<div dir="ltr"><div class="gmail_quote"><div dir="ltr">Two years ago in this forum, Andrew Johnson wrote (based on his searches) "There is no bobs-only exact 7-part."<div><br></div><div>Graham <br><br><div class="gmail_quote"><div dir="ltr" class="gmail_attr">---------- Forwarded message ---------<br>From: <strong class="gmail_sendername" dir="auto">Andrew Johnson</strong> <span dir="auto"><<a href="mailto:andrew_johnson@uk.ibm.com" target="_blank">andrew_johnson@uk.ibm.com</a>></span><br>Date: Fri, 24 Jun 2022 at 08:28<br>Subject: Re: [r-t] Bobs-only Stedman Triples - 12 complete B-block peals<br>To: <a href="mailto:ringing-theory@bellringers.org" target="_blank">ringing-theory@bellringers.org</a> <<a href="mailto:ringing-theory@bellringers.org" target="_blank">ringing-theory@bellringers.org</a>><br></div><br><br>>Andrew Johnson writes,<br>
> > These are the last of the bobs-only peals for now; I have some <br>
> > unpublished peals but they are not significantly different from the <br>
> > published ones.<br>
> > The collection is here: <a href="https://complib.org/collection/11309" rel="noreferrer" target="_blank">https://complib.org/collection/11309</a><br>
>This is amazing work, thank you Andrew. How comprehensive is this latest survey of yours? In other words, is<br>
>it possible there are significant peals (in terms of interest of construction, number of bobs, ease of calling,<br>
>etc) which are as yet undiscovered, or do you believe you have covered most possibilities?<br>
><br>
>MBD<br>
These are ad-hoc searches, and are not comprehensive. A lot of them were obtained by taking an existing peal I had found, keeping some parts fixed and varying other parts.<br>
<br>
For example, sometimes I took one peal, bobbed all the Q-sets to get an odd number of round blocks, and kept the paired B-blocks as fixed because paired B-blocks are useful as there are a 6 ways each pair can be expressed, which helps linking everything together, and also if they were variable a search could come up with any of those 6 forms which wouldn't be significantly different, but would just flood the results.<br>
<br>
I then took the remaining sixes and searched to see how many sets of round blocks I could obtain. I might also impose other limits like the total number of bobs (because otherwise the search might just turn up 84 B-blocks), and number of complete B-blocks, and also that any Q-sets should be bobbed (to restrict equivalent solutions). My searches couldn't easily filter results of odd numbers of round blocks vs even numbers so I had to do that once I had say 1000 results. If I got less than 1000 when the search terminated then I knew I had found all the blocks subject to the conditions. I could then try to link the blocks into a peal, but there would be a free choice of 6 for each of the paired B-blocks. Sometimes if I wanted to reduce the number of bobs I would choose the instances of the paired B-blocks to maximise the number of available Q-sets and then try to link the blocks without changing the types of the B-blocks. I could also try other restrictions such as not allowing 7 bobs in a row, and finding 8 plains in a row.<br>
<br>
The number of bobs now ranges from 438 to 456 (existing 10-part peals) and 528 to 711 (one-part peals). With 84 B-blocks there are 840 bobs, so 123 omits on Q-sets could link those to 2 blocks with 717 blocks. An upper limit on the number of bobs therefore seems to be 714, but that would require a 2 blocks to 1 link with 3 omits, or a 4 blocks to 1 with 6 omits etc. which seems unlikely, so perhaps 711 bobs is the limit.<br>
<br>
The 528 bob peal <a href="https://complib.org/composition/86761" rel="noreferrer" target="_blank">https://complib.org/composition/86761</a> had 4 Q-sets which could not be plained without splitting it back into multiple blocks, and there was another peal <a href="https://complib.org/composition/92761" rel="noreferrer" target="_blank">https://complib.org/composition/92761</a> where if all the Q-sets were plained there were 501 bobs (but multiple blocks), so perhaps a one-part peal using B-blocks can be found with less than 528 bobs but I'm guessing to achieve 500 bobs would be hard.<br>
<br>
I found several peals with 8 plains in a row but they can't be arranged to put all at the start or all at the end of a peal with a normal start; it needs to start or end with a bob, so at best with my peals you can end with 8 plains and a bob, or 7 plains. Perhaps this can be improved on.<br>
<br>
<a href="https://complib.org/composition/84174" rel="noreferrer" target="_blank">https://complib.org/composition/84174</a> is interesting as all the bobs come in multiples of 2 (2,4 or 6) except for 18 single bobs, but it is a one-part. The treble does 0,2 or 6 bobs behind. Compare with the exact 3-part from 2017 <a href="https://complib.org/composition/36006" rel="noreferrer" target="_blank">https://complib.org/composition/36006</a>, which has 9 runs of 1 bob, 3 runs of 3 bobs and 15 runs of 5 bobs and the 7 does either 2 or 6 bobs behind.<br>
<br>
Among the things that didn't work for me:<br>
7-part peals<br>
There is no bobs-only exact 7-part. There are multiple sets of blocks with a 7-part group with an odd number [5,21,35] of round blocks, but it doesn't seem possible to link them with bobs or omits. Here are some peals with singles to link the blocks<br>
<br>
35 round blocks, 49 complete B-blocks, signature 63:14+105<br>
<br>
3124567QS-----PPP--------PP----P---P---------P--------P---------PP---------P---*1(7)<br>
4125367SQ-------P--P------P--*1(7)<br>
4236175SQ----------*1(7)<br>
5316472SQ----------*1(7)<br>
1627354SQ----------*1(7)<br>
<a href="https://complib.org/composition/48931" rel="noreferrer" target="_blank">https://complib.org/composition/48931</a> 651 bobs, 42 singles<br>
<br>
5 round blocks, 14 complete B-blocks, signature 84:5+252<br>
<br>
3124567QS-----P-------PP-P-----PP--PP---------P--PP-----P-PPP----PP-PP-P-P-P-----P---*7(1)<br>
6421375QS--PP----PP---P---P--------PP*7(1)<br>
4125367SQ-----P*7(1)<br>
1536472QSP-----*7(1)<br>
7165324QSP-PP*7(1)<br>
<a href="https://complib.org/composition/48932" rel="noreferrer" target="_blank">https://complib.org/composition/48932</a> 560 bobs, 56 singles<br>
<br>
35 round blocks, 49 complete B-blocks, signature 63:14+105<br>
<br>
1234567QS--P--P--------P---------PP-------PP-P---------P--P---PPP--------P---------P-----*1(7)<br>
4267351SQ----------*1(7)<br>
1637452QS----------*1(7)<br>
5316472SQ----------*1(7)<br>
3617254SQ----------*1(7)<br>
<a href="https://complib.org/composition/48954" rel="noreferrer" target="_blank">https://complib.org/composition/48954</a> 651 bobs, 42 singles<br>
<br>
6-part peals<br>
There can't be an exact 6-part bobs-only peal, but the group can generate blocks which link in pairs to give 3 round blocks. When I was working with Dr Michael Haythorpe he ran my program in 2013 and we found the following sets of an odd number of blocks, shown in file odd.txt. Unfortunately I cannot see a way of linking the blocks with just bobs. That's an especial pity because the set giving 5 round blocks has only 402 bobs.<br>
<br>
We also found some single blocks, shown in file single.txt, which when expanded for a peal formed 2 or 6 round blocks (but unfortunately not 3). This shows that the 6-part graph is Hamiltonian. Example peals using singles are here:<br>
<br>
<a href="https://complib.org/composition/49158" rel="noreferrer" target="_blank">https://complib.org/composition/49158</a> 540 bobs, 6 singles<br>
<a href="https://complib.org/composition/49177" rel="noreferrer" target="_blank">https://complib.org/composition/49177</a> 540 bobs, 6 singles<br>
<br>
5-part peals<br>
As already published, there just one instance (+rotations/reversal) of a single block (actually from the 10-part group, so made of two identical sections), but it links to itself and cannot be linked by Q-sets of bobs or omits. <br>
<a href="https://complib.org/composition/28949" rel="noreferrer" target="_blank">https://complib.org/composition/28949</a> 450 bobs, 10 singles<br>
There are sets of multiple blocks with odd numbers of blocks when expanded to all the rows (e.g. the original 10-part peals) which can be linked with Q-sets of bobs or omits to provide the existing 10-part peals.<br>
<br>
The magic blocks appear in a 5-part search, together with the remaining B-blocks, but cannot be linked without changing the type of some of the B-blocks, so there isn't a 5-part based magic block peal. There might be other sets of an odd number of blocks - I haven't found them yet.<br>
<br>
4-part peals - group [6.26]<br>
Not much success here - but the search space gets larger as the groups get smaller.<br>
The graph is Hamiltonian, but the example I found, from a 20-part search) divides into 4 1-part blocks. This doesn't stop there being an exact 4-part out there.<br>
3124567QSP--PP--P---------P--PPP---------P--P--P--P*5(4)<br>
Generators [1543267, 1425376] group order 4<br>
<br>
4-part peals - group [4.04]<br>
Not directly useful for a bobs-only peal.<br>
This is the only group for Stedman Triples where I don't know whether the reduced graph is Hamiltonian i.e. can a search just turn up one block<br>
<br>
4-part peals - group [6.35]<br>
Not directly useful for a bobs-only peal. Is a Hamiltonian graph (i.e. search can find one block), but then need 4 of those blocks cover all the sixes.<br>
<br>
4-part peals - group [4.05m]<br>
An in course subgroup of the Thurstans' 20-part out of course group.<br>
Not directly useful for a bobs-only peal. Is a Hamiltonian graph.<br>
<br>
4-part peals - group [4.06m]<br>
Two pairs swapping independently<br>
Not directly useful for a bobs-only peal. Is Hamiltonian though, giving this nice peal with singles just swapping 12 or 34 in the parts:<br>
<a href="https://complib.org/composition/66787" rel="noreferrer" target="_blank">https://complib.org/composition/66787</a> 486 bobs, 6 singles<br>
<br>
3-part peals<br>
Lots of peals - but too big to search completely.<br>
I think I did complete a search choosing 4 pairs of complementary B-blocks and one block with the remaining sixes (200 sixes), anyway it was how I found the 42 and 48 complete B-block exact 3-parts.<br>
<br>
2-part peals - group [4.07]<br>
Nothing new after the 2012 peals<br>
<br>
2-part peals - group [2.01]<br>
One pair swapping. Not directly useful for a bobs-only peal.<br>
Is Hamiltonian though, and gives a peal which looks like Slack's two-part, but comes round half-way if you replace the single with a bob.<br>
<a href="https://complib.org/composition/66793" rel="noreferrer" target="_blank">https://complib.org/composition/66793</a> 508 bobs, 2 singles<br>
Slack's peal is actually made of 4 blocks and does not come round if you replace the single with a bob or a plain and has more bobs (596 vs 508), so mine is cleaner, but not of the same historical interest.<br>
<a href="https://complib.org/composition/29011" rel="noreferrer" target="_blank">https://complib.org/composition/29011</a> 596 bobs, 2 singles<br>
<br>
Andrew Johnson<br>
--<br>
<br>
<br>
<br>
<br>
<br>
<br>
<br>
<br>
<br>
Unless otherwise stated above:<br>
<br>
IBM United Kingdom Limited<br>
Registered in England and Wales with number 741598<br>
Registered office: PO Box 41, North Harbour, Portsmouth, Hants. PO6 3AU<br>
_______________________________________________<br>
ringing-theory mailing list<br>
<a href="mailto:ringing-theory@bellringers.org" target="_blank">ringing-theory@bellringers.org</a><br>
<a href="https://bellringers.org/listinfo/ringing-theory" rel="noreferrer" target="_blank">https://bellringers.org/listinfo/ringing-theory</a><br>
</div></div></div>
</div></div>