# Speed FMC Trial Competition

#### Pyjam

Well for me it's silly that U2 counts for 1 when M counts for 2... I don't see how it's cleaner but that's just me…

Anyway the question is: is it (good) optimized speed solves, or bad (or lucky) FMC in 10 min?

#### Mike Hughey

##### Super Moderator
Staff member
I admit, I haven't tried, but I have doubts about the formula.
If someone solves the cube with his left hand while writing down the solution with his right hand, without barely any attempt to optimize, he may very well solve in 50 moves and 50 seconds, for a score of 195.6. Better than almost every score posted until now.
But I might be wrong, I haven't tried.

Also, it's disappointing if the HTM metric is used because it disqualifies the Roux method which is one of the best method (if not the best) for speed FMC. But the world is ruled by CFOP users! (Just kidding)
That's sort of the purpose for these trial runs. If you would, please try it! Let's see what can be done, and try to find the flaws. Specifically I'd like to know if someone can really do 50 moves in 50 seconds. I suspect it's quite possible, but I know it's completely out of my reach.

Or if Pyjam can't, maybe someone else will?

Keep in mind that this is different from a normal solve. You have no inspection time, and you have to scramble before you start to solve, and you have to actually write everything down. A 50 move solve is a pretty efficient solve; really good solvers can do it, but generally only with a 15 second inspection. And that plus the scramble would equal at least a 75 second solve.

And if you want I can compare scores with slice HTM to see what they do too. Now is the time to experiment.

Last edited:

#### xyzzy

##### Member
Also, it's disappointing if the HTM metric is used because it disqualifies the Roux method which is one of the best method (if not the best) for speed FMC.
This statement reads like "Roux is good but only if we change the rules", which comes with the implication that Roux is actually not good under the current rules… which I don't think is "obviously" true. (It might be true; it just needs some real justification!) I mean, for the usual one-hour non-speed FMC, Roux is roughly on par with methods like Petrus and ZB even when counting the moves with the face turn metric, even though people keep talking about how Roux is bad for FMC because of M slices.

Specifically I'd like to know if someone can really do 50 moves in 50 seconds. I suspect it's quite possible, but I know it's completely out of my reach.

Or if Pyjam can't, maybe someone else will?

Keep in mind that this is different from a normal solve. You have no inspection time, and you have to scramble before you start to solve, and you have to actually write everything down. A 50 move solve is a pretty efficient solve; really good solvers can do it, but generally only with a 15 second inspection. And that plus the scramble would equal at least a 75 second solve.
I'll try this. I don't think I can hit 50 moves in 50 seconds, but 80-100 seconds might be doable.

(watch me fail terribly as I overestimate my solving and writing speeds)

edit: ha ha ha nope, I can't think and write simultaneously fast enough

#1: 42 in 3:26.405 (score = 223.85)
#2: 53 in 4:32.914 (score = 297.29)
#3: 43 in 5:30.549 (score = 249.43)
#4: DNF
#5: 37 in 3:54.707 (score = 201.96)

Average: 256.86

#1 (42):
F B' L2 U R2 // 222 (5/5)
F' L D' L2 F2 L // xxcross (6/11)
D R' D' R D' F D F' // third slot (8/19)
B D2 B' // EO (3/22)
D2 L' D2 L D2 L' D' L // last slot (8/30)
F' D B D' F D' B' D' B D' B' D // ZBLL (12/42)

#2 (53):
R D' R' F2 R F // p222 (6/6)
D' L D' L2 // another square (4/10)
R' B' R B2 L' F' // realign stuff (6/16)
D2 B' D B D' B' D' B // third slot (8/24)
D' R' D2 R D' F L D L' // VHLS (9/33)
D F' D F D' F' D F D2 F' // OLL (10/43)
B2 R D R' B2 L U' L U L2 // J perm (10/53)
This one was pretty bad. I actually know the ZBLL for the inverse of this case but switching to inverse seems like cheating…

#3 (43)
U F' R2 D B // EOline (5/5)
R2 U L R' U L // left sq (6/11)
U' R2 U R' L U2 L' // left (7/18)
R' U' R' U R' U' R U2 R' D // right+align (10/28)
U' L U2 B' L U L U' L' B L2 U L2 U L' // ZBLL (15/43)
(okay I "cheated" and used the inverse ZBLL this time; I haven't learnt this case yet, but I know the inverse)

#4 DNF

#5 (37)
D' B2 D R U R // 222 (6/6)
L B2 D2 B' // 223 (4/10)
D2 B D U B' U' // EO (6/16)
B2 L B' L' B2 // F2L-1 (5/21)
D L2 D' L' D L' D' L' // edges; ab3c (8/29)
B2 L' F' L B2 L' F L // L3C (8/37)

Last edited:

#### WoowyBaby

##### Member
Whoops, didn’t mean for my thoughts to go on and on. TL;DR 50 in 50 no, 50 in 100 yes.

—-

It seems to me everyone is talking about how a good OH solver could just get sub-200 points all the time by getting 50 or so moves in 50 or so seconds, but I do not think this is possible. I tried this out myself, but I realized it doesn’t matter how good of an OH solver you are, I’m pretty bad if I’m being honest, but the real bottleneck is writing speed.

It takes a bit over a second to write each move, even if it’s really sloppy, which I was lol. My best time for this strategy was 1:49, which is 109 seconds. It was also 51 moves, which suggests getting 50 moves all the time super quick is not much of a challenge, it’s pretty reasonable, but maybe not 50 seconds all the time. At the end of my 109 second attempt, what I had on the paper was what you call chicken scratch. Lol. I had trouble reading it myself because I was trying to go too fast, while in any other situation I can read my writing perfectly fine. Now, there’s no rules on how sloppy you can write, as long as you can read it yourself, without a cube of course. So, as a result of this, or just a side effect or whatever you call it, there is a rule in which once you stop the timer, you are not allowed to use a cube until you’ve submitted your solution. Element of surprise added as bonus hehe . Although, people are probably already doing this so this shouldn’t be an issue.

So, I have concluded that consistently getting 50 moves in 50 moves is not possible, but consistently getting 50 moves in 100 seconds just might be.

Edit: Forgot about computer keyboards. I need to do the same test but on a keyboard and see if it makes it any faster, perhaps I can get it down to 80 seconds, but maybe I’ll find out it’s not actually that good? I’ll come back later with results and see.

Last edited:

#### Pyjam

So, good news.
Now, if you get a 50 moves solutions in 2 minutes (240 pts), is it worth trying to get a better result?
That's a risky bet. Interesting!

#### Mike Hughey

##### Super Moderator
Staff member
...
This one was pretty bad. I actually know the ZBLL for the inverse of this case but switching to inverse seems like cheating…
...
(okay I "cheated" and used the inverse ZBLL this time; I haven't learnt this case yet, but I know the inverse)
Why would it seem like cheating? It's perfectly fine if it's within the rules. It's trickier because you have to place it properly in the solution, but it should be fine to do it. And wow, it's helpful for this event to know ZBLL!

#### xyzzy

##### Member
Why would it seem like cheating? It's perfectly fine if it's within the rules. It's trickier because you have to place it properly in the solution, but it should be fine to do it. And wow, it's helpful for this event to know ZBLL!
Oh, I meant it in the sense of trying to write down a solution as quickly as possible, rather than writing down a reasonably good solution in a reasonable amount of time. And it's faster to write down OLL+PLL (especially when it's a short PLL), rather than grabbing another cube to set up the last layer case on. I'd totally use NISS (if it was worth it) in a proper attempt.

#### Jacck

##### Member
#1-#4: DNF (time ran out much to soon)

#5: 32 in 9:46.09 (score = 203.95)
lucky skeleton in 5 min and fast insertions (and IF says: optimal), probably not beatable by my anymore
L2 * F2 L' # B D' L' D2 // 222
U' L' B U2 L' B // 223
U' L' U L U2 // F2L-1
L' B L2 B' L2 // L5C in 23
* L B L' F2 L B' L' F2 // 8-4/27
# D F2 D' B D F2 D' B' // 8-3/32
solution: L' B L' F2 L B' L2 D F2 D' B D F2 D2 L' U' D2 L' B U2 L' B U' L' U L U2 L' B L2 B' L2

#### Mike Hughey

##### Super Moderator
Staff member
#5: 32 in 9:46.09 (score = 203.95)
lucky skeleton in 5 min and fast insertions (and IF says: optimal), probably not beatable by my anymore
Optimal 5-corner insertion for speed FMC - there's something wrong about that.

I'm going to start the next round in a few hours. Any opinions on a possible value for k for a linear scoring formula? Or are we already convincing ourselves that the logarithmic formula is really the way to go?

#### Kit Clement

Optimal 5-corner insertion for speed FMC - there's something wrong about that.

I'm going to start the next round in a few hours. Any opinions on a possible value for k for a linear scoring formula? Or are we already convincing ourselves that the logarithmic formula is really the way to go?
I like 1/60, as then that scoring system translates to exactly "Moves plus minutes" which is incredibly easy to understand. I've done comparisons of results and it tends to agree with the log scoring system as to which result is better. The main difference is that moves plus minutes seems to prefer a quicker time more than a shorter solution, which I think fixes the issue that taking all 10 minutes in this strategy is easily the best use of your time if you aren't trying to go for ultra fast solutions. Maybe I'll run some code to see how moves+minutes compares to log.

I'd also like to say that at this point I support moves+minutes over log scoring.

#### Mike Hughey

##### Super Moderator
Staff member
I like 1/60, as then that scoring system translates to exactly "Moves plus minutes" which is incredibly easy to understand.
That does seem like a particularly elegant formula - hard to be much simpler than "moves plus minutes"!

#### Kit Clement

Here are some contour plots from WolframAlpha for each scoring system. The x-axis is the number of moves, and the y axis is the number of minutes. These plots draw lines that connect pairs of inputs with the same output. Below is the plot for moves+minutes. Thus, we can see that a 40 move solution in 0 minutes (lol) is worth the same as a 30 move solution in 10 minutes, as the line connects those two outcomes.

Here's the same contour plot for log scoring:

So when the solution gets to about 4+ minutes on log scoring, these contours seem to have around the same slope as the contour lines of the moves plus minutes, meaning that the trade-off of spending more time to get a shorter solution is roughly the same. The trade-off is far different for really quick solutions, which is what people have been communicating through examples - 50 moves in 50 seconds is a very good result, and you lose a lot of points fast if you spend more time.

I'm fully in support of the moves+minutes scoring system at this point.

#### Mike Hughey

##### Super Moderator
Staff member
Results of first round, based on logarithmic scoring:
1. TipsterTrickster 230.38
2. Kit Clement 242.75
3. xyzzy 256.86
4. WoowyBaby 262.23
5. qT Tp 269.07
6. Mike Hughey 292.44
7. fun at the joy 310.18
8. irontwig DNF
8. Jacck DNF

Results of first round, based on linear scoring (moves + minutes):
1. TipsterTrickster 45.19
2. Kit Clement 47.92
3. xyzzy 50.50
4. qT Tp 51.62
5. WoowyBaby 51.82
6. Mike Hughey 55.58
7. fun at the joy 60.52
8. irontwig DNF
8. Jacck DNF

Interesting - we do have a rearrangement of fourth and fifth place. It might be worth discussing the relative merits of the two systems based on those differences.

Thanks everyone for giving this a try.

Last edited:

#### Mike Hughey

##### Super Moderator
Staff member
Round 2, closes at midnight GMT on December 29, in a little under 3 days. That gives me time to make last minute changes before it goes live in competition (assuming I can really get it done in time, which looks promising at the moment).

Scrambles for round 2:
1. R' U' F D B2 F2 U L2 F2 L2 D R2 B2 D' L2 B' D F L' R B U2 L R' U R' U' F
2. R' U' F B R2 F' L2 B D2 F' L2 B R2 U2 R F D U R' B2 U2 F' L2 D' R' U' F
3. R' U' F D F2 L U2 L' B2 D2 R' D2 R2 B2 F' L2 U' B2 F' U B L D R' U' F
4. R' U' F R2 B2 R2 D F2 U' B2 D2 L2 D B2 L' B' R B' D2 B' L' D U L' R' U' F
5. R' U' F U2 F D2 R2 F' L2 B R2 B D2 U F' R' U' L' B U' R B' U' R' U' F

No time limit this time!

Official scoring for this round will be:
score = moves + minutes (measured in hundredths of a second, with overall score rounded to hundredths of a point)

Last edited:

#### Kit Clement

Have a lot going on right now so I don't have time to post solutions, but here's what I got

1. 44 in 10:00.00 (solution written and circled just before timer expired)
2. 41 in 9:46.58
3. 36 in 6;34.52
4. 35 in 5:54.21
5. DNF

That gives an average of 40.78.

It probably should be discussed if the situation for #1 should be DNF or not, which I didn't think about until afterward (and probably wouldn't have DNF'd 5 if #1 was DNF'd). I think that it should work similar to MBLD, where you get the current result you have as time expires, and in real FMC you still get to turn in your solution/write your name once time expires. But whatever we think, we should decide before we put this on the SS comp!

Last edited:

#### Jacck

##### Member
#5: 32 in 9:46.09 (score = 203.95)
lucky skeleton in 5 min and fast insertions (and IF says: optimal), probably not beatable by my anymore
Did I wrote that? Take a look at #4:
1. DNF
2. DNF
3. 39 in 9:41.10 (score = 48.69; ln-score = 248.23)
4. 30 in 7:40.81 (score = 37.68; ln-score = 183.99)
5. DNF (45 in 10:07.05)

3.
R F' L // EO
D B D' // 222
U2 L2 U2 B U2 B' F U' F' // 223
U B U B U2 // L6C in 20
U2 B' D' B U2 B' D B // 8-4/24 L5C
B D F2 D' B' D F2 D' // 8-1/31 L3C
L U' R2 U L' U' R2 U // 8/39
solution: R F' L D B D' U2 L2 U2 B U2 B' F U' F' U B U D' B U2 B' D B2 D F2 D' B' D F2 D' L U' R2 U L' U' R2 U

4. 30 in 7:40.81
D' L R2 D' // 222
F D' L D L' // 223
L' U L // F2L-1
U R U2 R' U F' U F // L5C in 19 after 3:45 min
ssf: D' # L R2 D' F D' L D L2 U L U R U2 R' U F' * U F
* F R F' L F R' F' L' // 8-2/25 L3C
# D' L D R2 D' L' D R2 // 8-3/30 (IF says 29)
solution: D2 L D R2 D' L' D L D' F D' L D L2 U L U R U2 R' U R F' L F R' F' L' U F

#### qT Tp

##### Member
1. 41 in 7:28.64 score:48.48
2. 40 in 9:37.02 score:49.62
3. 42 in 8:11.15 score:50.19
4. 37 in 5:18.64 score:42.31
5. 34 in 9:27.95 score:43.47

average: 47.19

Solve 1:
D U R2 F2 R2 F2 //222
U B L B2 L' F U B' F' //F2L-2
U2 R' U2 R //F2L-1
F U F' U2 F U' F' //F2L
R U B U' B' //OLL (R')
R2 U L' U2 R U' R' U2 R L U //J-PERM

F2L in 26, regular OLL-PLL

solve 2:
R F2 D2 L2 F U R' //222
D2 F' L U L2 U' F2 // F2L-2 (L2)
L F' D2 F2 D' F' D L D B // 2E2C
B //SETUP N-PERM
L D' R D2 L' D R' L D' R D2 L' D R' D B'

23 moves to 2e2c, was short on time so resorted to a setup-N-perm.

solve 3:
B2 F' L U B2 U //222
D2 L' B' L' D' B2 //F2L-2
L2 U' L2 F L' F' L' U // F2L
B L B' L' B' U B U' //OLL
D L2 D' L2 D F' D' L' D L D F D2 L //R-PERM

F2L in 20, regular OLL-PLL

solve 4:
D' L R2 D' //222
F2 D F' D' //223
F' U2 F L' B' U B L U2 //F2L-1
F R' F R F' R' F2 R //F2L/OLL
U2 B' R F' R F R' B U2 L F' L' //G-PERM

8 move 223, followed by some "strange" block building to setup F2L/EO, forced OLL skip and G-perm

solve 5:
U F D B D2 B //222
U L U L2 U F' //223
B L B2 U2 B L** U L' *U2 //5C
*= R D2 R' U R D2 R' U' (8-1)
**= L D L' U' L D' L' U (8-2)

5 corners left in 21 with the timer at 4 minutes. I tried some insertions finishing in the last minute. It would have probably been a better score if I just did regular 3-cycle commutators at the end

Last edited:

#### Mike Hughey

##### Super Moderator
Staff member
It probably should be discussed if the situation for #1 should be DNF or not, which I didn't think about until afterward (and probably wouldn't have DNF'd 5 if #1 was DNF'd). I think that it should work similar to MBLD, where you get the current result you have as time expires, and in real FMC you still get to turn in your solution/write your name once time expires. But whatever we think, we should decide before we put this on the SS comp!
At first I was confused by this - it seemed to me that of course it should count as a success. But then I realized - what you're talking about here is that you didn't stop the timer? Actually, after thinking about this, I really think we'd need to count that as a DNF. Just like with a regular solve, you must finish the solve, release the cube, and stop the timer in order for the solve to count, I think you need to release the pen (and/or any cubes) and stop the timer for this to count. If you don't stop the timer before it runs out, I think it should count as a DNF.

I am currently working on the timer for the weekly competition. It's a mess to deal with, and I'm not sure if I'll have it ready by Monday, but the manual entry page is already ready, so I intend to have the event added - the timer might be disabled, depending on how much luck I have with fixing it before the start of the competition. But I do intend to make it so that if it reaches 10 minutes, it sets the result for the attempt as a DNF; I think that's the right thing to do. You must stop the timer before time runs out in order for it to count.

I would probably like to have an audible alarm letting people know time is getting close to running out. Again, this probably won't happen right away - it will probably be several weeks before that is implemented, even if I get the basic timer done in time, and we would want to define the times that should be announced. Perhaps one minute, 15 seconds, 5 seconds? Don't know. Also, there are some browsers where it seems like the audio announcements are a little flaky with my code; people will need to be wary of that until/unless I can fix it.

#### Kit Clement

At first I was confused by this - it seemed to me that of course it should count as a success. But then I realized - what you're talking about here is that you didn't stop the timer? Actually, after thinking about this, I really think we'd need to count that as a DNF. Just like with a regular solve, you must finish the solve, release the cube, and stop the timer in order for the solve to count, I think you need to release the pen (and/or any cubes) and stop the timer for this to count. If you don't stop the timer before it runs out, I think it should count as a DNF.

I am currently working on the timer for the weekly competition. It's a mess to deal with, and I'm not sure if I'll have it ready by Monday, but the manual entry page is already ready, so I intend to have the event added - the timer might be disabled, depending on how much luck I have with fixing it before the start of the competition. But I do intend to make it so that if it reaches 10 minutes, it sets the result for the attempt as a DNF; I think that's the right thing to do. You must stop the timer before time runs out in order for it to count.

I would probably like to have an audible alarm letting people know time is getting close to running out. Again, this probably won't happen right away - it will probably be several weeks before that is implemented, even if I get the basic timer done in time, and we would want to define the times that should be announced. Perhaps one minute, 15 seconds, 5 seconds? Don't know. Also, there are some browsers where it seems like the audio announcements are a little flaky with my code; people will need to be wary of that until/unless I can fix it.
Yeah, that's the weird thing with SpeedFMC, as if you look at FMC, there's no timer to stop, and if you have your pen down with the solution written down in time, then the solution counts. Same with MBLD -- while you have to stop the timer to get the attempt under 1 hour, your attempt still counts even if you just finish a cube right before the 1 hour mark and put your hands down after the 1 hour mark. But with speedsolving a 7x7, if you're around 10 minutes, you need to include the timer stop in that solve, which is different than time in MBLD. I think either choice is fine and can be justified with how either speedsolving events are handled or with how long events are handled, but I figured it was good to discuss now what we would do about it.

#### Mike Hughey

##### Super Moderator
Staff member
If you imagine what this event would actually be like in an official competition, well, the first thing that would occur to you is how impractical it would be. That's the benefit of being able to do this in an at-home online competition - we don't have to worry about all those pesky details! But it would be absolutely necessary in an official competition to have a judge at every competitor, watching for the end of the solve, to make sure the timer was properly stopped and that nothing else was written after the timer was stopped. I guess if you did this in an official competition with two rounds, it might be doable - the idea would be that you would have two groups (with two different sets of scrambles!) first round, and half the competitors would judge the other half. Then the half of the competitors that didn't make the cut to the second round would be required to stay and judge the half that made it, so that you could have enough simultaneous judges to judge everyone with the same set of scrambles to make the finals fair. It would be truly a nightmare! But I think it would be necessary; the chaos that already exists after regular FMC would be so much worse for this event, there's no way it could work without a separate judge for each competitor. Even one judge per two competitors would be really inadequate for difficult scenarios - in fairness it would have to be one judge per competitor.

In any event, my vote is that a timer stop should be required. I intend to go with that unless there are significant objections. But since this was just a trial run, I have no objections with counting your solve for this round anyway, since the rules had not been clearly stated before the round began.