pipkiksass
Veteran
Interesting talk of improving averages vs. singles. I've found that my averages improve more consistently, but I've actually had a fair few PB singles recently, as my averages have improved.
I think there's two things going on here - improvement and consistency.
Timed solves make you more consistent, i.e. reduce the variance in your times. No matter how many timed solves you do, you won't get faster, you'll just get more consistent. With consistency comes improved averages. For example, my PB Ao12 is 15.96, and that was 12 solves all between 15 and 18 seconds - no amazing singles, no terrible ones.
What tends to screw up an otherwise great average is a terrible single - if you average 20 seconds and get a 30 second single in an average, that makes the 24/25 second solve you do later in the average count. Without the 30 second single, the 24/25 is the worst, and is eliminated from the average. Hence, more consistent = better averages.
Statistically speaking, the likelihood of getting a PB single with purely timed solves is very low, but the probability of improving your averages is very high. In fact, the more consistent you get, the less likely you are to improve your PB single! Maths, eh?
When you actually improve, the mathematical mean of your solves becomes lower. You become "better" at some part of the solve, whether it's lookahead, cross>F2L transition, LL algs, or what-have-you. This happens through specific, targeted practice. As the mean of your solves is lower, the "outliers" i.e. the solves that are n standard deviations from the mean, will also be faster.
What you also tend to find, as you improve, is that the variance in your times goes up, albeit temporarily. This is generally because sometimes your new habits abandon you, and you revert to the old (bad ones). It's often during these periods of dramatic improvement, and increased variance, that you will get that ridiculous PB single.
So, in short:
deliberate practice => improvement => PB singles;
spamming timed solves => consistency => PB averages.
I think there's two things going on here - improvement and consistency.
Timed solves make you more consistent, i.e. reduce the variance in your times. No matter how many timed solves you do, you won't get faster, you'll just get more consistent. With consistency comes improved averages. For example, my PB Ao12 is 15.96, and that was 12 solves all between 15 and 18 seconds - no amazing singles, no terrible ones.
What tends to screw up an otherwise great average is a terrible single - if you average 20 seconds and get a 30 second single in an average, that makes the 24/25 second solve you do later in the average count. Without the 30 second single, the 24/25 is the worst, and is eliminated from the average. Hence, more consistent = better averages.
Statistically speaking, the likelihood of getting a PB single with purely timed solves is very low, but the probability of improving your averages is very high. In fact, the more consistent you get, the less likely you are to improve your PB single! Maths, eh?
When you actually improve, the mathematical mean of your solves becomes lower. You become "better" at some part of the solve, whether it's lookahead, cross>F2L transition, LL algs, or what-have-you. This happens through specific, targeted practice. As the mean of your solves is lower, the "outliers" i.e. the solves that are n standard deviations from the mean, will also be faster.
What you also tend to find, as you improve, is that the variance in your times goes up, albeit temporarily. This is generally because sometimes your new habits abandon you, and you revert to the old (bad ones). It's often during these periods of dramatic improvement, and increased variance, that you will get that ridiculous PB single.
So, in short:
deliberate practice => improvement => PB singles;
spamming timed solves => consistency => PB averages.
.
.
