UKBouldering.com

Pullups, to 1 armers, to bouldering grade (Read 6515 times)

petejh

Offline
  • *****
  • forum hero
  • Posts: 3281
  • Karma: +351/-16
Talk of 'a grade' being easier for a tall people doesn't mean much does it? Because there isn't a 'standard benchmark grade' in climbing. Unless you're talking about a specific style of route or boulder. For example which 8c+ is easier for tall people, Liquid Ambar? Evolution? Or Era Vella? (ok it's given 9a by some/8c+ by others but you get the point).
Anecdotal; but from talking to various people involved with lattice assessments, and reading their comments on ukb etc., most often the goal of doing the lattice training program seems to be success on a specific European sport route or general success on a Euro sport trip, at one of the well-known French or Spanish crags. Great goal. But it doesn't reflect all of climbing 'at that grade'.
If a Lattice trainee then comes back and reports climbing 'success at grade xyz' and this feeds into the lattice data then it might bias the data towards 'success on grade xyz' for a certain style of route.

Just in the cave, same rock type, venue, you have Left Wall High and Rock Atrocity - both 7C but plenty of people can't do both.

Edit, I see TT beat me to it.
« Last Edit: December 15, 2017, 12:29:40 pm by petejh »

Danny

Offline
  • ****
  • forum abuser
  • Posts: 620
  • Karma: +26/-3
+1 TT for essentially making the 'I think you'll find it's more complicated' argument.

However, I think it's still a useful/interesting exercise to look at these things. Correlations that exist despite a mountain of other explanatory variables are always striking...modern modelling techniques allow us to account for multiple categorical and continuous variables, fixed and random effects, atypical distributions, and the rest. I see no reason why—given enough data—and even with measurement errors, we (not me) can't find out some useful stuff...even if it's purely pub banter material.

In my field (ecology)—which is bewilderingly complex, filled with sampling errors and seemingly random at times—certain things (like body size) explain or correlate with many things.   

nai

Offline
  • *****
  • forum hero
  • Posts: 3254
  • Karma: +143/-0
  • In my dreams
Two more data points for you:

Me 1990s
pullups 25
one armers 2
General bouldering grade 6B (but could do about 8 6c-7A+s at Longridge plus Gorilla Warfare and Green Traverse. One Trick Pony?  :-\)

Me 2000s
pullups 15
one armers 0
general bouldering grade 7A+ (but did 20 odd harder problems up to 7C in a variety of styles)

Perhaps people who can do silly amounts are generally short, light, young people?

Luke Owens

Offline
  • *****
  • forum hero
  • Posts: 1297
  • Karma: +66/-0
    • My Blog
I'm guessing a lot of the people climbing 7C and above who said they can't do a full one armer can probably lock off on one arm though?

slackline

Offline
  • *****
  • forum hero
  • Posts: 19136
  • Karma: +633/-26
    • Sheffield Boulder

Correlations that exist despite a mountain of other explanatory variables are always striking...modern modelling techniques allow us to account for multiple categorical and continuous variables, fixed and random effects, atypical distributions, and the rest. I see no reason why—given enough data—and even with measurement errors, we (not me) can't find out some useful stuff...even if it's purely pub banter material.


And that is the folly...the answer to the worlds "problems"* is not more data.

The well refined and highly successful scientific method which has been used to great effect for several hundred years now, whilst requiring data follows a cyclical, self-correcting process of observation > hypothesis > testable theory (i.e. predicitons) > testing these predictions under controlled conditions > checking results against predictions > refinement of hypothesis (or if enough evidence accrues against it a paradigm shift).

To do this you need to have the correct data, accurately and carefully collated, reducing measurement error as much as possible should be the starting point, otherwise you have Garbage In, Garbage Out.  This thread and the survey I've done in the past is not a good example, the data Lattice collect is a big improvement, but as I think abarro81 or someone at least commented when I'd set up the hanging on edges survey, conditions are fickle, humidity can affect things, someones nutritional/rest state too.

And things like Generalised Linear Mixed Models are brilliant, but they still require knowledge about what variables are important, you don't just throw everything in.  Forwards/backwards stepwise selection has repeatedly been shown to converge on sub-optimal (local maxima) solutions, LASSOs are a bit better.  Principle Components Analysis in my view just masks things.  Neural Nets can do clever things, but are subject to the bias they are fed (big example would be racial profiling because they are being fed biased data reflecting racism in humans).  Even then all these models need training and validation data on which to test the predictive ability of any derived model.

I've said it before but will repeat myself, there will not be money to do this detailed level of research in climbing in a pure research environment.


* substitute inquisitiveness if you like, I just used the most common phrasing of the problem.

36chambers

Offline
  • ****
  • junky
  • Posts: 866
  • Karma: +46/-1
I'm guessing a lot of the people climbing 7C and above who said they can't do a full one armer can probably lock off on one arm though?

I expect this is generally the case. It certainly is with myself and the people I climb with.

Danny

Offline
  • ****
  • forum abuser
  • Posts: 620
  • Karma: +26/-3
I agree with most of that slackers, but we do have *some* knowledge of what's important. I have neither the time nor inclination to do a deep dive on modelling simplification (as it happens I'm doing backwards selection on a lme as I write this). I don't agree on the garbage data point entirely. Plenty of data is like this in the real world: you work with what you have, and—often—come out with huge uncertainties at the other end. But it's still information. Everything on the rosetta mission was n=1. Still useful info. We tagged a single eel making a migration to Sargasso. Still useful. Tree rings are a wobbly proxy for climatic conditions, we can get n=loads and glean some useful info.  Plus, this thread has been qualitatively enlightening from my perspective.   

slackline

Offline
  • *****
  • forum hero
  • Posts: 19136
  • Karma: +633/-26
    • Sheffield Boulder
(as it happens I'm doing backwards selection on a lme as I write this).

I would highly recommend checking out Least Absolute Shrinkage And Selection Operator (LASSO) originally proposed by Tibshirani and implemented in R in the glmnet package.

What are some of the problems with stepwise regression?

References
Altman, D. G. and P. K. Andersen. 1989.Bootstrap investigation of the stability of a Cox regression model. Statistics in Medicine 8: 771–783.
Copas, J. B. 1983.Regression, prediction and shrinkage (with discussion). Journal of the Royal Statistical Society, Series B 45: 311–354.

^ Shows why the number of CANDIDATE variables and not the number in the final model is the number of degrees of freedom to consider.

Derksen, S. and H. J. Keselman. 1992. Backward, forward and stepwise automated subset selection algorithms: frequency of obtaining authentic and noise variables. British Journal of Mathematical and Statistical Psychology 45: 265–282.

Hurvich, C. M. and C. L. Tsai. 1990. The impact of model selection on inference in linear regression. American Statistician 44: 214–217.

Mantel, Nathan. 1970. Why stepdown procedures in variable selection. Technometrics 12: 621–625.

Roecker, Ellen B. 1991. Prediction error and its estimation for subset—selected models. Technometrics 33: 459–468.

^ Shows that all-possible regression can yield models that are too small.

Tibshirani, Robert. 1996. Regression shrinkage and selection via the lasso. Journal of the Royal Statistical Society, Series B 58: 267–288.


And whilst qualitatively interesting you are trying to quantify relationships.  People's weights fluctuate, measurement error from varying devices, reporting of "best" performance all add noise, and whilst the data may be useful in guiding hypothesis formation if you want to tease out whats going on and relationships you do need to collect a large standardised data set over time (repeated measures on individuals) in multiple samples (lots of different groups of people) since repeatability is fundamental to science.  The Lattice plots show little relationship, and as remus noted the fact that the outcome (grade) isn't exact makes it even more challenging.
« Last Edit: December 15, 2017, 03:16:49 pm by slackline »

highrepute

Offline
  • ****
  • junky
  • Posts: 942
  • Karma: +80/-0
  • Blah
This builds on earlier points by TT, PJH and Danny regarding lattice testing...

I've not convinced about the climbing is easier for the tall conclusion from the lattice data - and I'm a short climber so could do with the excuse.

It could easily be argued the opposite way - the lattice tests are easier for the short.

Danny

Offline
  • ****
  • forum abuser
  • Posts: 620
  • Karma: +26/-3
Ta. I'm aware of most of the issues. My factors/interactions are mechanistically meaningful. I'm not going 'all the way', as it were. Back to the one-armers...

Danny

Offline
  • ****
  • forum abuser
  • Posts: 620
  • Karma: +26/-3
This builds on earlier points by TT, PJH and Danny regarding lattice testing...

I've not convinced about the climbing is easier for the tall conclusion from the lattice data - and I'm a short climber so could do with the excuse.

It could easily be argued the opposite way - the lattice tests are easier for the short.

This. Even when we're talking about correlations we can't help but drift into causality.

36chambers

Offline
  • ****
  • junky
  • Posts: 866
  • Karma: +46/-1
This builds on earlier points by TT, PJH and Danny regarding lattice testing...

I've not convinced about the climbing is easier for the tall conclusion from the lattice data - and I'm a short climber so could do with the excuse.

It could easily be argued the opposite way - the lattice tests are easier for the short.

This. Even when we're talking about correlations we can't help but drift into causality.

With regards to gritstone bouldering, you don't need lattice data to conclude that the number of problems that favour taller folk significantly outnumber the ones that favour shorter ones :worms: ;)

nik at work

Offline
  • *****
  • forum hero
  • Posts: 3239
  • Karma: +271/-2
Have i misunderstood or have the lattice chaps not simply said
"For a given max grade shorter climbers tend to be able to do more pull ups than taller ones"
This is clearly very different to saying climbing is easier for tall climbers isn't it?

(Obviously climbing is far easier for cheaty lankmeisters and everyone knows this, I just don't think that is what the lattice brains are saying...)

ghisino

Offline
  • ****
  • forum abuser
  • Posts: 642
  • Karma: +35/-0
I'm guessing a lot of the people climbing 7C and above who said they can't do a full one armer can probably lock off on one arm though?

During the time window where i climbed most of my 7C's i wasn't training pullups or lockoffs at all, and was relatively weak at both. Certainly not holding a one arm lockoff. I would fingerboard or campus more often, and be relatively much better at those.

Interestingly, i've been resuming pullup training lately and the progress seems relatively quick (i got the one arm lockoff very fast),
 probably the muscular structure (fibers) is there, it's just a neuromuscular issue.

i should add that since 99% of my bouldering has been in font, this is quite logic...in font you want to be able to press hard into slopers and heels, to be explosive at times, but the times where you want to lock off statically on a reasonable hold with shit feet are almost non-existent.

I remember more situations where i would have wanted a better lockoff when sport climbing on limestone...


Sasquatch

Offline
  • *****
  • forum hero
  • Posts: 1831
  • Karma: +138/-1
  • www.akclimber.com
    • AkClimber
I'm guessing a lot of the people climbing 7C and above who said they can't do a full one armer can probably lock off on one arm though?

I expect this is generally the case. It certainly is with myself and the people I climb with.

I can't do a 1-arm lockoff either. 

lagerstarfish

Offline
  • *****
  • forum hero
  • Weapon Of Mass
  • Posts: 7948
  • Karma: +659/-10
  • "There's no cure for being a c#nt"
I used to be able to do three one armers

closest I ever came to bouldering 7C was toproping Angel's Share (which means it must be 7B+) and a few ascents of WSS before the pebble went (only topped out once)