• 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 35,000+ people from around the world today!

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

Measuring Standard Deviation of solve averages

Joined
Jul 21, 2014
Messages
2,896
Likes
1,532
Location
Spain
WCA
2018GUTI13
YouTube
QBAndo
Thread starter #1
Hi all!
I beat my PB Ao12 yesterday and I got a standard deviation of 2.87. Another guy posted his PB Ao12 right after me (both posts are consecutive here in the accomplishment thread). His Ao12 was 0.1 worse than mine but the std.dev was 0.82.

So I started to think perhaps my consistency is not very good. In a typical Ao100 I still get like 20% over 25sec solves, even 1 or 2 over 30sec and, at the same time I get around 15% sub-20 with several 17-18 sec solves. Typical std. dev for an Ao100 ranges from 2.2 to 2.6.

Most of the over 24-25 sec times are due to missed OLL / PLL, wrong slot for an F2L pair and that kind of silly mistakes. Do you guys experience the same problem or is it just me and my recognition / visual memory being bad?

Also a question for the mathematician out there. I have seen that the way csTimer calculates std. dev. for , e.g., an Average of 5 is not with the 5 values and the full population formula but with just the 3 counting values and the "sample of population" formula. Not sure if this is the most meaningful way to calculate std.dev when you have the whole population to calculate with.
 
Joined
Oct 19, 2014
Messages
3,327
Likes
318
Location
Ottawa, Canada
WCA
2015MANS03
YouTube
P3NGU1N5D0NTFLY
#2
Hi all!
I beat my PB Ao12 yesterday and I got a standard deviation of 2.87. Another guy posted his PB Ao12 right after me (both posts are consecutive here in the accomplishment thread). His Ao12 was 0.1 worse than mine but the std.dev was 0.82.

So I started to think perhaps my consistency is not very good. In a typical Ao100 I still get like 20% over 25sec solves, even 1 or 2 over 30sec and, at the same time I get around 15% sub-20 with several 17-18 sec solves. Typical std. dev for an Ao100 ranges from 2.2 to 2.6.

Most of the over 24-25 sec times are due to missed OLL / PLL, wrong slot for an F2L pair and that kind of silly mistakes. Do you guys experience the same problem or is it just me and my recognition / visual memory being bad?

Also a question for the mathematician out there. I have seen that the way csTimer calculates std. dev. for , e.g., an Average of 5 is not with the 5 values and the full population formula but with just the 3 counting values and the "sample of population" formula. Not sure if this is the most meaningful way to calculate std.dev when you have the whole population to calculate with.
Yeah I am also quite inconsistent. I fail times and get like 17s because of wrong slots. They are quite common, maybe 1/25 ish solves. Then I get like 3 10s and like 2 14s and there goes my chance at a good average.
 
Joined
Jun 25, 2014
Messages
429
Likes
51
Location
Atlantic Canada
WCA
2014BOUD01
#3
Those silly mistakes happen to a lot of cubers actually. I mess up my cross a lot when I plan out a move wrong and execute it trying to look for pairs. It's so annoying because I'm half way-3 quarters done my cross by the time I notice I have to place my final cross edge.

Also, what is the question exactly in the last paragraph? I'm no mathematician and I don't know how the standard deviation is calculated, but I do agree the standard deviation should include the entire average.
 

Kit Clement

Premium Member
Joined
Aug 25, 2008
Messages
1,269
Likes
455
Location
Portland, OR
WCA
2008CLEM01
YouTube
KitClement
#4
Also a question for the mathematician out there. I have seen that the way csTimer calculates std. dev. for , e.g., an Average of 5 is not with the 5 values and the full population formula but with just the 3 counting values and the "sample of population" formula. Not sure if this is the most meaningful way to calculate std.dev when you have the whole population to calculate with.
But this isn't a population of values, it is a sample. In reality, you are sampling from a theoretical distribution of all possible times you could receive, so even your whole session of solves is considered a sample.
 
Joined
Jul 21, 2014
Messages
2,896
Likes
1,532
Location
Spain
WCA
2018GUTI13
YouTube
QBAndo
Thread starter #5
Those silly mistakes happen to a lot of cubers actually. I mess up my cross a lot when I plan out a move wrong and execute it trying to look for pairs. It's so annoying because I'm half way-3 quarters done my cross by the time I notice I have to place my final cross edge.

Also, what is the question exactly in the last paragraph? I'm no mathematician and I don't know how the standard deviation is calculated, but I do agree the standard deviation should include the entire average.
Not sure if I should find it reassuring to know that it happens to other people as well, but it is frustrating nonetheless.

Regarding the std. dev. question I just find it curious that csTimer does not take into account the 2 dropped values of an Ao5 or Ao12 when calculating the standard deviation. It does not reflect my lack of consistency because it is very rare that I get an Ao5 without one horrible time, which, if counted would raise the std dev a lot, like, e.g.: 20.x, 17.x, 19.x, 19.x, 26.x. This is so true that I get track of my "best worst time in an Ao5 or Ao12". I thought of changing that to track std.dev, but given the fact it does not include those times it does not serve my purpose of tracking if my consistency improves.

Perhaps in terms of consistency it is better to track std. dev over large averages, like Ao50, 100 or bigger.

Thanks for your feedback
 
Joined
Jul 21, 2014
Messages
2,896
Likes
1,532
Location
Spain
WCA
2018GUTI13
YouTube
QBAndo
Thread starter #6
But this isn't a population of values, it is a sample. In reality, you are sampling from a theoretical distribution of all possible times you could receive, so even your whole session of solves is considered a sample.
What you say is very true!!!! I hadn't thought about that.
 
Joined
Sep 5, 2014
Messages
275
Likes
4
Location
Tennessee
#7
Yeah, probably 5-10% of my solves at least I do something stupid like messing up my COLL or putting in a pair wrong or putting the cross in wrong. It really bugs me because my most important statistic is ao100 and so those times still count.
 
Joined
Jul 21, 2014
Messages
2,896
Likes
1,532
Location
Spain
WCA
2018GUTI13
YouTube
QBAndo
Thread starter #8
Yeah, probably 5-10% of my solves at least I do something stupid like messing up my COLL or putting in a pair wrong or putting the cross in wrong. It really bugs me because my most important statistic is ao100 and so those times still count.
It's the same in my case, it can spoil any long average, but oh, when it spoils a good *short* average... grrrrr, I have lots and lots of Ao5 with 4 times sub-20 or at least sub-20.5 and then... a 25.x, e.g.: 19.56, 20.33, (17.63), 17.64, (25.90) bummer!
 
Joined
Sep 5, 2014
Messages
275
Likes
4
Location
Tennessee
#9
It's the same in my case, it can spoil any long average, but oh, when it spoils a good *short* average... grrrrr, I have lots and lots of Ao5 with 4 times sub-20 or at least sub-20.5 and then... a 25.x, e.g.: 19.56, 20.33, (17.63), 17.64, (25.90) bummer!
At least there the highest time is ignored. The way I see it it just means you have more room to improve. Someone with a low std is stuck where they are, but if you can figure out what you're doing right in the fast solves and what you're doing wrong in the slow ones, you can improve a lot more. I have found that metronome training helps you move slower and solve better, even if you gradually speed it up and then just turn it off, it helps. Frantically trying to go as fast as possible just leads to more mistakes.
 
Joined
Jul 21, 2014
Messages
2,896
Likes
1,532
Location
Spain
WCA
2018GUTI13
YouTube
QBAndo
Thread starter #10
I think I should perhaps take your advice and try that, with or without metronome, but definitely slow solving because I get more and more frustrated every day. Today I got 5 sub-20s in a row for the first time and I thought "maybe this is the tipping point towards real improvement" and right after that I started to do 25s all over again. It's like all of a sudden I want to turn faster than I really can and solving becomes a fight between my hands and the cube, which the cube always wins :)
 
Joined
Sep 5, 2014
Messages
275
Likes
4
Location
Tennessee
#11
One thing that helps me when I can feel myself getting nervous or getting bad times is I just think "I'm not trying to beat a pb here, I'm trying to improve my cubing ability. Each solve is a learning experience, and if I can learn what it's trying to teach, the pb will come on its own."
 
Joined
Jul 21, 2014
Messages
2,896
Likes
1,532
Location
Spain
WCA
2018GUTI13
YouTube
QBAndo
Thread starter #12
One thing that helps me when I can feel myself getting nervous or getting bad times is I just think "I'm not trying to beat a pb here, I'm trying to improve my cubing ability. Each solve is a learning experience, and if I can learn what it's trying to teach, the pb will come on its own."
Yeah, I try to think along the same lines myself but I don't always succeed. :(

By the way, we have very similar times right now but I couldn't help but notice in your "progress" post that you have rapidly improved your Ao100 this month. When did you start speedcubing?
 
Joined
Jul 21, 2014
Messages
2,896
Likes
1,532
Location
Spain
WCA
2018GUTI13
YouTube
QBAndo
Thread starter #13
One thing that helps me when I can feel myself getting nervous or getting bad times is I just think "I'm not trying to beat a pb here, I'm trying to improve my cubing ability. Each solve is a learning experience, and if I can learn what it's trying to teach, the pb will come on its own."
I think it is working! I find myself being able to concentrate on lookahead more easily and with less warmup and my 50 & 100 averages are coming down from sub-23 to sub-22. I still have those moments when there's a 24-to-26 solve or two in a row but it is rarer now to do worse than that unless the cube explodes or I miserably fail OLL/PLL and have to start solving again.

I have also changed my main from Chilong (core lubed with silicone, pieces lubed with Maru) to a Sulong (core and pieces lubed with different weight silicone). I find that the latter is a bit slower but more controllable, which seems to help my consistency.

I'm also starting to practice OH a bit (with my left hand, like most people) but I don't know if this has anything to do with my slight but noticeable improvement.

Thanks to all that replied for your comments. I have started to keep record of the standard deviation of my averages and they vary greatly but I've got some really consistent ones so , I'll keep trying.
 
Top