Rate of improvement theory (project)

[goldencuber]

For those of you who log your times, would you mind sending me a log file of your times for all events to
[email protected]. If you have anything (as long as it's not just 4 solves or something, send it: I need as much data as I can get)
In the email, please use this format
name (optional)
timer name
age (optional)
and other info needed (This would include if your times only represent half of your solves, or if you rarely cubed for a month from solve 456 to solve 460, or anything else that would be helpful)

It will likely be a while until I get complete results.

I'm working of finding the general rate of improvement rate per X solved cubes. Obviously, this is virtually impossible, as many factors affect this. However, it is definitely possible to find a general trend.
Part of my goal is to find out:
-if a person takes X solves to reach say...sub 1 on a 3x3, and XY represents the solves needed from there to reach sub 30, than what is Y?
-A general power regression equation that would apply for most people
For example: I received a 7x7 recently, and I've solved and logged the times 25 times. Using my calculator, I found the power regression of my times to be about
y=1738.656X^-.375587
Where X is the solve number, and Y is the time in seconds.
Currently, with 25 solves, I'm in the 9:00-10:00 zone, verging on sub 9. This equation predicts that my 50th solve would be 6:40.03, a fairly reasonable number. However, it predicts my 500th solve as 2:48???

There is obviously a equation/horizontal asymptote adjustment needed, and i need more times. (and power regression is only accurate up to a point)

 

[JLarsen]

Long story short-

I'm working of finding the general rate of improvement rate per X solved cubes. Obviously, this is virtually impossible, as many factors affect this. However, it is definitely possible to find a general trend.
Part of my goal is to find out:
-if a person takes X solves to reach say...sub 1 on a 3x3, and XY represents the solves needed from there to reach sub 30, than what is Y?
-A general power regression equation that would apply for most people
For example: I received a 7x7 recently, and I've solved and logged the times 25 times. Using my calculator, I found the power regression of my times to be about
y=1738.656X^-.375587
Where X is the solve number, and Y is the time in seconds.
Currently, with 25 solves, I'm in the 9:00-10:00 zone, verging on sub 9. This equation predicts that my 50th solve would be 6:40.03, a fairly reasonable number. However, it predicts my 500th solve as 2:48???

There is obviously a equation/horizontal asymptote adjustment needed, and i need more times. (and power regression is only accurate up to a point)

anyway, for those of you who log your times, would you mind sending me a log file of your times (from start to current) at [email protected]. Sadly, I can't just ask the dates that you subbed X, as it wouldn't be helpful.
In the email, write you name, timer name, age, gender, and other info needed (This would include if your times only represent half of your solves, or if you rarely cubed for a month from solve 456 to solve 460, or anything else that would be helpful)

It will likely be a while until I get complete results.

YOU JUST HIT THE JACKPOT MO F***KA! Rofl

http://www.speedsolving.com/forum/showthread.php?t=10157&highlight=3500+solve

 

[goldencuber]

I'll await your email.
(lol I don't know what I'd say if I didn't force myself to stay calm.)

 

[JLarsen]

hahahahahaha ill send it to you right now. I feel super happy someone is going to put it to good use. and all this time I saw it a complete waste.

Edit: I'm trying to send the entire folder but it's only letting me attach individual notepad files. I'm not sure what to do here.

Edit again: are you only interested in 3x3 speed or all of the events that I have. and also you don't need the actual notepad file with all the info do you?

 

[goldencuber]

everything would be best

You can't compress everything into one .zip file?

edit:hmm...maybe you could send both the "complete" files, and another reduced version. The other info could be useful for other stuff

 

[JLarsen]

Good idea. I'm on it.

Sent.

 

[mazei]

I think the gradient of the graph reduces the further you progress. Its easier for you to improve when you're at 1 min compared to when you're at 15 seconds.

 

[goldencuber]

yeah you're right. I'm trying to find the differences in the amount of improvement for each level though.

 

[abr71310]

Interesting project. I was going to look into something myself when i snapped my wrist playing badminton (since i can't cube for a week, I figured I was going to study some theory, anyway), so good job picking up my n00b slack (LOL I was probably never going to get around to it anyway XD)

My theory was that improvement is a standard exponential (negative?) curve up until we hit some marker X. The progression then changes to a small negative linear function until we hit present day Y.

I've still yet to see why this works, but so far it's worked with everyone in my school who cubes + me. That marker X for me was I think 33.40 seconds, since improvement has been scarce in the past weeks.

It might have something to do with genetics though, since I'm pretty sure some people will progress far faster than a standard curve. If we took a normal distribution of "improvement" over all of the speedcubers in the world i'm pretty sure we'd get exactly what we looked at; some would progress faster than the possible "standard curve" (that is, the equation for "normal" improvement), and some would progress slower...

anyway, my hand is tired (only using left hand right now). I'll type more when i have hte brainpower.

 

[cmhardw]

I have about 8 years of saved .txt logs on my hard drive. That is the good news. The bad news is the controller card appears to be dead, and possibly the drive itself. I'll be getting a new controller card soon to test option 1. If that doesn't work, and the drive is dead, at some point in the next year or so I will send it off to have the discs taken out and the data extracted. This costs a lot of money, but I have a lot of sentimental stuff on this drive (including my practice logs) that I want to save. So when I get this figured out I can send you about 8 years worth of practice logs.

Chris

 

[rahulkadukar]

Will send you my times for Square 1 and Megaminx as I am going to start soon and I have no Prior idea

 

[Poke]

I have about 8 years of saved .txt logs on my hard drive. That is the good news. The bad news is the controller card appears to be dead, and possibly the drive itself. I'll be getting a new controller card soon to test option 1. If that doesn't work, and the drive is dead, at some point in the next year or so I will send it off to have the discs taken out and the data extracted. This costs a lot of money, but I have a lot of sentimental stuff on this drive (including my practice logs) that I want to save. So when I get this figured out I can send you about 8 years worth of practice logs.

Chris

Hope it works for ya!
My dad's hard drive fired and they couldn't get the data... they even tried freezing it... Maybe you could freeze it and put it in your computer... Or you could leave the easily breakable objects to the experts. :rolleyes:

Probably the second option.

 

[shelley]

I logged my times for about half a year back in 2006. It's certainly not a complete log (it just documents my improvement from 30 s average to 20 s), but would that be useful to you?

On looking at it, I didn't log individual times either. I just recorded the date, best time, worst time and average (of 12).

 

[Cloud_9ine]

Well with not data avalible, I'm figuring it to be logarithmic. Starts improving slowly, then much improvement til a certain y point, then small improvement over time. But I do know of some cubers that make many large steps too.

 

[goldencuber]

Interesting project. I was going to look into something myself when i snapped my wrist playing badminton (since i can't cube for a week, I figured I was going to study some theory, anyway), so good job picking up my n00b slack (LOL I was probably never going to get around to it anyway XD)

My theory was that improvement is a standard exponential (negative?) curve up until we hit some marker X. The progression then changes to a small negative linear function until we hit present day Y.

I've still yet to see why this works, but so far it's worked with everyone in my school who cubes + me. That marker X for me was I think 33.40 seconds, since improvement has been scarce in the past weeks.

It might have something to do with genetics though, since I'm pretty sure some people will progress far faster than a standard curve. If we took a normal distribution of "improvement" over all of the speedcubers in the world i'm pretty sure we'd get exactly what we looked at; some would progress faster than the possible "standard curve" (that is, the equation for "normal" improvement), and some would progress slower...

anyway, my hand is tired (only using left hand right now). I'll type more when i have hte brainpower.

I'm still thinking of ways to write the equation. Maybe your limit has to do with your method?

I have about 8 years of saved .txt logs on my hard drive. That is the good news. The bad news is the controller card appears to be dead, and possibly the drive itself. I'll be getting a new controller card soon to test option 1. If that doesn't work, and the drive is dead, at some point in the next year or so I will send it off to have the discs taken out and the data extracted. This costs a lot of money, but I have a lot of sentimental stuff on this drive (including my practice logs) that I want to save. So when I get this figured out I can send you about 8 years worth of practice logs.

Chris

Hopefully I won't have to get the data next year

I logged my times for about half a year back in 2006. It's certainly not a complete log (it just documents my improvement from 30 s average to 20 s), but would that be useful to you?

On looking at it, I didn't log individual times either. I just recorded the date, best time, worst time and average (of 12).

that would be useful. just make note of it in the email

Well with not data avalible, I'm figuring it to be logarithmic. Starts improving slowly, then much improvement til a certain y point, then small improvement over time. But I do know of some cubers that make many large steps too.

Yeah I was thinking of somthing like that, though I found that times improve fast and then slow down. I'd say the derivative of improvement is logarithmic. I might have done a log regression, but i couldn't get it to work on my calc, so i did power regress.

 

[abr71310]

I'm still thinking of ways to write the equation. Maybe your limit has to do with your method?

I'd hope not; that would mean that Fridrich itself has a limit, and I really see no chance of that happening until Roux becomes so advanced that a cube can be solved in the entire 15 second inspection.

 

[Lord Voldemort]

Would averages over time help too?
If so... I have a log from mid-December

 

[Swordsman Kirby]

that would mean that Fridrich itself has a limit,

Well duh. I don't really see sub-0 averages occuring in the future. (nobody mention negative time solving)

 

[soccerking813]

I'm still thinking of ways to write the equation. Maybe your limit has to do with your method?

I'd hope not; that would mean that Fridrich itself has a limit, and I really see no chance of that happening until Roux becomes so advanced that a cube can be solved in the entire 15 second inspection.

That may actually be possible. Use different algorithms to solve 4 pieces at a time, and they don't affect anything but those four pieces.

I don't have very many sessions saved yet, but I am building them up. I will send them to you someday maybe.

 

[abr71310]

I have a lot of sessions, but i've only been cubing since november of 2008, so i dunno how useful i can be.