#### CreativeCubing

##### Member
Heise is incredibly fun, and I find it somewhat useful to know for fewest moves (very rare to submit a whole Heise solve for fewest moves, but the principles from it definitely can help, and using it for starts can get you pretty tolerable averages, even if they rarely result in winning results), but I know of no one who has ever gotten really fast with Heise. It typically takes me at least a minute to complete a full Heise solve; I average 20 seconds with CFOP. I know there are plenty of people who can sub-20 Heise, but most of them are probably twice as fast with CFOP.
Thank you for the information!

##### Member
Is Heise really all that efficient? What if you did steps 1 and 2 of Heise the same, and then inserted the F2L pair while maintaining EO, and then did ZBLL to complete? Seems like that should have a lower movecount.

Here's an attempt at some estimates of average movecount from F2L-1+EO cubestate using rough guesses.

Heise
Step 3: I have no idea, but if it's more than 10 moves, I would seriously doubt Heise's claim to being super efficient.
Step 4: ~8 (idk how commonly conjugates are needed, but 8 should be a lower bound if I understand commutators correctly)

Heise-a
LS: 6
ZBLL: 12 HTM (https://www.speedsolving.com/wiki/index.php/ZBLL)
= ~18 HTM?

Given that the only step I had hard numbers for was ZBLL, I tried to bias my estimates against Heise-a being better, if I was unsuccessful in that endeavor, so be it, though I'd appreciate being made aware of it so that I may be less unsuccessful the next time I try something like this.

Also, I'm not saying that either this or Heise is a good speedsolving method (nor the contrary), just trying to sort out whether or not Heise's reputation of being extremely efficient is deserved.

PS: Does anyone have numbers for average Heise movecounts? I assume it would be a tad lower than LLOB (41-45 (https://docs.google.com/document/d/1gs3THtRU5UCckKcM_5zjjm5RkJmGxHAJUnv6ta4lJhw/edit)), and LMCF (41-45 (https://drive.google.com/file/d/0B2QnZ3uD6I8kNkpHSURSbzluc2s/view)).

#### 2180161

##### Member
If Heise as a method should average approximately 40 moves (according to the wiki) and the LSLL (as I'll call it for now) takes approximately 9 moves on average for the commutator, and approximately 11 for the blocks, I would say that would be about 20 moves for the F2L-1, which would be about 2 moves less.

The thing with Heise's efficiency is that in a speedsolve, of course it won't average 40. In reality, it might be around 50. I ran Heise with a set block through HARCS and got about 39 moves per solve. Now this seems very low, until you realize that in HARCS, Roux/ZZ/Petrus all also get around 39 moves on average, despite the best solvers using those usually are high 40's low 50's.

##### Member
If Heise as a method should average approximately 40 moves (according to the wiki) and the LSLL (as I'll call it for now) takes approximately 9 moves on average for the commutator, and approximately 11 for the blocks, I would say that would be about 20 moves for the F2L-1, which would be about 2 moves less.

The thing with Heise's efficiency is that in a speedsolve, of course it won't average 40. In reality, it might be around 50. I ran Heise with a set block through HARCS and got about 39 moves per solve. Now this seems very low, until you realize that in HARCS, Roux/ZZ/Petrus all also get around 39 moves on average, despite the best solvers using those usually are high 40's low 50's.
Thanks for the response. I guess the gap between a lot of the methods in terms of raw average movecount is fairly close. I still want to learn Heise eventually, but if one can use another method with better ergonomics and lookahead and recognition that doesn't lose too much in terms of movecount for speedsolving, I guess that really does sink Heise in that department.

##### Member
If Heise as a method should average approximately 40 moves (according to the wiki) and the LSLL (as I'll call it for now) takes approximately 9 moves on average for the commutator, and approximately 11 for the blocks, I would say that would be about 20 moves for the F2L-1, which would be about 2 moves less.

The thing with Heise's efficiency is that in a speedsolve, of course it won't average 40. In reality, it might be around 50. I ran Heise with a set block through HARCS and got about 39 moves per solve. Now this seems very low, until you realize that in HARCS, Roux/ZZ/Petrus all also get around 39 moves on average, despite the best solvers using those usually are high 40's low 50's.
How did you get harcs to run heise?

#### mprimesarefun

##### Member
I've been trying to learn the Heise method for FMC purposes for a while now, and I get most of it, but 5E2C is what really gets me, and I don't know how exactly to get good solutions with it. My blockbuilding is also kind of crap even though I use Roux, so if anyone could direct me towards some good resources (besides Ryan's website, which I already know about), that would be amazing.

#### Mike Hughey

Staff member
I try to use Heise a lot for FMC, but to be honest, I'm pretty sloppy about it (and generally don't get that good results). But for what it's worth, I generally only try to do 5E instead of 5E2C, and then hope that what I get is good - either the best case where 5E2C luckily happens, or a case where my remaining 5 corners are nice to solve. 5E2C has always seemed pretty difficult to me.

#### mprimesarefun

##### Member
Thanks man. I know edges first and comms and stuff like that pretty well, so I should be set, I guess.

#### Mike Hughey

Staff member
For me, when I use Heise for FMC, it's all about trying to get a good setup to 5 or 3 corners, and I also check for easy 4th pair/OLL/3 corners if there's no nice 5 edge solution. If I'm trying to go with Heise, I try as many starts as possible and count the moves to 5 or 3 corners, looking for as many as I can get before about 40 minutes are up. Then I take the best one and try to find good insertions for it. I used that strategy when I got my one decent mean of 3 (29 moves) in official competition. With Heise, it's usually possible to find 8 or 10 decent 2x2x2 pseudo blocks to start with, and then go from there and take the best one.

#### Drotitis

##### Member
Hi

I have been lurking in the speedcubing community for a couple of years, mostly focusing on sovling my cubes not so much speed. I am now trying to learn speedcubing with roux and heise as my main methods since I love using intuitive moves to solve the cube.

I feel though that with step 3 of heise I often get stuck with 2 corner and 2 edges that are not solved corectly instead of 3 corners. I can often understand what caused the mistanke but I dont understand how to proceed.

I find it hard to use communtators, since either I would try to solve the edges or the corners first. But when i try to do it step wise I end up with the same problem. Anyone have any tips to how you can solve this state using commutators?

##### Member
Obligatory heise is too complicated to speedsolve really (Assuming vanilla heise- there's speedheise variants but i haven't looked into those for a while)

That aside, have a look at some of the 3bld "parity" algs and use them to solve 2e2c or similar (j perm or similar is common but there might be faster ones for your case)

#### Tao Yu

##### Member
Hi

I have been lurking in the speedcubing community for a couple of years, mostly focusing on sovling my cubes not so much speed. I am now trying to learn speedcubing with roux and heise as my main methods since I love using intuitive moves to solve the cube.

I feel though that with step 3 of heise I often get stuck with 2 corner and 2 edges that are not solved corectly instead of 3 corners. I can often understand what caused the mistanke but I dont understand how to proceed.

I find it hard to use communtators, since either I would try to solve the edges or the corners first. But when i try to do it step wise I end up with the same problem. Anyone have any tips to how you can solve this state using commutators?
Try and find some algs for these cases that you can understand in terms of commutators and just set up to them

for example:

J perm = [R U R' F': [R U R' U', l']] U'
T perm = [F R U' R': J perm]

#### brododragon

##### Member
I've been learning Heise from Ryan Heise’s guide and I almost have it down except for corner twist cases. I’ve learned commutators from J Perm but can't seem to find a way to solve something like this:
(Both corners are permuted but incorrectly oriented).

#### Mike Hughey

Staff member
For pure Heise, you probably want to avoid this situation as much as possible. But sometimes of course it can come up to where you're on your last 3 corners, and this is what you get. Ryan Heise's site says this:

Another, slightly less elegant way to solve this is to just perform any convenient 8 move 3-cycle involving these 2 corners and any third corner (perhaps the one between them). Then you'll have a nice ordinary 3 corner cycle (which might be more than 8 moves, but probably not more than 9) you can do to solve all 3.

#### brododragon

##### Member
I'm learning Heise (basically just from the wiki article), how do I perform EO while pairing up the Heise blocks? I'm also having trouble with solving the edges + 2 corners while finishing F2L
Learn from Heise's website. For EO, just solve the blocks and then do EO like your would from F2L-1 in Petrus. For the edges+two corners, solve exactly two LL edges, misplace the third, and make sure the LL edge is in the LL slot. Then, just do a simple 3-move keyhole-type insert. Then, do a comm for the 2 corners. Finally, do L3C. All of this is explained way better on Heise's website, so just use that.

#### brododragon

##### Member
For pure Heise, you probably want to avoid this situation as much as possible. But sometimes of course it can come up to where you're on your last 3 corners, and this is what you get. Ryan Heise's site says this:

Another, slightly less elegant way to solve this is to just perform any convenient 8 move 3-cycle involving these 2 corners and any third corner (perhaps the one between them). Then you'll have a nice ordinary 3 corner cycle (which might be more than 8 moves, but probably not more than 9) you can do to solve all 3.
Here, just tried it out, and wondering if this can be improved upon or if I've found the optimal comm: https://alg.cubing.net/?setup=(R-_D-_R_D)2_U-_(R-_D-_R_D)4_U&alg=x-_[R_U_R-_U-_R_U_R-,_D]

My main worry is for the orienting of the first corner it was very similar to R' D' R D.

#### Tao Yu

##### Member
The move optimal comm is pretty cool and is quite different to the comms taught on Heise's website.

[F' R D2 R' F, U2]

Essentially swapping two corners and swapping them back in a different way.

#### brododragon

##### Member
Essentially swapping two corners and swapping them back in a different way.
Thank you! Also, how do you do a 2c2c/2e2e comm?

#### ProStar

##### Member
Thank you! Also, how do you do a 2c2c/2e2e comm?
Solve two edges/corners, with the third affected edge/corner simply moving places with one comm, then use another comm to solve the remaining 3 edges/corners. Having 2c2c/2e2e is similar to hitting your buffer in BLD, where you have to now start a new cycle/com