The long and tortuous road to sub-20

[Foreright]

Only the 17 timed solves today as work was quite demanding plus it was pretty cold in my home office so was feeling it in my hands. It's a small sample size but there were some good solves today:

Step	Time	Recog	Execution	Goal	Diff	STM	TPS
FB	2.42​	0.00​	2.42​	4.00​	1.58​	6.43​	2.82​
SB	6.94​	1.00​	5.94​	6.00​	-0.94​	17.50​	3.05​
CMLL	4.36​	2.06​	2.30​	4.00​	-0.36​	12.00​	6.61​
LSE	5.57​	1.34​	4.23​	6.00​	0.43​	15.71​	3.81​
Average	19.29​	4.40​	14.89​	20.00​	0.71​	51.64​	2.68​

Finished with Ao100 on 19.88 and Ao1000 finally below 21 at 20.96. The graph is amusing though as that Ao1000 is dropping like a stone right now.

I spent some of this afternoon on untimed / slow solves just concentrating on getting my second block efficiency as high as possible and it resulted in some good (but slow!) solves. The following is STM for 30 solves:

FB	8.00
​ SB​	​ 15.20​
-- SB Square	-- 8.52
-- SB Last Pair	-- 5.92
CMLL	10.28
LSE	14.76
Total for solve:	48.68

So a sub-49 STM average - if I could maintain that whilst turning faster I'd be extremely happy! In actual fact I suspect this could have been lower as I wasn't focussing at all on first block so there's some higher-than-there-should-have-been counts for that in there and I had a couple of less optimal efforts towards the end. I suspect that my best course of action here is going to be to literally solve like this for some time to ingrain some of the more efficient solutions and forget about timing for a while as it's reinforcing bad habits.

 

[Foreright]

Interesting couple of days - I'm getting faster but it doesn't feel like it as with the speed comes reduced efficiency which is annoying. Did 15 timed solves today with an average of 19.27 @ 51.7 STM / 2.7 TPS which is better although a small sample size. Ao100 is now at it's best at 19.64 and Ao1000 at 20.84 and still coming down nicely. Highlight of those solves was the last one which was a 13.39 which came out of nowhere as the rest of my solves had all been above 17 (!):

13.38s Speedsolving solve by BewareTheWereHamster on Cubeast

cbst.link

Stats for those 15 solves:

Step	Time	Recog	Execution	Goal	Diff	STM	TPS
FB	3.21​	0.00​	3.21​	4.00​	0.79​	7.69​	2.48​
SB	6.73​	1.04​	5.70​	6.00​	-0.73​	16.85​	3.01​
CMLL	4.03​	1.25​	2.79​	4.00​	-0.03​	12.31​	6.01​
LSE	5.26​	1.21​	4.06​	6.00​	0.74​	15.62​	3.93​
Average	19.24​	3.49​	15.75​	20.00​	0.76​	52.46​	2.73​

Spent the rest of day playing with untimed solves concentrating purely on second block efficiency and inbetween meetings and work managed 100 solves with the move count stats looking like this:

STAGE	MOVE COUNT
FB	7.94
SB	15.63
... SB square	... 8.56
... SB last pair	... 6.54
CMLL	10.63
LSE	15.16

Average STM for the session was 49.29 which is nice as it shows I'm capable of solving in a relatively efficient way albeit with an average time of 41s. There were 5 solves in there with sub-40 move counts too with a low of 34 which was nice with the vast majority being in the 40-52 range which is my ultimate target for STM, especially since at 52 STM, you need a TPS greater than 2.6 to break 20 and that's a fair bit lower than my speedsolving TPS so should be ultimately achievable

I did notice that I was getting a bit faster towards the end of the session so maybe this is the way to practice - literally forget about the clock for a while and just steadily try to increase the speed keeping the efficiency until I'm back at full speed (?) - anyone had any success in doing this?

 

[Foreright]

Been a week now and a little progress I think and some frustration also. I've been doing a lot less timed fast solves and have been doing a LOT more untimed and / or timed slow solves with the smart cube concentrating on getting second block move count down consistently with some success. I have 295 of these solves recorded with an average STM of 49.7 but an average time of 41.2s. I've had plenty of sessions in that when STM was well below that (ie. 48 average and below) which is what my long term target will be for speedsolving I think. Interestingly the biggest fact on whether I get a 45 move solve or a 50+ seems to be LSE which makes sense as I haven't done any EOLR-type training at all and am kinda "winging it" a bit at that stage. When I slow right down to a crawl it's much easier to have an efficient LSE.

Anyway, in terms of times solves latest Ao100 is 19.53 with my Ao500 stubbornly staying above 20 at 20.02 (!) and Ao1000 at 20.72. My complete graph since I learnt full CMLL looks like this:

At the start I knew all of H / Pi / T / U and some L / S / As but there was a LOT of 2L CMLL solves in there hence the quick drop in times. The big hump is where I committed fully to learning S / As / L set and using them all. It's interesting to put this in the context of my complete solve history for this year which is 3200 solves or so and looks like this:

There's three humps here - the first at 1100 solves where I started learning CMLL and I guess learnt H / Pi and some of T / U and the second big hump where I learnt T / U and some of the remaining algs and then the third when I learnt the rest.

In terms of move count, the complete graph looks like this for STM with the blue line being Ao12 and red Ao100:

For individual steps the blue line is Ao100:

Quite a decent improvement I think!

In terms of time, I'm pretty much able to maintain a *just* below 20 average now in most sessions. Latest 20 timed solves I did this afternoon was an average of 19.79s / 51.2 stm @ 2.65 tps which is pretty standard right now. I'm still pausing WAY too much in my second block though and that's what's killing my average overall otherwise that would be sub-19 at least. Move count again would be better but I find it faster / easier to "spam" in the LSE stage rather than slowing down and actually look properly

Highlight of today was a 12.6 - my second best time ever - completely out of the blue as I had a 15 and a couple of 17s prior to that but nothing amazing.

12.59s Speedsolving solve by BewareTheWereHamster on Cubeast

cbst.link

This was a nice scramble as I could plan right into second block which always helps and the pieces were just "there" at every stage with an easy CMLL to boot.

 

[Deleted member 36445]

graphs

 

[Foreright]

graphs

Yep got to love a graph - I probably spend more time in excel than solving

 

[OtterCuber]

Yep got to love a graph - I probably spend more time in excel than solving

Move to R and you'll be doing a lot less solving. And we have similar goals and progress. You should push for Ao100 sub-10.

 

[Foreright]

Move to R and you'll be doing a lot less solving. And we have similar goals and progress. You should push for Ao100 sub-10.

Im not sure sub-10 is a realistic target for me - having a job and kids means I can’t practice 8 hours a day and I’m not naturally talented / fast enough to wing it.

 

[Foreright]

40 times solves today with an average time of 19.37 - so pretty flat on last week which is ok. STM is at 52.78 for the day which again is ok as I am trying to turn a bit faster so that's inevitably going to suffer a bit.

One encouraging thing is that the number of solves where I actually get good lookahead during the second block phase is increasing markedly. I'm finding it a bit easier to either full plan the first block (and importantly ignore it whilst I'm solving and look for SB pieces!) or to at least plan the square and track the last pair and DR edge and that's definitely showing up in my second block times which for today is below 7 seconds with just about a quarter of my solves today being sub-6 (ie. on target!) for the second block. Given a week ago I was lucky to get one or two solves below 6 seconds for second block per session that's a big improvement.

I can really see now that second block lookahead (or rather lack of...) is the only factor holding me back from being significantly sub-20 at this point. Out of my 40 solves today, 11 were sub-18 with 2 of those being sub-15 and 24 were sub-20.

One of the stats I put in my spreadsheet today is a prediction as to when I'll reach an Ao1000 of 20 seconds based on the slope/intercept of my Ao1000 graph. It currently reckons 310 solves. Current Ao1000 is at 20.65 with Ao500 now at 19.90. Given I'm doing an average of 20-40 solves per day (weekdays only!) that's approx. 2 weeks time assuming a steady rate of progress.

Trimmed mean (ie. discard top/bottom 10%) for each step for today looks like this:

Step	Time	Recog	Execution	Goal	Diff	STM
FB	3.41​	0.00​	3.41​	4.00​	0.59​	8.12​
SB	6.95​	1.00​	5.83​	6.00​	-0.95​	16.94​
CMLL	3.67​	1.13​	2.34​	4.00​	0.33​	11.52​
LSE	5.11​	1.12​	4.00​	6.00​	0.89​	16.39​

 

[espeed]

Reminds me of playing poker. Those were the times. Give me more graphs.

 

[scrubizilla]

you should try sub 9 but gl :;

 

[Cubinginatree]

40 times solves today with an average time of 19.37 - so pretty flat on last week which is ok. STM is at 52.78 for the day which again is ok as I am trying to turn a bit faster so that's inevitably going to suffer a bit.

One encouraging thing is that the number of solves where I actually get good lookahead during the second block phase is increasing markedly. I'm finding it a bit easier to either full plan the first block (and importantly ignore it whilst I'm solving and look for SB pieces!) or to at least plan the square and track the last pair and DR edge and that's definitely showing up in my second block times which for today is below 7 seconds with just about a quarter of my solves today being sub-6 (ie. on target!) for the second block. Given a week ago I was lucky to get one or two solves below 6 seconds for second block per session that's a big improvement.

I can really see now that second block lookahead (or rather lack of...) is the only factor holding me back from being significantly sub-20 at this point. Out of my 40 solves today, 11 were sub-18 with 2 of those being sub-15 and 24 were sub-20.

One of the stats I put in my spreadsheet today is a prediction as to when I'll reach an Ao1000 of 20 seconds based on the slope/intercept of my Ao1000 graph. It currently reckons 310 solves. Current Ao1000 is at 20.65 with Ao500 now at 19.90. Given I'm doing an average of 20-40 solves per day (weekdays only!) that's approx. 2 weeks time assuming a steady rate of progress.

Trimmed mean (ie. discard top/bottom 10%) for each step for today looks like this:

Step	Time	Recog	Execution	Goal	Diff	STM
FB	3.41​	0.00​	3.41​	4.00​	0.59​	8.12​
SB	6.95​	1.00​	5.83​	6.00​	-0.95​	16.94​
CMLL	3.67​	1.13​	2.34​	4.00​	0.33​	11.52​
LSE	5.11​	1.12​	4.00​	6.00​	0.89​	16.39​

Nice!

 

[abunickabhi]

Been a week now and a little progress I think and some frustration also. I've been doing a lot less timed fast solves and have been doing a LOT more untimed and / or timed slow solves with the smart cube concentrating on getting second block move count down consistently with some success. I have 295 of these solves recorded with an average STM of 49.7 but an average time of 41.2s. I've had plenty of sessions in that when STM was well below that (ie. 48 average and below) which is what my long term target will be for speedsolving I think. Interestingly the biggest fact on whether I get a 45 move solve or a 50+ seems to be LSE which makes sense as I haven't done any EOLR-type training at all and am kinda "winging it" a bit at that stage. When I slow right down to a crawl it's much easier to have an efficient LSE.

Anyway, in terms of times solves latest Ao100 is 19.53 with my Ao500 stubbornly staying above 20 at 20.02 (!) and Ao1000 at 20.72. My complete graph since I learnt full CMLL looks like this:

View attachment 21153
At the start I knew all of H / Pi / T / U and some L / S / As but there was a LOT of 2L CMLL solves in there hence the quick drop in times. The big hump is where I committed fully to learning S / As / L set and using them all. It's interesting to put this in the context of my complete solve history for this year which is 3200 solves or so and looks like this:

View attachment 21154
There's three humps here - the first at 1100 solves where I started learning CMLL and I guess learnt H / Pi and some of T / U and the second big hump where I learnt T / U and some of the remaining algs and then the third when I learnt the rest.

In terms of move count, the complete graph looks like this for STM with the blue line being Ao12 and red Ao100:

View attachment 21155

For individual steps the blue line is Ao100:

View attachment 21156

Quite a decent improvement I think!

In terms of time, I'm pretty much able to maintain a *just* below 20 average now in most sessions. Latest 20 timed solves I did this afternoon was an average of 19.79s / 51.2 stm @ 2.65 tps which is pretty standard right now. I'm still pausing WAY too much in my second block though and that's what's killing my average overall otherwise that would be sub-19 at least. Move count again would be better but I find it faster / easier to "spam" in the LSE stage rather than slowing down and actually look properly

Highlight of today was a 12.6 - my second best time ever - completely out of the blue as I had a 15 and a couple of 17s prior to that but nothing amazing.

12.59s Speedsolving solve by BewareTheWereHamster on Cubeast

cbst.link

This was a nice scramble as I could plan right into second block which always helps and the pieces were just "there" at every stage with an easy CMLL to boot.

Beautiful graphs and analysis yo.

How does your Bluetooth cube keep track of M moves? My Giiker registers it as 2 face turns in the reconstruction or move count analysis.

 

[Foreright]

Beautiful graphs and analysis yo.

How does your Bluetooth cube keep track of M moves? My Giiker registers it as 2 face turns in the reconstruction or move count analysis.

M/E/S (and wide) move detection is going to be down to the software. If you look at how the hardware works it can only register face moves and in some cases send gyro information. Cubeast at least detects the two face turns being simultaneous / near simultaneous and records them as M (or whatever) instead. This is the cases for both my iCarry and Moyu Ai.

I guess that whatever software the Giiker uses doesn’t do that.