factor missed by sieving

[Frodo42]

I think we had another discussion about something similar to this somewhere, but I can't find it so I'll start a new.

I just found this

P-1 found a factor in stage #2, B1=30000, B2=292500.
4847*2^7411287+1 has a factor: 25606348911511

This factor should have been found been found by sieving long ago.

Someone mentioned that the some version of the siever had missed a few factors, but this range is too low to have been sieved with the mentioned siever as I remember it.

The range it should have been found in is:
25600 25650 Netbrian

Could it be that there are more factors to find by resieving this range or is this just a single miss?

[Mystwalker]

Just tested it with proth_sieve 0.42 - factor is found.

How about resieving Netbrian's range to check if other factors are missed? I could do this quickly (< 1 day)...

[Frodo42]

Originally posted by Mystwalker
Just tested it with proth_sieve 0.42 - factor is found.

How about resieving Netbrian's range to check if other factors are missed? I could do this quickly (< 1 day)...

That would probably be a good idea, just go ahead and do that.

If there are more factors here it could be that Netbrians other ranges also should be checked.

[Mystwalker]

Ok, started computation...
ETA: 3 hours

[vjs]

I had a quick question about sieve vs P-1

If for example a k/n pair has a factor within the sieve range of 2^49-2^50 could this factor be missed by P-1???

I know that sieve and P-1 are different animals, sieve is basically trial division so if there is a factor in the range sieve will find it. But I'm not exactly sure how P-1 works if there is a factor in that range (2^49-2^50) is P-1 guarenteed to find the factor?

[Mystwalker]

P-1 will only find a factor if (P-1) - with P as the factor - is smooth, which means that (P-1) is (highly) composite and all of its factors are very small*.

*all factors have to be < B1, except one factor, which may be < B2.


ETA Update: still 2 hours...
no additional factors found yet

[Death]

no new factors or no factors at all?

[Mystwalker]

No new factors - but meanwhile, there are some:

25608226881791 | 10223*2^16324409+1
25618827138587 | 4847*2^10937991+1

2 of the results were new results and saved to the database.

[vjs]

all factors have to be < B1, except one factor, which may be < B2

Interesting, I didn't understand what the b1, b2 values meant perhaps this should be included in the p-1 reservations section.

So basically you want to keep b1 and b2 as high as posssible.

When I did my range I tryed to keep b1 low and b2 high...

Which was obviously the wrong thing to do (b1 low)... well at least I found one factor.

[vjs]

Netbrains ranges:

1800 - 1850 50 644 87 0 1177.076
4350 - 4399 49 253 38 0 1108.356
6350 - 6400 50 184 44 0 1175.531
7700 - 7750 50 149 33 0 1153.169
8900 - 8950 50 121 21 0 1081.705
9050 - 9100 50 104 25 0 945.777
9650 - 9700 50 120 17 0 1162.489
11610 - 11650 40 79 18 0 920.949
12230 - 12300 70 141 33 0 1732.428
13800 - 13850 50 66 12 0 913.728
15100 - 15150 50 57 19 0 864.624
15155 - 15200 45 63 18 0 958.733
15300 - 15350 50 76 18 0 1167.512
15850 - 15900 50 63 18 0 1002.767
16100 - 16150 50 65 18 0 1050.604
21230 - 21300 70 49 9 3 1044.181
22640 - 22700 60 31 4 4 703.487
25600 - 25650 50 32 9 3 822.362
26750 - 26800 50 15 8 3 403.778
27000 - 27050 50 12 4 5 325.054
27700 - 27750 50 37 5 0 1027.320
32000 - 32050 50 28 7 0 898.945
32630 - 32700 70 36 12 0 1179.682
39450 - 39500 50 33 6 0 1305.043
40800 - 40850 50 26 5 0 1063.494
42600 - 42650 50 20 5 1 854.580
42900 - 42950 50 13 4 3 559.666
43400 - 43450 50 16 5 2 8.684

Total 1454G

Your right some of them look interesting...

I'll start retest,

25600-25650 Mystwalker (complete) no additional through sieve
26750-26800 VJS
27000-27050 VJS

Or were you stating that you would retest all of them?

LMK I already started the range but I could stop, should have these finished by tommorrow

These also look interesting:

11610-11650
15100-15150
42900-42950
43400-43450

[Mystwalker]

Sieving complete, no additional previously-unknown factors found


Or were you stating that you would retest all of them?

I only meant that specific one. Maybe I'll retest others in the next days, but I wait until maybe some results show up and only after having reserved here.


25600-25650 Mystwalker (complete) no additional through sieve
26750-26800 VJS
27000-27050 VJS

You seem to have picked the most likely (read: "underpopulated") ranges. I did a quick check, including factor density:

Start End Size Factors Factors / G Corrected Factors / G (=Start*Factors/(10000*rangeSize))
1800 1850 50 644 12.88 2.32
4350 4399 49 253 5.16 2.25
6350 6400 50 184 3.68 2.34
7700 7750 50 149 2.98 2.29
8900 8950 50 121 2.42 2.15
9050 9100 50 104 2.08 1.88
9650 9700 50 120 2.4 2.32
11610 11650 40 79 1.98 2.29
12230 12300 70 141 2.01 2.46
13800 13850 50 66 1.32 1.82
15100 15150 50 57 1.14 1.72
15155 15200 45 63 1.4 2.12
15300 15350 50 76 1.52 2.33
15850 15900 50 63 1.26 2
16100 16150 50 65 1.3 2.09
21230 21300 70 49 0.7 1.49
22640 22700 60 31 0.52 1.17
25600 25650 50 32 0.64 1.64
26750 26800 50 15 0.3 0.8
27000 27050 50 12 0.24 0.65
27700 27750 50 37 0.74 2.05
32000 32050 50 28 0.56 1.79
32630 32700 70 36 0.51 1.68
39450 39500 50 33 0.66 2.6
40800 40850 50 26 0.52 2.12
42600 42650 50 20 0.4 1.7
42900 42950 50 13 0.26 1.12
43400 43450 50 16 0.32 1.39

Next interesting would be:

42900 42950 50 1.12
22640 22700 60 1.17
43400 43450 50 1.39
21230 21300 70 1.49
25600 25650 50 1.64
32630 32700 70 1.68
42600 42650 50 1.7
15100 15150 50 1.72
32000 32050 50 1.79
13800 13850 50 1.82
9050 9100 50 1.88

Btw.:
"25600-25650 Mystwalker (complete) no additional through sieve" <-- I found 2 additional factors, frodo 1, so there were 3

[vjs]

read: "underpopulated") ranges

Do you have a link???

Yes I did a quick "look see" at the numbers and those looked promising, thanks for your table.

I also looked at some other users and how many factors they recieved in near-by ranges to see if (0.3 f/G) was reasonable


<del>

But it looks like anything with a factors/G of <0.4 is questionable. But anything with factors/G >0.4 is probably O.K.

Perhaps this has been done already etc. I'll make up another table of questionable in a few minutes.

[vjs]

O.K. new table,
paramters for table
Only ranges <2^48
Only range sizes >50G
CF < ~1.2 (I used a value simlar to your CL)
CL - Confidence level - Instead of 10000 it's something smaller but same formula


Here is the table, it's note worthy to see that some popular names are included
mklasson for example
as well as :Pokes: frodo42

CL CF Fact/G Start Stop Range Fact Excl Dup Name
21.0 0.26 0.010 262400 262500 100 1 0 0 Troodon
40.0 0.50 0.160 31250 31300 50 8 0 0 jmblazek
40.6 0.51 0.020 254000 254100 100 2 0 0 alexr
40.8 0.51 0.020 254950 255000 50 1 0 0 Troodon
43.1 0.54 0.169 31870 32000 130 22 0 0 jmblazek
51.8 0.65 0.240 27000 27050 50 12 4 5 netbrian
53.4 0.67 0.060 111200 111250 50 3 0 0 rosebud
59.6 0.75 0.185 40300 40500 200 37 1 0 jmblazek
63.5 0.79 0.030 264500 264600 100 3 0 16 frodo42
64.2 0.80 0.300 26750 26800 50 15 8 3 netbrian
65.4 0.82 0.036 227000 227500 500 18 1 0 expmod8
66.6 0.83 0.110 75700 75800 100 11 5 0 Foobar
67.5 0.84 0.055 153300 153500 200 11 0 0 hc_grove
67.7 0.85 0.040 211500 211600 100 4 0 0 frodo42
69.0 0.86 0.032 269500 269750 250 8 13 277 royanee
73.0 0.91 0.060 152000 152100 100 6 0 0 Deoje362 (Deoje)
73.3 0.92 0.130 70500 70600 100 13 2 1 jimmy
79.2 0.99 0.180 55000 55050 50 9 4 0 Moothecow (Moo_the_co
81.0 1.01 0.050 202500 202600 100 5 0 0 Tully
81.6 1.02 0.060 170100 170300 200 12 4 216 Joe_O
82.8 1.03 0.456 22710 22800 90 41 6 3 cmprince
84.2 1.05 0.045 234000 234200 200 9 5 241 BigMajaX
85.0 1.06 0.071 148800 149500 700 50 8 0 expmod8
85.4 1.07 0.040 266850 267500 650 26 0 0 darkman
85.7 1.07 0.070 153000 153100 100 7 4 62 Deoje362 (Deoje)
88.0 1.10 0.400 27500 27600 100 40 0 0 jmblazek
88.3 1.10 0.280 39400 39450 50 14 6 114 mklasson
88.3 1.10 0.120 92000 92050 50 6 3 179 BigMajaX
89.2 1.12 0.260 42900 42950 50 13 4 3 netbrian
93.2 1.16 0.541 21520 21605 85 46 19 6 louie
93.6 1.17 0.517 22640 22700 60 31 4 4 netbrian
94.7 1.18 0.690 17150 17250 100 69 2 6 ceselb
95.5 1.19 0.280 42650 42700 50 14 1 1 mklasson
97.0 1.21 0.240 50500 50550 50 12 1 3 jimmy
97.1 1.21 0.053 229000 230000 1000 53 1 0 priwo
97.6 1.22 0.049 247250 248000 750 37 1 0 Mat67
97.7 1.22 0.070 174400 174500 100 7 7 171 jimmy
97.8 1.22 0.080 152800 152900 100 8 6 222 Mischa
98.4 1.23 0.057 215150 215500 350 20 0 0 darkman

Can someone modify this table so it looks good, don't know how to Code: etc.

I'm also not sure how accurate CL/CF is when it comes to larger p values >>281T and of course small ranges produce more error in the calculated CL/CF. So alot of ranges were not included in this table some obviously had small ranges ~10G and 2 or less factors, they ranked very high of course.

[Joe O]

As per usual, I will redo my own range:

81.6 1.02 0.060 170100 170300 200 12 4 216 Joe_O

[vjs]

Joe_O,

I wouldnt redo your range unless your thinking that there may be a problem there.
Also before we start distributing these we should infact make a double check thread.

You have a CL of ~80 with a fairly large p, I don't think there are missed factors there.

If you'd like to recheck a range do

31870-32000 jmblazek

It looks like a much greater offender than yours,

I'm going (31250-31300 jmblazek) already to see if the theory holds.

[Mystwalker]

Originally posted by vjs
Do you have a link???

Link? :confused:
I draw that conclusion from the table I offered afterwards...


But it looks like anything with a factors/G of <0.4 is questionable. But anything with factors/G >0.4 is probably O.K.

As factor density decreases, it's important to multiply the density with the range depth (start or end - I found out that it basically makes no real difference). So when a desity for depth x is acceptable, it's density/2 for 2*x...

[vjs]

link?

I just wanted to read how one came up with the =Start*Factors/(10000*rangeSize)

The table does draw its own conclusions pretty well.

Anyways I was thinking it should be something like start^1.05 or some other very small exponent b/c it does break down above 2^48.

But more importantly, I've checked most of the 27000-27050 last night before I lost power and havn't found any new factors in fact.txt.

However I did find about 400 new results in factexcl.txt from the 500 I submitted and 1 new result factrange.txt from the 17 of 44 verified.

I will continue with the following two ranges

31250-31300

and

26750-26800

But I don't think anything will be found there that is both new and important, but I will keep all informed.

I think this double checking thread is at least one reason why factexcl.txt should be included in submissions.

[Mystwalker]

Originally posted by vjs
I just wanted to read how one came up with the =Start*Factors/(10000*rangeSize)

It was just my very own spontaneous idea...

It started with Factors/rangeSize - but as we have to think of the decreasing factor density, I added the multiplier "Start" to even this out. And the 10000, well, that's only to get good-looking numbers... :D

[vjs]

31250-31300 Completed double check no new factors in fact.txt
27000-27050 Completed double check no new factors in fact.txt

Well I complete these two ranges found alot of new factors in factexcl.txt and factrange.txt but nothing in fact.txt.

Regardless I submitted all of these factors even tough they were dups.

I'm going to consider that the rest of these ranges are correct and not worth the time nor energy to redo, even though they look promissing.

Probably be better off doing supersecret double checks etc not double check sieve.

Hopefully this will bring this thread to a close.

[Frodo42]

OK I found another weird factor using P-1

23171537407129 | 10223*2^8341577+1

for one thing it should have been found by sieving long ago.

The other is that it doesn't seem to want to submit to result.txt it doesn't show up in the scores and I can't find it in results.txt.
It does get verified and the first time I submitted it, it did say it was added to the database.
It could be there is a few more factors.

Anyways I'm not sure this is worth resieving ... the factor should be found in this range

23170 23200 Troodon

I think (unless I counted the number of digits wrong here):rolleyes:

[vjs]

O.K. I'll resieve that range...

Please do not resieve any of the ranges less than 75T at this moment, or if you must please contact Joe_O or myself before doing so...

Also thanks Frodo, if any of you P1 guys find missing factors that should have been found through sieve let us know asap.

BTW is it worth making a double check sieve thread for any of these resieves???

Just to report reservations only...

Jason

23170 23200 Troodon [Double Check by VJS]

I'll have it done by tommorrow and report my findings.

[MikeH]

The other is that it doesn't seem to want to submit to result.txt it doesn't show up in the scores and I can't find it in results.txt. That's because it's too low to show up on the results.txt - only factors above 25T.

I'll manually add your find to my low results file, then you'll get credit. Unfortunately it's below 40T, so it won't score much. Sorry.

[vjs]

Mike, There is another factor in that range <20M

it's not pressing... Joe will send it to you.

[Frodo42]

Originally posted by MikeH
That's because it's too low to show up on the results.txt - only factors above 25T.

I'll manually add your find to my low results file, then you'll get credit. Unfortunately it's below 40T, so it won't score much. Sorry.

Well that doesn't matter to much, what is important to me is that is won't be handed out, but since I submitted it via. the form I hope that's been taken care of.
But thanks anyway.

[vjs]

23170 23200 Troodon [Double Check Completed by VJS] 1 extra factor at 19M plus frodos factor

[vjs]

This has come up in another thread so I would ask that nobody double check sieve ranges less than 26000, (26000G or 26T), consider them already checked or that someone is investigating them "already reserved", I will report progress...


0-26000G [Reserved or Completed]


I would also ask that anyone finding factors through P-1 in/within a range that has already been sieve report it here. I'm generally keeping tab's etc but you never know.