kcl
Legend
Really? I need to re watch the video then haha it looked like he forced it...
Edit: yeah you're right. He only paused because of PLL recognition though.
How do top-cubers lookahead?
- Thread starter bundat
- Start date
-
- Tags
- f2l look ahead
- Status
Really? I need to re watch the video then haha it looked like he forced it...
Edit: yeah you're right. He only paused because of PLL recognition though.
Escher
Babby
Just thought I'd respond with some ideas to help you improve the elusive thing called lookahead.
First off, I've been consistently sub 10 for roughly 3 years now, at my peak I was low 8 avg 100, and after all that experience, I honestly don't think that the lookahead that 'pros' do and the lookahead that 12-13s+ cubers do is the same thing at all. There are a few comparisons I think work to describe how it feels. A good juggler can do it blindfolded because as soon as they've made the throw, as they know where the ball will land based on the feel of the throw. A piano player, once they have learnt a piece, can do it without sheet music. When an 'amateur' cuber such as yourself solves the cube, one is constantly checking the 'sheet music' to see where they'll go next. The sheet music in this analogy is pretty much the algorithms they know for f2l slots (whether they have learned intuitively or not). 'Real' lookahead has no checks or doubletakes, there is hardly any self-correction. A pro cuber looks at the state of the f2l and reads it like music - they have a rough idea of how the solve will go based on their 'peripheral vision' of the general patterns that are popping up, but they'll instinctually solve the 'notes' that are in front of them. There are few feelings quite like nailing a sub 4 f2l, because it feels like it happens on a strange balance between conscious decision making and subconscious reflexes.
If there is anything I can recommend for you to start emulating this, it would be to scramble a cube, solve the cross, and stare at the cube for a very long time. Identify each pair, and going off whatever seems nicest, imagine the pieces move in your head as you would solve them. Visualise where affected pieces will end up. Close your eyes, execute your pair, and check whether your intuition was correct - if not, why? What were the reasons for your mistake? Don't just notice that you have made one - really consider what the reasons were, and write it down. Alternately, solve the cross without checking any f2l pieces, if you can, and then as soon as you look at the cube time how long it takes you to identify each pair.
I just did an avg of 5 of that last exercise, times of which I consider pretty slow, though I'm quite exhausted:
You need to be so, so comfortable with f2l that you can execute a pair and be comfortable with predicting, instantly, how it will affect the pieces around it. One possible exercise would be to do the inverse of the pair algorithm on a solved cube, and check what happens to the pieces. It's quite hard to describe all of the things you could possibly do - one advantage to the hobby of cubing over others is the practically infinite number of different exercises you can create for yourself to try and reach a state of 'unthinkingness' while solving. You want to devote as few mental cycles as possible to whats in front of you, and do your best to sharpen your 'peripheral vision' of what is to come next. Where most intermediate cubers still devote so much time to staring at stickers they have already processed information for, you never want to be checking the same piece of information twice, whether that be a sticker or the arrangement of the 3rd pair while you do the first.
I guess the point of this ramble is really that lookahead isn't so much as a 'trick', or singular 'thing', rather it's the process of optimising your information management, and it begins when you no longer have to double-check the simplest operations.
First off, I've been consistently sub 10 for roughly 3 years now, at my peak I was low 8 avg 100, and after all that experience, I honestly don't think that the lookahead that 'pros' do and the lookahead that 12-13s+ cubers do is the same thing at all. There are a few comparisons I think work to describe how it feels. A good juggler can do it blindfolded because as soon as they've made the throw, as they know where the ball will land based on the feel of the throw. A piano player, once they have learnt a piece, can do it without sheet music. When an 'amateur' cuber such as yourself solves the cube, one is constantly checking the 'sheet music' to see where they'll go next. The sheet music in this analogy is pretty much the algorithms they know for f2l slots (whether they have learned intuitively or not). 'Real' lookahead has no checks or doubletakes, there is hardly any self-correction. A pro cuber looks at the state of the f2l and reads it like music - they have a rough idea of how the solve will go based on their 'peripheral vision' of the general patterns that are popping up, but they'll instinctually solve the 'notes' that are in front of them. There are few feelings quite like nailing a sub 4 f2l, because it feels like it happens on a strange balance between conscious decision making and subconscious reflexes.
If there is anything I can recommend for you to start emulating this, it would be to scramble a cube, solve the cross, and stare at the cube for a very long time. Identify each pair, and going off whatever seems nicest, imagine the pieces move in your head as you would solve them. Visualise where affected pieces will end up. Close your eyes, execute your pair, and check whether your intuition was correct - if not, why? What were the reasons for your mistake? Don't just notice that you have made one - really consider what the reasons were, and write it down. Alternately, solve the cross without checking any f2l pieces, if you can, and then as soon as you look at the cube time how long it takes you to identify each pair.
I just did an avg of 5 of that last exercise, times of which I consider pretty slow, though I'm quite exhausted:
Average of 5: 1.58
1. 1.68 L' D2 R' D2 F2 R' D2 B2 D2 L' R' U' F' L2 D' R2 D L2 B' D F
2. 1.62 F2 R' D2 R' U2 R2 B2 F2 R' F2 D2 B' F2 D B U' F L F2 D2 L2
3. 1.44 U F2 L2 D2 B2 R2 F2 U' R2 D' U' L F2 D2 F D' R2 B' L D F
4. (1.32) R2 D R2 B2 L2 R2 U' L2 U' L2 F2 L' D B' F L' B2 R2 F L'
5. (2.05) R U2 L U2 F2 L2 R U2 B2 L2 D2 F U B' D2 R' F' U2 L' F
1. 1.68 L' D2 R' D2 F2 R' D2 B2 D2 L' R' U' F' L2 D' R2 D L2 B' D F
2. 1.62 F2 R' D2 R' U2 R2 B2 F2 R' F2 D2 B' F2 D B U' F L F2 D2 L2
3. 1.44 U F2 L2 D2 B2 R2 F2 U' R2 D' U' L F2 D2 F D' R2 B' L D F
4. (1.32) R2 D R2 B2 L2 R2 U' L2 U' L2 F2 L' D B' F L' B2 R2 F L'
5. (2.05) R U2 L U2 F2 L2 R U2 B2 L2 D2 F U B' D2 R' F' U2 L' F
You need to be so, so comfortable with f2l that you can execute a pair and be comfortable with predicting, instantly, how it will affect the pieces around it. One possible exercise would be to do the inverse of the pair algorithm on a solved cube, and check what happens to the pieces. It's quite hard to describe all of the things you could possibly do - one advantage to the hobby of cubing over others is the practically infinite number of different exercises you can create for yourself to try and reach a state of 'unthinkingness' while solving. You want to devote as few mental cycles as possible to whats in front of you, and do your best to sharpen your 'peripheral vision' of what is to come next. Where most intermediate cubers still devote so much time to staring at stickers they have already processed information for, you never want to be checking the same piece of information twice, whether that be a sticker or the arrangement of the 3rd pair while you do the first.
I guess the point of this ramble is really that lookahead isn't so much as a 'trick', or singular 'thing', rather it's the process of optimising your information management, and it begins when you no longer have to double-check the simplest operations.
Riley
Veteran
Thanks for this. I'll definitely keep it in mind and try to put it to practice.
immortalchaos29
Veteran
Thanks Rowan, I appreciate your thoughts. I think using some of these tricks really will start to teach how to visualize the cube with your peripheral vision, as you say. The OP was right when he guessed that simply tracking the next pair one at a time is insufficient for top level cubing.
Did you mostly have to teach yourself these lookahead techniques in order to propel you to sub 10 much, or did you find that they started coming more and more natural to you as you did lots and lots of solves?
6.63, 3.88, 7.58, 3.66, 3.78 = 4.78 on the exercise. I clearly have room to improve. I seem to have trouble particularly when pieces were in adjacent B slots.
Did you mostly have to teach yourself these lookahead techniques in order to propel you to sub 10 much, or did you find that they started coming more and more natural to you as you did lots and lots of solves?
6.63, 3.88, 7.58, 3.66, 3.78 = 4.78 on the exercise. I clearly have room to improve. I seem to have trouble particularly when pieces were in adjacent B slots.
rj
Veteran
It still caught him off guard.
kcl
Legend
Not really if he predicted it..
rj
Veteran
And PLL recog slowed him down, because he's not used to regnizing PLL during F2L.
kcl
Legend
That makes no sense.. I really don't feel like being in a debate when I should be in bed, so I'll rest my case at this: 1. He predicted the OLL skip. 2. He took a moment for PLL recog which is normal. Nothing caught him off guard..
Escher
Babby
Not much time to reply now, but I'll tell you that from around 14-15 seconds, I began doing a lot more to improve my solving than just mindlessly grinding. In fact I would say I did more non-timed solves than I did timed speedsolves, concentrating on making my f2l as pro as possible. Yeah, for sure this kind of thing becomes more natural with experience, but for each solve you do, the more mindful you are, the better you'll get given each solve. Given the insane amounts I used to grind I think I was actually very inefficient, which might be why I think about this kind of thing so much! Personally it's kind of a chicken/egg argument as to which came first, but I think working on these kind of things earlier is better. That said, there is a basic level of competence to actually usefully apply the slightly deeper things, which I think is probably around 16 seconds average - that can simply be reached by turning faster and learning more 'hard' things (like algorithms) rather than 'soft'.
Hypocrism
Veteran
Your advice on this thread is really useful, I'm going to adopt your prediction ideas in my training along with "regular" lookahead training to try and finally break sub-15!
Thanks a lot Rowan!
I can feel this will greatly help me improve, especially for the "psychological confirmation" you gave me that such levels of prediction is possible.
(Just like in the way I mentioned earlier somewhere in this thread on how after I saw a sub-1 non-MU PLL was possible, that I managed to train enough to achieve it, despite being inconsistent sub-2 before).
Also, your description on your mental processes gave me a LOT of insight.
I had been training to follow all four pairs, but it does make more sense to understand the affected pieces. Also, you're right that it seems to be more than just one thing... your ability to be able to find all 4 pairs in under 2 seconds (and not just find the 8 pieces, since you say there's no "self-checking", I assume you also instantly know all 4 algs to solve any of the 4 pairs before you stopped the timer), combined with the ability to predict the changed pieces of your current pair's alg, I can imagine many things you can instead occupy your mind with (i.e. choosing to solve pairs that you predict will be unaffected and only shifted around on the u face, or actually just choose to insert a pair another way such that another pair is unaffected and just shifted around or something, or becomes an easier case... actually, 1-look F2L [blindfolded] doesn't seem too far of a stretch, since you know all 4 pairs, and you know what pieces are affected and how, so you can probably just update in your mind the alg(s) for the affected pair(s)... even if it's just doable only when its not too complicated, that's still an amazing thought).
Definitely explains why reconstruction of Feliks' and Mats' solves are so elegant with their F2L (it's like they're always lucky and all the cases are simple, or harder cases magically get simplified during the other pairs).
And also thanks for the exercises. Especially provides me insight on how you expect them to work, and your thought process while looking ahead.
Definitely not just tracking pieces of the next pair.
Very well thought out post. :tu
Also your avatar always catches my attention... is that girl your "#1 fan"? lol. She looks... uh... pretty well endowed lol
I can feel this will greatly help me improve, especially for the "psychological confirmation" you gave me that such levels of prediction is possible.
(Just like in the way I mentioned earlier somewhere in this thread on how after I saw a sub-1 non-MU PLL was possible, that I managed to train enough to achieve it, despite being inconsistent sub-2 before).
Also, your description on your mental processes gave me a LOT of insight.
I had been training to follow all four pairs, but it does make more sense to understand the affected pieces. Also, you're right that it seems to be more than just one thing... your ability to be able to find all 4 pairs in under 2 seconds (and not just find the 8 pieces, since you say there's no "self-checking", I assume you also instantly know all 4 algs to solve any of the 4 pairs before you stopped the timer), combined with the ability to predict the changed pieces of your current pair's alg, I can imagine many things you can instead occupy your mind with (i.e. choosing to solve pairs that you predict will be unaffected and only shifted around on the u face, or actually just choose to insert a pair another way such that another pair is unaffected and just shifted around or something, or becomes an easier case... actually, 1-look F2L [blindfolded] doesn't seem too far of a stretch, since you know all 4 pairs, and you know what pieces are affected and how, so you can probably just update in your mind the alg(s) for the affected pair(s)... even if it's just doable only when its not too complicated, that's still an amazing thought).
Definitely explains why reconstruction of Feliks' and Mats' solves are so elegant with their F2L (it's like they're always lucky and all the cases are simple, or harder cases magically get simplified during the other pairs).
And also thanks for the exercises. Especially provides me insight on how you expect them to work, and your thought process while looking ahead.
Definitely not just tracking pieces of the next pair.
Very well thought out post. :tu
Also your avatar always catches my attention... is that girl your "#1 fan"? lol. She looks... uh... pretty well endowed lol
Last edited:
uberCuber
Legend
Rowan, thank you for that post. For quite awhile I've noticed (and laughed at) how I'm so slow at 3x3 compared to what I can do on bigcubes. I always thought it was just because of the simple fact that getting good times on 3x3 requires faster turning than with bigcubes and that I just wasn't used to looking ahead (i.e. simply finding the next pair while solving this one) while turning at even as much as 4-5 TPS. Now I realize that I wasn't even trying to do as much as is possible with lookahead. You've given me a lot to think about, and I may try to finally improve at 3x3 now that I have a better idea of what the goal is.
Forte
Veteran
- Joined
- Jul 12, 2009
- Messages
- 1,151
- Location
- Waterloo, Ontario
- WCA ID
- 2009SHIN02
- YouTube
- Visit Channel
He's saying that if Mats had normally gone F2L -> OLL -> PLL, he might have predicted what case he was going to get in each step during the previous one (predicting his OLL before his F2L is done, predicting his PLL before his OLL is done), so he would not have to pause to recognize. However, since Mats got an OLL skip, he did not have as much intuition to predict what PLL he was going to get while doing his last F2L pair, so he had to pause to recognize it.
Also really amazing write-up Rowan
Escher
Babby
Lol yes that's my girlfriend, she changed my profile pic for a troll like 2 years ago, and uh, I guess it's a flattering photo.
Anyway, glad some people find this useful. I don't think that this stuff is particularly revolutionary or insightful (plenty of conversations irl with Kirjava and others confirm this), I guess I'm just good at writing stuff down - an English/Philosophy degree is good for something at least
The final thing I'll say is that yeah, for a few people I'm sure totally blindfolded f2l post-cross is quite possible. It's not possible for me right now without beginning to train again (which I haven't in over a year), but what I personally mean by locating each of the 8 pieces is that I develop a plan, which mostly combines solving a couple of pairs at a time; mostly like (1+2)(3+4), sometimes (1+2+3)(4), the more rarely (1+2+3+4). Sometimes you get scrambles where each pair affects fairly few other f2l pieces and these are lovely. Part of pair selection is learning to reduce the amount of f2l pieces that will be affected by your choices - with the other, more difficult part, as choosing the ones that will make each successive pair better.
This kind of response is tempting me to write up the pair selection guide that I've been thinking on for a year or two.
PeelingStickers
Veteran
Lol yes that's my girlfriend, she changed my profile pic for a troll like 2 years ago, and uh, I guess it's a flattering photo.
Anyway, glad some people find this useful. I don't think that this stuff is particularly revolutionary or insightful (plenty of conversations irl with Kirjava and others confirm this), I guess I'm just good at writing stuff down - an English/Philosophy degree is good for something at least
The final thing I'll say is that yeah, for a few people I'm sure totally blindfolded f2l post-cross is quite possible. It's not possible for me right now without beginning to train again (which I haven't in over a year), but what I personally mean by locating each of the 8 pieces is that I develop a plan, which mostly combines solving a couple of pairs at a time; mostly like (1+2)(3+4), sometimes (1+2+3)(4), the more rarely (1+2+3+4). Sometimes you get scrambles where each pair affects fairly few other f2l pieces and these are lovely. Part of pair selection is learning to reduce the amount of f2l pieces that will be affected by your choices - with the other, more difficult part, as choosing the ones that will make each successive pair better.
This kind of response is tempting me to write up the pair selection guide that I've been thinking on for a year or two.
Please do this, F2L movecount and general efficiency in lookahead is exactly my problem
teller
REAL Fingertricks!
Bless you, Rowan. And yes please.
Escher
Babby
No pressure, but I'm still waiting for the one on rotationlessness you mentioned here...
Ahha, my to-do list grows by the day... Although I'm a little less concerned with rotationlessness these days and how relevant it is to being fast, some little tricks and ideas from there would still be useful in getting faster.
Thanks Teller :3
D
Deleted member 19792
Guest
I think the top speedcubers use Turbo Tracking.
(Take one corner on a cube, then follow that corner by doing a crazy hand scramble. Move the pieces very fazt)
(Take one corner on a cube, then follow that corner by doing a crazy hand scramble. Move the pieces very fazt)
Ninja Storm
Veteran
That doesn't make any sense.
- Status
Similar threads
