Data driven route difficulty ratings

Route difficulty ratings are intensely subjective and regional.

What's the most objective measurement a person can make about difficulty? I think it's a very simple comparison: yes/no to "was route A harder than route B".

My suggestion: for people who tick routes, some kind of opt-in feature that would ask them to compare the difficulty of routes they've climbed. Just yes/no, no rating of how much harder one is from the other.

If there are enough such binary rankings from many climbers/routes/areas, that data could be fed to an optimization problem which could compute a less-subjective per route difficulty rating. Something like find the rankings that minimize the distance to existing difficulty while maximizing the agreement in collected ranking data.

I wouldn't suggest using this to replace existing ratings, but it would be really interesting to augment.

In the unlikely case that MP would want to do this, I'd be happy to make the math/backend part happen, lotta experience around that sort of thing.

I would suggest starting with defining an objective definition of "harder." It likely won't be directly measurable, since if it were, we could just use that.

Big data might allow us to eventually look at a route's ticks and see which people onsighted, flashed, or fell/hung. You could then compare that to their personal rates of sending vs whipping for different grades, and then compare that back to the original route. 

For example, I can usually flash 5.10, sometimes flash 5.11-, and usually hang on 5.11, so if I (and a bunch of similar climbers) are hanging a 10, maybe that means the 10 is stiff.

I am not a data scientist and this is not a well thought out idea, so be gentle with me lol

Alex R wrote:

I would suggest starting with defining an objective definition of "harder." It likely won't be directly measurable, since if it were, we could just use that.

Can't because it's not objective, but it would seem to me like the least subjective measurement, perhaps the best you can do with human subjectivity. I'm actually inspired by the CIE 1931 color space experiments, which were a little more involved than "is A more blue than B" but similar - it's about the best you can do.

Camdon Kay wrote:

Big data might allow us to eventually look at a route's ticks and see which people onsighted, flashed, or fell/hung. You could then compare that to their personal rates of sending vs whipping for different grades, and then compare that back to the original route. 

For example, I can usually flash 5.10, sometimes flash 5.11-, and usually hang on 5.11, so if I (and a bunch of similar climbers) are hanging a 10, maybe that means the 10 is stiff.

I am not a data scientist and this is not a well thought out idea, so be gentle with me lol

That's actually a really good idea, that data would be a little less rich than deliberate comparisons (person whipping on 10- might've been having a hard day and wouldn't want to rate that) but it would be muuuuuch more abundant.

I prefer to ask AI to just tell me I'm a good climber and a person worthy of love.

I just climb the N side of Beacon Rock, bam, I CAN ONSIGHT 10d TRAD NOW!

(rampant sand and helium bagging inspired this thread)

One could probably borrow from the ELO methodology used to rank chess players