13 3.81E06 30 1.76E13 1.97E06 31 4.68E14 1.01E06 32 1.25E14 5.11E07 33 3.31E15 2.57E07 34 8.79E16 1.28E07 35 2.33E16 6.35E08 36 6.20E17 3.12E08 37 1.64E17 1.52E08 38 4.35E18 7.31E09 39 1.15E18 3.49E09 40 3.04E19 1.65E09 

20210427, 21:03  #28  
Dec 2008
you know...around...
677 Posts 
Quote:
I took one of If needed, I also have this Pari oneliner to turn a list of CRT offsets into a number o (mod p#), with input r as a vector: Code:
r=readvec("offset_res.txt");o=0;p=1;q=1;for(i=1,#r,q=nextprime(q+1);o=o+((q(r[i]+o)%q)*bezout(p%q,q)[1])%q*p;p=p*q);print(o) Last fiddled with by mart_r on 20210427 at 21:17 

20210427, 23:47  #29 
Mar 2021
53 Posts 
Are those all new records? Sorry, I thought you had some new CRT offsets you wanted me to look at.

20210428, 08:33  #30 
Dec 2008
you know...around...
677 Posts 
Last time I checked, and that was the day before yesterday, these were 18 improvements in merit and 3 new entries.

20210502, 07:53  #31  
Just call me Henry
"David"
Sep 2007
Cambridge (GMT/BST)
61·97 Posts 
Quote:
I think it makes sense that the average gap case would be bad. It punishes central candidates so strongly it ignores wider candidates. I would imagine for extremely low merit it would be excellent. I am a little confused by why 15 and 20 were mirrors of each other. I suppose it makes some sense as P(gapstarting value) doesn't actually change. It is just the amount of times that they were cumulated that changed. I used old slow code for implementing probabilities as it was simpler. I think I can add this to my faster code which will allow optimizing two at once again. I also have Robert's idea to try. Time is the issue... 

20210505, 18:16  #32 
Mar 2021
110101_{2} Posts 
The CRT offsets listed in this thread so far are all improvements on 7993#/13#. These work best for merits in roughly 2535 range. The first missing gaps are in the 15 merit range. You would probably be better off using 7993#/210 for these. A quick optimization gives 43*7993#/210. I think you will find about 50% more gaps with merit of 15 using this.

20210505, 23:17  #33 
Mar 2021
53_{10} Posts 
I took another look at the CRT offsets. I was surprised by how much worse the CRT_6727 did at finding 220k gaps even though it had significantly fewer remaining candidates. We have CRT offsets with 6776, 6751, and 6727 remaining candidates. There is a 4th that was optimized for a merit of 15 that has 6840 remaining candidates. These are all variations on 7993#/13# so I added a fifth to the test, 2347*7993#/13# with 6850 remaining candidates.
The ranking based on probability of finding a gap>220k is OPT_15, 7993#/13#, 6776, 6751, 6727 with 6727 significantly worse than the others. The rankings are almost the inverse of the number of candidates remaining. The table below shows the number of candidates remaining for gaps from 20k to 440k and the probability of finding a gap>220k. Code:
Gap Size /13# CRT_6776 CRT_6751 CRT_6727 OPT_15 20000 570 530 519 476 560 40000 907 883 871 834 891 60000 1344 1315 1306 1288 1327 80000 1841 1834 1832 1822 1823 100000 2386 2388 2400 2398 2372 120000 3053 3026 3020 3039 3034 140000 3744 3723 3717 3734 3727 160000 4468 4441 4435 4464 4449 180000 5227 5194 5182 5230 5205 200000 6012 5980 5962 5988 6000 220000 6850 6776 6751 6727 6840 240000 7703 7717 7709 7738 7694 260000 8591 8626 8639 8748 8583 280000 9472 9560 9599 9768 9461 300000 10411 10530 10593 10804 10407 320000 11357 11511 11589 11856 11358 340000 12329 12507 12601 12906 12335 360000 13306 13522 13636 13988 13317 380000 14303 14528 14662 15075 14320 400000 15319 15584 15733 16154 15342 420000 16344 16657 16821 17268 16372 440000 17391 17716 17894 18390 17423 prob > 220000 7.13E06 6.92E06 6.83E06 5.99E06 7.30E06 
Thread Tools  
Similar Threads  
Thread  Thread Starter  Forum  Replies  Last Post 
Primorial offsets  robert44444uk  Prime Gap Searches  7  20181129 08:40 