Loading ...
Sorry, an error occurred while loading the content.

[Speed cubing group] Re: A scheme and searching for some moves

Expand Messages
  • Josef Jelinek
    Ok ok... it is completed, enjoy it :) For all found solutions see http://rubikscube.info/duncan/ There is also a discussion about improvements etc. The
    Message 1 of 6 , Mar 1, 2004
    • 0 Attachment
      Ok ok... it is completed, enjoy it :)

      For all found solutions see http://rubikscube.info/duncan/
      There is also a discussion about improvements etc.
      The computation was quite fast (several minutes for all sequences).
      The manual processing was then slower (evening part) :).

      Some comments to the previously found sequences:
      - most of short sequences were optimal and it is double-checked now.
      - the original sequence #6 seems bad (first time I thougth that
      my program has a bug, because it found longer sequence, but
      then I checked the original one and it seems to be the
      unflipped/unflipped case).
      - all sequences are at most 11 face-turns long

      It would be nice if you discuss any further progress to see
      how the method is going.
      Contact me (gloom@...) if you want other sequences.
      However, the previous steps are quite well known and
      the last step can be solved in any optimal cube solver
      (reid's, kociemba's, mine...). So it could be useful for
      further experiments and method improval/change.

      Let me know, if my work was useful..

      Best regards,

      Josef

      PS: If anyone in this group would like to get the solution
      to a similar problem, let me know...

      --- In speedsolvingrubikscube@yahoogroups.com, "Duncan Dicks"
      <duncan@d...> wrote:
      > Josef,
      > That would be fantastic. An explanation of the cases and what I
      need is
      > below but if it is hard to follow or you want it in a different
      notation or
      > something let me know and I will try to send whatever you need.
      >
      > The way I am solving has the LL as the F face and the first
      (completed
      > layer) as the B face. At the point I am trying to solve I have
      the whole of
      > B solved and the UL, LD and RD edges solved.
      >
      > The final middle layer edge is currently sitting in LF in an
      orientation
      > such that when it goes to UR, L goes to U and F goes to R.
      >
      > At the same time I want to orient the corner in the UFL spot.
      > I also want to take the edges currently at UF and RF and place
      them and
      > orient them in LF and UF (that is either side of the newly
      oriented UFL
      > corner). It doesn't matter which of the two edges goes to which
      place.
      >
      > The twelve cases are defined by the current orientation of the UFL
      corner (3
      > possibilities) and the current orientation of the UF and RF edges
      (2
      > possibilities each (3x2x2=12). In table form I have the following:
      >
      > UFL
      UF RF
      > Solution
      > 1 Untwisted Unflipped
      Unflipped
      > R'F'RFUFU' (7,7)
      > 2 Untwisted Unflipped
      Flipped
      > R'FRU'RUR' (7,7)
      > 3 Untwisted Flipped
      Unflipped
      > L'UB'R'URBLU' (9,9)
      > 4 Untwisted Flipped
      Flipped
      > L'URUR'F'U'FL (9,9)
      > 5 Clockwise Unflipped
      Unflipped
      > D'L2B'UBL2D (7,9)
      > 6 Clockwise Unflipped
      Flipped
      > L2B'UBL2DF'D' (8,10)
      > 7 Clockwise Flipped
      Unflipped
      > No solution yet
      > 8 Clockwise Flipped
      Flipped
      > No solution yet
      > 9 AntiClockwise Unflipped
      Unflipped
      > R2FL'F2RF'LR'FL'F2RF'LFR (16,19) !!! too long!
      > 10 AntiClockwise Unflipped
      Flipped
      > No solution yet
      > 11 AntiClockwise Flipped
      Unflipped
      > FRFR'DBR'B'D' (9,9)
      > 12 AntiClockwise Flipped
      Flipped
      > R'F'RUF'U'F2R'FRF'UFU' (14,15) also too long
      >
      > May have my clockwise and anticlockwises the wrong way round!
      Does this
      > make sense to you?
      >
      > Eagerly awaiting a response (been working on this for a few weeks
      off and
      > on!).
      >
      > Duncan
      >
      >
      >
      >
      >
      >
      > ----- Original Message -----
      > From: "Josef Jelinek" <gloom@e...>
      > To: <speedsolvingrubikscube@yahoogroups.com>
      > Sent: Thursday, February 26, 2004 4:30 PM
      > Subject: [Speed cubing group] Re: A scheme and searching for some
      moves
      >
      >
      > > Hi,
      > >
      > > I have written a program based on my (now quite old) ACube 2.6
      > > program. I can be used to search for either optimal or
      > > suboptimal sequences for uncompletely specified cube.
      > > You can specify positions of only some cubies as well as
      > > their orientations, etc.
      > > It can also find optimal sequences in three different
      > > metrics (quarter, half, and slice-turn).
      > > It can also ignore specified leading and trailing move
      > > (U/U'/U2 in this case).
      > > Is is quite fast as it uses a lot of pruning tables etc.
      > >
      > > However, it is now text only and without documentation.
      > > Until I finish the program to publish it, you can send
      > > me cube configurations that you want to solve (while
      > > ignoring some orientations and/or permutations of the
      > > last layer in this case) and I will (try to) solve them
      > > and give you all optimal results...
      > >
      > > I used this program to find sequences for Waterman's method
      > > and (although waterman stated that they were checked for
      > > optimality up to 13 moves) found many shorter sequences
      > > (up to 2 turns safe).
      > >
      > > Regards,
      > >
      > > Josef
      > >
      > > --- In speedsolvingrubikscube@yahoogroups.com, "duncandicks"
      > > <duncan@d...> wrote:
      > > > Hi,
      > > > I am searching for some moves to help towards a reduced 2 look
      > > LL.
      > > > Ultimately I hope it'll get me to one look - but we all know by
      > > now
      > > > that that is always harder than you think its going to be.
      > > >
      > > > My first step is to put the last edge of the middle layer in
      and
      > > at
      > > > the same time orient two edges and a corner between them. The
      > > > tricky bit is arranging their position at the same time. They
      > > don't
      > > > have to be properly positioned wrt their colours just next to
      each
      > > > other so that you get a block of three in a corner that is
      > > properly
      > > > oriented. I have nine of the twelve possible cases. Three of
      > > them
      > > > found using one of the cube-solving applets. Trouble is that
      I am
      > > > now up to a minimum of 11 moves and its a bit unmanageable.
      > > >
      > > > Ultimately my goal is to start using these moves on my second
      to
      > > > last middle edge and then I need 30 to 40 more moves (I think
      its
      > > a
      > > > while since i came up with this) to put the last middle edge in
      > > > while retaining the structure I have and flipping the last two
      > > edges
      > > > and another corner. This would leave a LL with two corners
      > > > unoriented and nothing positioned which should be manageable in
      > > one.
      > > >
      > > > When I had it clear in my head I figured it would only save 4-8
      > > > moves on a normal fridrich F2L and a 2-look LL but its the
      > > principle
      > > > of the thing!
      > > >
      > > > Anyway the point of the post is to ask whether anyone has any
      > > views
      > > > on the scheme and whether anyone has any brilliant ideas for
      > > finding
      > > > those extra moves - is there stuff out there that will help.
      I'm
      > > > usually pretty good at working stuff out in RL but the cube
      applet
      > > > has been useful so far - it just seems to be at its limit at 11
      > > > moves.
      > > >
      > > > Happy to share what I have so far if anyone is interested.
      > > >
      > > > Duncan
      > >
      > >
      > >
      > >
      > >
      > > Yahoo! Groups Links
      > >
      > >
      > >
      > >
      > >
      > >
    • Gilles Roux
      ... Wow, Josef, are you doing all this free of charge? :-) Everybody wants to find his original method and put his name on it, so you may receive a lot of
      Message 2 of 6 , Mar 1, 2004
      • 0 Attachment
        --- In speedsolvingrubikscube@yahoogroups.com, "Josef Jelinek"
        <gloom@e...> wrote:
        > Ok ok... it is completed, enjoy it :)
        > [...]

        Wow, Josef, are you doing all this free of charge? :-)

        Everybody wants to find his original method and put his name on it, so
        you may receive a lot of work.

        There are many slightly different approaches based on F2L, the problem
        is they don't seem faster than Jessica's.

        I'm a bit surprised noone proposed something like this:
        - cross
        - 2 F2L pairs
        - 1 F2L pair + orient remaining edges (*)
        - 1 F2L pair + orient LL corners (**)
        - LL permutation

        It looks like another petrofridrich hybrid. Just in case you are
        interested in searching for (*) and (**) sequences ;-)

        Gilles.
      • Josef Jelinek
        Yes, it is free of charge, but noone should blame me if I reject a request (there was not such a case anyway) ;). I do it for fun and to improve my program
        Message 3 of 6 , Mar 1, 2004
        • 0 Attachment
          Yes, it is free of charge, but noone should blame me if
          I reject a request (there was not such a case anyway) ;).
          I do it for fun and to improve my program also for free.
          (However, any kind of support is of course welcome :))

          Is somebody willing to make some contribution to my program
          (mainly GUI and documentation - program is in Java now)?

          I do not think many people are going to ask me for help.

          Your proposed method looks interesting. However, the third pair
          + orienting edges is not as straightforward as it may seem, because
          you still have the last F2L edge unsolved and it may result
          in increasing the number of sequences and slower recognition.
          The similar problems apply to the last pair + orienting all corners.

          If you prepare the situations for (*) and (**) (as Duncan did)
          I can try to get some results.
          (Even pointing a bad way can be useful to followers :)).

          Regards,

          Josef

          PS: I do not know If fixed, but several years ago I tried to
          optimize Fridrich's sequences to place corner-edge pairs for F2L
          and found an improvement (I do not remember more details).
          Has someone already checked and/or optimized them.

          --- In speedsolvingrubikscube@yahoogroups.com, "Gilles Roux"
          <grrroux@f...> wrote:
          > --- In speedsolvingrubikscube@yahoogroups.com, "Josef Jelinek"
          > <gloom@e...> wrote:
          > > Ok ok... it is completed, enjoy it :)
          > > [...]
          >
          > Wow, Josef, are you doing all this free of charge? :-)
          >
          > Everybody wants to find his original method and put his name on
          it, so
          > you may receive a lot of work.
          >
          > There are many slightly different approaches based on F2L, the
          problem
          > is they don't seem faster than Jessica's.
          >
          > I'm a bit surprised noone proposed something like this:
          > - cross
          > - 2 F2L pairs
          > - 1 F2L pair + orient remaining edges (*)
          > - 1 F2L pair + orient LL corners (**)
          > - LL permutation
          >
          > It looks like another petrofridrich hybrid. Just in case you are
          > interested in searching for (*) and (**) sequences ;-)
          >
          > Gilles.
        • Duncan Dicks
          Many many thanks Josef. I have yet to go through it all but it looks very easy to follow and the results are great. I will certainly be working on the method
          Message 4 of 6 , Mar 1, 2004
          • 0 Attachment
            Many many thanks Josef. I have yet to go through it all but it looks very
            easy to follow and the results are great. I will certainly be working on
            the method in the future and will let you know how it is progressing. If
            the next step starts to look feasible to move a step closer to a one look LL
            I shall certainly request some more algorithms - hope you will give up your
            time to help out again!

            Duncan
            ----- Original Message -----
            From: "Josef Jelinek" <gloom@...>
            To: <speedsolvingrubikscube@yahoogroups.com>
            Sent: Monday, March 01, 2004 10:23 AM
            Subject: [Speed cubing group] Re: A scheme and searching for some moves


            > Ok ok... it is completed, enjoy it :)
            >
            > For all found solutions see http://rubikscube.info/duncan/
            > There is also a discussion about improvements etc.
            > The computation was quite fast (several minutes for all sequences).
            > The manual processing was then slower (evening part) :).
            >
            > Some comments to the previously found sequences:
            > - most of short sequences were optimal and it is double-checked now.
            > - the original sequence #6 seems bad (first time I thougth that
            > my program has a bug, because it found longer sequence, but
            > then I checked the original one and it seems to be the
            > unflipped/unflipped case).
            > - all sequences are at most 11 face-turns long
            >
            > It would be nice if you discuss any further progress to see
            > how the method is going.
            > Contact me (gloom@...) if you want other sequences.
            > However, the previous steps are quite well known and
            > the last step can be solved in any optimal cube solver
            > (reid's, kociemba's, mine...). So it could be useful for
            > further experiments and method improval/change.
            >
            > Let me know, if my work was useful..
            >
            > Best regards,
            >
            > Josef
            >
            > PS: If anyone in this group would like to get the solution
            > to a similar problem, let me know...
            >
            > --- In speedsolvingrubikscube@yahoogroups.com, "Duncan Dicks"
            > <duncan@d...> wrote:
            > > Josef,
            > > That would be fantastic. An explanation of the cases and what I
            > need is
            > > below but if it is hard to follow or you want it in a different
            > notation or
            > > something let me know and I will try to send whatever you need.
            > >
            > > The way I am solving has the LL as the F face and the first
            > (completed
            > > layer) as the B face. At the point I am trying to solve I have
            > the whole of
            > > B solved and the UL, LD and RD edges solved.
            > >
            > > The final middle layer edge is currently sitting in LF in an
            > orientation
            > > such that when it goes to UR, L goes to U and F goes to R.
            > >
            > > At the same time I want to orient the corner in the UFL spot.
            > > I also want to take the edges currently at UF and RF and place
            > them and
            > > orient them in LF and UF (that is either side of the newly
            > oriented UFL
            > > corner). It doesn't matter which of the two edges goes to which
            > place.
            > >
            > > The twelve cases are defined by the current orientation of the UFL
            > corner (3
            > > possibilities) and the current orientation of the UF and RF edges
            > (2
            > > possibilities each (3x2x2=12). In table form I have the following:
            > >
            > > UFL
            > UF RF
            > > Solution
            > > 1 Untwisted Unflipped
            > Unflipped
            > > R'F'RFUFU' (7,7)
            > > 2 Untwisted Unflipped
            > Flipped
            > > R'FRU'RUR' (7,7)
            > > 3 Untwisted Flipped
            > Unflipped
            > > L'UB'R'URBLU' (9,9)
            > > 4 Untwisted Flipped
            > Flipped
            > > L'URUR'F'U'FL (9,9)
            > > 5 Clockwise Unflipped
            > Unflipped
            > > D'L2B'UBL2D (7,9)
            > > 6 Clockwise Unflipped
            > Flipped
            > > L2B'UBL2DF'D' (8,10)
            > > 7 Clockwise Flipped
            > Unflipped
            > > No solution yet
            > > 8 Clockwise Flipped
            > Flipped
            > > No solution yet
            > > 9 AntiClockwise Unflipped
            > Unflipped
            > > R2FL'F2RF'LR'FL'F2RF'LFR (16,19) !!! too long!
            > > 10 AntiClockwise Unflipped
            > Flipped
            > > No solution yet
            > > 11 AntiClockwise Flipped
            > Unflipped
            > > FRFR'DBR'B'D' (9,9)
            > > 12 AntiClockwise Flipped
            > Flipped
            > > R'F'RUF'U'F2R'FRF'UFU' (14,15) also too long
            > >
            > > May have my clockwise and anticlockwises the wrong way round!
            > Does this
            > > make sense to you?
            > >
            > > Eagerly awaiting a response (been working on this for a few weeks
            > off and
            > > on!).
            > >
            > > Duncan
            > >
            > >
            > >
            > >
            > >
            > >
            > > ----- Original Message -----
            > > From: "Josef Jelinek" <gloom@e...>
            > > To: <speedsolvingrubikscube@yahoogroups.com>
            > > Sent: Thursday, February 26, 2004 4:30 PM
            > > Subject: [Speed cubing group] Re: A scheme and searching for some
            > moves
            > >
            > >
            > > > Hi,
            > > >
            > > > I have written a program based on my (now quite old) ACube 2.6
            > > > program. I can be used to search for either optimal or
            > > > suboptimal sequences for uncompletely specified cube.
            > > > You can specify positions of only some cubies as well as
            > > > their orientations, etc.
            > > > It can also find optimal sequences in three different
            > > > metrics (quarter, half, and slice-turn).
            > > > It can also ignore specified leading and trailing move
            > > > (U/U'/U2 in this case).
            > > > Is is quite fast as it uses a lot of pruning tables etc.
            > > >
            > > > However, it is now text only and without documentation.
            > > > Until I finish the program to publish it, you can send
            > > > me cube configurations that you want to solve (while
            > > > ignoring some orientations and/or permutations of the
            > > > last layer in this case) and I will (try to) solve them
            > > > and give you all optimal results...
            > > >
            > > > I used this program to find sequences for Waterman's method
            > > > and (although waterman stated that they were checked for
            > > > optimality up to 13 moves) found many shorter sequences
            > > > (up to 2 turns safe).
            > > >
            > > > Regards,
            > > >
            > > > Josef
            > > >
            > > > --- In speedsolvingrubikscube@yahoogroups.com, "duncandicks"
            > > > <duncan@d...> wrote:
            > > > > Hi,
            > > > > I am searching for some moves to help towards a reduced 2 look
            > > > LL.
            > > > > Ultimately I hope it'll get me to one look - but we all know by
            > > > now
            > > > > that that is always harder than you think its going to be.
            > > > >
            > > > > My first step is to put the last edge of the middle layer in
            > and
            > > > at
            > > > > the same time orient two edges and a corner between them. The
            > > > > tricky bit is arranging their position at the same time. They
            > > > don't
            > > > > have to be properly positioned wrt their colours just next to
            > each
            > > > > other so that you get a block of three in a corner that is
            > > > properly
            > > > > oriented. I have nine of the twelve possible cases. Three of
            > > > them
            > > > > found using one of the cube-solving applets. Trouble is that
            > I am
            > > > > now up to a minimum of 11 moves and its a bit unmanageable.
            > > > >
            > > > > Ultimately my goal is to start using these moves on my second
            > to
            > > > > last middle edge and then I need 30 to 40 more moves (I think
            > its
            > > > a
            > > > > while since i came up with this) to put the last middle edge in
            > > > > while retaining the structure I have and flipping the last two
            > > > edges
            > > > > and another corner. This would leave a LL with two corners
            > > > > unoriented and nothing positioned which should be manageable in
            > > > one.
            > > > >
            > > > > When I had it clear in my head I figured it would only save 4-8
            > > > > moves on a normal fridrich F2L and a 2-look LL but its the
            > > > principle
            > > > > of the thing!
            > > > >
            > > > > Anyway the point of the post is to ask whether anyone has any
            > > > views
            > > > > on the scheme and whether anyone has any brilliant ideas for
            > > > finding
            > > > > those extra moves - is there stuff out there that will help.
            > I'm
            > > > > usually pretty good at working stuff out in RL but the cube
            > applet
            > > > > has been useful so far - it just seems to be at its limit at 11
            > > > > moves.
            > > > >
            > > > > Happy to share what I have so far if anyone is interested.
            > > > >
            > > > > Duncan
            > > >
            > > >
            > > >
            > > >
            > > >
            > > > Yahoo! Groups Links
            > > >
            > > >
            > > >
            > > >
            > > >
            > > >
            >
            >
            >
            >
            >
            > Yahoo! Groups Links
            >
            >
            >
            >
            >
            >
          • gregvdyke
            ... Hi Have you worked out the number of cases for **? there are 27 orientations and severall permutations, giving well over a hundred cases... I tried once: I
            Message 5 of 6 , Mar 1, 2004
            • 0 Attachment
              >
              > I'm a bit surprised noone proposed something like this:
              > - cross
              > - 2 F2L pairs
              > - 1 F2L pair + orient remaining edges (*)
              > - 1 F2L pair + orient LL corners (**)
              > - LL permutation
              >
              > It looks like another petrofridrich hybrid. Just in case you are
              > interested in searching for (*) and (**) sequences ;-)
              >
              > Gilles.

              Hi

              Have you worked out the number of cases for **? there are 27
              orientations and severall permutations, giving well over a hundred
              cases... I tried once: I usually got between 11 and 13 turns with
              ron's solver, which compared with 7 + 7 with petrus pair + fridrich
              orientation: that's almost never a worthwhile trade.

              I think the best evolution is in knowing how to be lucky: If there are
              two petrus algs for the pair, of which one orients all the corners,
              then you've saved yourself 7 moves: that has to be 2-3 seconds in
              speedcubing...

              greg
            Your message has been successfully submitted and would be delivered to recipients shortly.