Message #1777

From: Matt Galla <mgalla@trinity.edu>
Subject: Re: [MC4D] Re: Ranking list for the 120-Cell
Date: Fri, 10 Jun 2011 20:11:20 -0500

Well yes I suppose you are right. Just remember that if my program says 100% complete then it really is 100% complete. So maybe an even more accurate way to calculate it is to have a diminishing background noise factor. Probably just linear from -0.8% at 0.8% complete and -0.0% at 100% complete.

In any case I didn’t start using it until I was beyond the 50% mark because I was curious to see how much I had left. In that range it can be considered quite accurate. +/- 0.4% is my guess.

-Matt Galla

Sent from my iPhone

On Jun 10, 2011, at 7:29 PM, Nan Ma <mananself@gmail.com> wrote:

> Hi Matt,
>
> You have an interesting point. For a fully scrambled 120 cell, your program will give a non zero percentage, because the colors of some stickers are correct by chance. A very coarse computation (incorrectly assuming statistical independence across different stickers) is that the 1/120 ~ 0.8% of the stickers are correct by chance. So 0.8% is the "background noise". If we subtract it from 1.1%, we get 0.3%, which is Ed’s original estimate. I think Ed’s estimate was very accurate.
>
> Nan
>
>
> On Fri, Jun 10, 2011 at 5:18 PM, Matthew Galla <mgalla@trinity.edu> wrote:
>
> Ed,
>
> I ran your log file through my program and these are the results:
> 2-Color pieces solved: 14 / 720
> 3-Color pieces solved: 8 / 1200
> 4-Color pieces solved: 6 / 600
> Total solved: 1.111 %
>
> The way this works is it counts the number of 2C stickers that are correct and divides by 2 (throwing away the remainder) to estimate the number of 2C pieces solved. It repeats for 3C and 4C, dividing by 3 and 4 respectfully to get the above estimations. Now it is possible that a couple of pieces happen to have one or two stickers correct, even though the pieces themselves are not in the right spots, and there is no detection for this case, so the numbers that appear above should be considered overestimates and the actual numbers may be slightly less than above, but certainly not bigger.
>
> It looks like you can claim a little more is done that you originally said! ;)
> Then again, maybe this isn’t accurate enough to claim much at all. What can be said is my program is more accurate the colser to solved you are, so it is best used for those cases. Anyways enjoy.
>
> If anyone else would like me to run their log file through the program, just let me know. Sorry Nan, but you are right, my program only works for the Magic120Cell program’s log files for now. :/
>
> -Matt Galla
>
> On Fri, Jun 10, 2011 at 3:02 PM, Eduard Baumann <baumann@mcnet.ch> wrote:
>
> [Attachment(s) from Eduard Baumann included below]
> 
>
> Hi Matt,
>
> Here is my 120cell_partial_solve log file for evaluation.
>
> Kind regards
> Ed
>
> —– Original Message —–
> From: Matt Galla
> To: <4D_Cubing@yahoogroups.com>
> Sent: Friday, June 10, 2011 9:36 PM
> Subject: Re: [MC4D] Re: Ranking list for the 120-Cell
>
>
> While I was still solving the 120Cell I wrote a short java program that gives the exact number of stickers in the correct spot and makes a guess as to how many of which type of piece is solved based on the stickers. It then estimates a percentage complete of the whole puzzle (# of correct PIECES / 2520). It is by no means complex or super sophisticated but it is set up and ready to read any Magic120Cell log file. If any of the partial solvers would like a more accurate estimation of their completion percentage I’d be happy to run your log file through the program and relay the output to you. Just send me an email with your most current log file attached.
>
> -Matt Galla
>
> Sent from my iPhone
>
> On Jun 10, 2011, at 2:21 PM, Nan Ma <mananself@gmail.com> wrote:
>
>>
>> Hi,
>>
>> Let me put 3% in it. That’s by no means accurate.
>>
>> 1. Noel Chalmers 100%
>> 2. Matt Galla 100%
>> 3. Adam Ford 100%
>> 4. Andrey Astrelin 10%
>> 5. Nan Ma 3%
>> 6. Eduard Baumann 0.3%
>>
>> Thanks.
>>
>> Nan
>>
>>
>> On Fri, Jun 10, 2011 at 3:57 AM, Eduard <baumann@mcnet.ch> wrote:
>>
>> Nan Ma: Please complete the list below!
>>
>> 120cell performers (including partial solves):
>>
>> Update 06/10/2011
>>
>>
>>
>> 1. Noel Chalmers 100%
>> 2. Matt Galla 100%
>> 3. Adam Ford 100%
>> 4. Andrey Astrelin 10%
>> 5. Nan Ma ??%
>>
>> 6. Eduard Baumann 0.3%
>>
>> — In 4D_Cubing@yahoogroups.com, "Eduard" <baumann@…> wrote:
>> >
>> > Update 03/13/2011
>> >
>> > 1. Noel Chalmers 100%
>> > 2. Matt Galla 100%
>> > 3. Adam Ford 100%
>> > 4. Andrey Astrelin 10%
>> > 5. Eduard Baumann 0.3%
>> >
>> > — In 4D_Cubing@yahoogroups.com, "Eduard" <baumann@> wrote:
>> > >
>> > > 1. Noel Chalmers 100%
>> > > 1. Matt Galla 100%
>> > > 3. Andrey Astrelin 10%
>> > > 4. Eduard Baumann 0.3%
>> > >
>> > > Others?
>> > >
>> >
>>
>>
>
>
>
>