• Welcome to the Speedsolving.com, home of the web's largest puzzle community!
    You are currently viewing our forum as a guest which gives you limited access to join discussions and access our other features.

    Registration is fast, simple and absolutely free so please, join our community of 40,000+ people from around the world today!

    If you are already a member, simply login to hide this message and begin participating in the community!

HD vs EG: 2x2 method showdown

Sue Doenim

Member
Joined
Nov 9, 2016
Messages
448
HD or Higgs-Demars, is a method for 2x2 speedsolving, comparable to EG. It breaks the solve into two steps: orient all corners and force a V separation case, and solve the rest with one of 36 algs. Here are movecount comparisons for EG and HD.

HD:
V+CO: 3.90 moves (from 100 solves by myself)
NLL: 8.13 moves (from current algs, factoring in probability)
Total (AUFs included): 13.53 moves
Moves: number of cases, chance, cumulative chance
0: 1, 0.56%, 0.56%
1: 0, 0%, 0.56%
2: 0, 0%, 0.56%
3: 3, 5%, 5.56%
4: 2, 4.44%, 10%
5: 3, 6.67%, 16.67%
6: 3, 6.67%, 23.33%
7: 7, 24.44%, 47.78%
8: 5, 11.11%, 58.89%
9: 15, 33.33%, 92.22%
10: 1, 2.22%, 94.44%
12: 1, 1.11%, 95.56%
13: 1, 4.44%, 100%

EG:
Face (optimal): 3.80 (unsure)
EG case: 9.54 (factoring in probability, using algs from Chris Olsen's website)
Total (AUFs included): 14.84 moves

Other arguments:

Alg count:
HD: 41
EG: 128

One looking:
HD: Track permutation of 8 pieces
EG: Track permutation of 8 pieces, orientation of 4

Ergonomics:
HD: First step decent, second step good, rather undeveloped algs
EG: First step decent, second step great, developed algs

Add ons:
HD: possibly more NLLs but for non V separation cases
EG: TCLL, LEG, possibly TEG if you are crazy

Sorry for not making this way earlier. Credit to myself, @Thermex, @Shiv3r, and @Neuro for the method, also thanks a little bit to @efattah for helping with movecount. HD is way cool, I really hope some people prove its viability with fast times. Note that these statistics are not guaranteed to be perfectly accurate, but they should be. Also note that the method called HD here is known more specifically as HD-G, or HD-Guimond, as the first step is similar to that of Guimond. Please post any more arguments or aspects I missed, and feel free to debate against what I have here.
 

Reed Merrill

Member
Joined
May 14, 2017
Messages
111
Location
Canada
Sounds sweet! Next time I feel super motivated to learn CLL I might go for the way lower alg count and do HD instead. I'm not anything close to a good 2x2 solver, and too many algs is mainly what keeps me away from trying get a lot better.
 

efattah

Member
Joined
Feb 14, 2016
Messages
711
I was pretty excited about HD and even considered learning it (despite already known EG1 & CLL), but at the present moment I find the HD algorithms slow and non-fingertricky. I believe they need to be expanded and improved, possibly choosing longer algorithms that are faster and being less obsessed with movecount. Look at some of the CFOP PLL's that are 19-21 moves that some people can execute in sub-1. HD is a great method, it just needs more community participation to develop the algorithms. The HD algorithms (currently) have a lot of half turns (too many F2's in particular). I am losing faith in STM as a means of measuring the length of an algorithm. Personally I have been switching to QSTM (M = 1 turn, M2 = 2 turns, F2 = 2 turns, etc.) although I think something like 1.5-STM is better (M2 = 1.5 turns, F2 = 1.5 turns). Counting an F2 or B2 as 1 turn I think is quite deceptive when you end up comparing something like R U R' U' with R' F2 R2 F2 and calling them the same length.
 

Thermex

Member
Joined
Mar 21, 2017
Messages
188
Location
The Milky Way
HD or Higgs-Demars, is a method for 2x2 speedsolving, comparable to EG. It breaks the solve into two steps: orient all corners and force a V separation case, and solve the rest with one of 36 algs. Here are movecount comparisons for EG and HD.

HD:
V+CO: 3.90 moves (from 100 solves by myself)
NLL: 8.13 moves (from current algs, factoring in probability)
Total (AUFs included): 13.53 moves
Moves: number of cases, chance, cumulative chance
0: 1, 0.56%, 0.56%
1: 0, 0%, 0.56%
2: 0, 0%, 0.56%
3: 3, 5%, 5.56%
4: 2, 4.44%, 10%
5: 3, 6.67%, 16.67%
6: 3, 6.67%, 23.33%
7: 7, 24.44%, 47.78%
8: 5, 11.11%, 58.89%
9: 15, 33.33%, 92.22%
10: 1, 2.22%, 94.44%
12: 1, 1.11%, 95.56%
13: 1, 4.44%, 100%

EG:
Face (optimal): 3.80 (unsure)
EG case: 9.54 (factoring in probability, using algs from Chris Olsen's website)
Total (AUFs included): 14.84 moves

Other arguments:

Alg count:
HD: 41
EG: 128

One looking:
HD: Track permutation of 8 pieces
EG: Track permutation of 8 pieces, orientation of 4

Ergonomics:
HD: First step decent, second step good, rather undeveloped algs
EG: First step decent, second step great, developed algs

Add ons:
HD: possibly more NLLs but for non V separation cases
EG: TCLL, LEG, possibly TEG if you are crazy

Sorry for not making this way earlier. Credit to myself, @Thermex, @Shiv3r, and @Neuro for the method, also thanks a little bit to @efattah for helping with movecount. HD is way cool, I really hope some people prove its viability with fast times. Note that these statistics are not guaranteed to be perfectly accurate, but they should be. Also note that the method called HD here is known more specifically as HD-G, or HD-Guimond, as the first step is similar to that of Guimond. Please post any more arguments or aspects I missed, and feel free to debate against what I have here.
It sounds like HD-G (I think we should come up with a better name for it) could really change the viability of the method; though I wish you could go a little more in depth on how you're doing the first two steps for your statistics. Otherwise I really like how this is shaking out, I think times very comparable to EG are possible with more developed algs. (I'll try to add some of this stuff to the wiki page I created).
I was pretty excited about HD and even considered learning it (despite already known EG1 & CLL), but at the present moment I find the HD algorithms slow and non-fingertricky. I believe they need to be expanded and improved, possibly choosing longer algorithms that are faster and being less obsessed with movecount. Look at some of the CFOP PLL's that are 19-21 moves that some people can execute in sub-1. HD is a great method, it just needs more community participation to develop the algorithms. The HD algorithms (currently) have a lot of half turns (too many F2's in particular). I am losing faith in STM as a means of measuring the length of an algorithm. Personally I have been switching to QSTM (M = 1 turn, M2 = 2 turns, F2 = 2 turns, etc.) although I think something like 1.5-STM is better (M2 = 1.5 turns, F2 = 1.5 turns). Counting an F2 or B2 as 1 turn I think is quite deceptive when you end up comparing something like R U R' U' with R' F2 R2 F2 and calling them the same length.
I agree with this; I feel like it's one of the main flaws of the method. I wish there was some sort of movecount-system that could measure the ergonomy and speed of an algorithm. Something that counts certain strings of moves with certain values (ex: F, U = 3 moves, R, U = 1.5 moves) that could be used to measure the true potential of a method. I plan on maybe going back and editing some of the NLL's this winter break.
 

Sue Doenim

Member
Joined
Nov 9, 2016
Messages
448
I'm working on algs, so I'll post alternatives here for you guys to judge. Tell me what needs work, what's good, and what's best. If you have a better alg, please share!
V permuted set:
F opp.
(U2) R2 U R2 U' R2 (original)
R opp.
y' (U2) R2 U' R2 U R2 (original)
F bar
(U') F2 U R2 U' B2 D R2 (original)
y' (U') R U' R' U2 R U R' D R2
y (U') R2 U R2 U' R2 D R2
(U') R' U' L U2 R' U R U2
x' F' R' F' R U2 R' U R U2
(U') x' R' F' R U2 R' U R U2
(U') x' L' U' L U2 R' U R U2
x' F' L' U' L U2 R' U R U2
R bar
(U) R2 U' F2 U R2 U' B2 (original)
(U) R2 D' R2 U R2 U' R2
x' U2 R' U' R U2 L' U R
x' U2 R' U' R U2 R' F R
x' U2 R' U' R U2 L' U L
Solved (This one's really difficult, I'd appreciate some help with it.)
R U R' F2 R F' R U R2 F2 (original)
y' L U' L D R U' R2 D L U2 L
R2 U R2 U' R2 D R2 U' R2 U R2
R U R' F2 R U R2 F R U' F
Diagonal
R2 U' B2 U2 R2 U' F2 U' R2 (original)
R2 U' F2 D' R2 D R2 D R2
R2 D' R2 D' R2 U R2 U R2
y2 R' U' R D' R2 D R' U R
F R F' U R2 U' F R' F'
y R U R' D R2 D' R U' R'
 

efattah

Member
Joined
Feb 14, 2016
Messages
711
For the solved case you can use some set up moves then an Ortega permute
U' R2 U' F2 R2 U2 + (Ortega PBL=) R2 U' B2 U2 R2 U' R2 [13]
It looks slow but if you finger trick it properly it has no regrips and is really fast, the F2 is done with the right ring finger. However the one given, R U R' F2 R U R2 F R U' F [11], if finger tricked properly is also pretty good. Does require regrip though.
 

Neuro

Member
Joined
Dec 23, 2016
Messages
597
Sure! I'm using CN HD-Guimond

1: 13
x' y2 R U R' U' R U2 R'
y2 U R2 F2 U' R2 U2

2: 14
x' y U2 R' U R'
x2 R' U R' F2 R F' R' F2 R2 U'

3: 14
x' U R' U2 R U2 R
F' U R' U2 R U' F U2

4: Scrambles for 3/4 are the same, using this instead: U F' U F' U2 R' U F U'. 15
z' y2 D F R2 U' F
y2 R2 U' B2 U R2 U' R2 D' B2 U2

5: 8
z' F' U2 F'
U2 R2 U' R2 U

MEAN: (13+14+14+15+8)/5= 12.8 avg movecount
Disregarding highest/lowest: 13.66 avg movecount
Not very indicative of "true" average movecount, but I just got a 12.8 mean of 5 with HD-Guimond. Used standard algs to do this and didn't remove cancellations (past V+CO.) The 8 move solve is pretty cool, shows what can happen if you're lucky. Scrambles and OG post on The 2x2 Example Solve Thread here on the forums
 

Duncan Bannon

Member
Joined
Sep 9, 2017
Messages
1,817
Location
Here
I did a little average game against him and with Pure CLL no EG got 14 mean and 14.33 Ao5. Ill do a quick Ao5 here with full EG and see what I can get. @Neuro Let do it again right here. Ill use EG. I know the alg count is very different but lets go.

Straight from CS timer.

1. R2 U2 R' F2 R' U' F U R'
2.R' U' F2 R F' U R2 U R2
3.F' U R' F2 R2 U R' F2 U'
4.F R U' R F' U2 F R U2
5.U2 R' U2 R U R2 U F2 R2 U2
 

Duncan Bannon

Member
Joined
Sep 9, 2017
Messages
1,817
Location
Here
1.
Z’, X’
U’,F’, U2, F’
U2, R’, U’,R, U’, R’, U2, R, U - 13

2.
Z
U, R’, U, R, U’, R’
R' F' U R' F R2 U2’ R' U R U2 - 17

3.
X, Y2
R,U,R’
U F' R U R' U' R U R2' F' R, U’ -15

4.
X, Y’
U’,R’
F' L F L' U2 L' U2 L U’ - 11

5.
X2, Y’
R2, U’, R’
U2 R U' R' F U2 R2 F' R F' R - 14

Didn’t want to keep you waiting, so I did this pretty quick. Me and Neuro are going to do some in a PM, May tell you guys how it goes.

Ao5- 14
Mean of 5- 14
 

Neuro

Member
Joined
Dec 23, 2016
Messages
597
Gonna try and really abuse the efficiency in these solves while still using the normal algs

1: 9
z y U R' U' R
U R2 U R2 y U2

2: 12
z U R U2 R y
U R2 U R2 U' F2 R2 U2

3: 10
z U2 R' U R'
y2 U R2 U' F2 R2 U

4: 11
x y' U' R U' R2' F R F'
R2 U R2 U2

5: 13
U' R U2 R'
x' U L U' R2 U L' U F' U2 x

Mean of 3: 11 moves
Mean of 5: 11 moves

This is indicative of what the method is capable of; although it may not be truly realistic by a human in 15 secs. It's fun in context of trying to get the most efficient solves though ;) When I do the Ao50 I'll provide solutions most indicative of a speedsolve
 
Top