Discussion: Call for Routes as Grade References

  • Started: 6 weeks ago on Tue 12th Jan 2021

Public discussion This is a public discussion in Australia.

started this discussion 6 weeks ago.

Call for Routes as Grade References

Some of you already are aware of our new grAId feature, a piece of software that helps grading routes based on ascent data that is currently being implemented on theCrag (check the respective route pages in AUS to see it or read more in this article by VL Magazine.

For verification purposes we are looking for a couple of routes that most climbers would consider to be so-called grade references - meaning their grading is undisputed, not really influenced by climber morphology and of course repeated by many.

We would appreciate to hearing your proposals by simply posting the grade and link to the route here as a forum comment. Eg.

  • 21 Kachoong (

Please come up with as many as you wish, grouped by sport climbing, trad climbing and boulder routes.

Thanks for helping us with this effort!!

replied 6 weeks ago.

SE Queensland Boulders

White Rock Conservation Area

V8 Tuscan Raider

V6 Shinkansen

Cedar Creek

V4 Sit Start

V1 Cosmic Dance Stand

I have only reference to mainly Brisbane and Sunny Coast bouldering, but grading seems to be all over the place. Potentially better areas of QLD for benchmarking purposes but it seems difficult to find boulders with many, many ascents to suit. Open for more discussion on bouldering in QLD.

JSBC replied 6 weeks ago.

Canada Sport:

Definitely “Telltale Heart p2” 5.13A or 28. Adam Ondra after onsighting it said “never ever 5.13a)

Bobloblaw 5.12D or 27. Hardest 27 in the world. A v7 dyno to run out steep sustained 27 climbing. Mega classic!

Atlantis 5.12C or 26 every style of Climbing in 30ms to a burly roof to clip the chains.

Bugs life 5.12B or 25 - one of the pumpest lines around. A long run out boukder problem 30ms above the deck on the side of a massive Candian Rockies mountain!

The Masque 5.12A or 24 - Megos used this and the 27 extension above it as his warm up for “Fight club” it was cool watching him clip 3 draws in 20m’s on it.

Australia Sport:

Spoonman 13A or 5.13A

Resistance is Futile 27 or 5.12D

False Signage 26 or 5.12C

Wet jigsaw puzzle 25 or 5.12B

Honed and Buffed 24 or 5.12A

Boulder Canada;

Mantra V8

Anubis V7

Minor Threat v6

Timeless v5

Supafly v4

Crackhead V3

Slingshot v2

Cheese Grater V1

Boulder Australia:

Wild Honey V8

If The Shoe Slips V7

Pendeltag V6

Ric Whites V5

Tree fall V4

Explorers Club V3

Lazarus V2

Dyno V1

Funny enough all these routes and problems would be some of my favourites in the two countries. I’ll edit this comment later on with all the links - as I’m just eating desert.


replied 6 weeks ago.

Can I also suggest that grade's added by 'grAId' are done under an account called grAId so when you stumble across it without context and click on the name you can get a bit more info.

Mark Gamble replied 6 weeks ago.


Spoonman 13A? Is this a different "Spoonman" than the one at Coolum?

JSBC replied 6 weeks ago.

5.13A is 28 . North American YDS grading System to Aussie ewbacks. Climbing school with Jimmy

replied 25 days ago.

Thanks for all your contributions! Please add more if you know of any.

JSBC I assume Bobloblaw 5.12D or 27. Hardest 27 in the world. A v7 dyno to run out steep sustained 27 climbing. Mega classic! is not a good reference as it is called the hardest 27 in the world

Will Monks replied 25 days ago.

As I've previously mentioned the outcomes produced by graid are not yet credible, at least for the Ewbank (Australian) scale. The Ewbank scale has always had reference routes, namely those graded by John Ewbank himself, whether his own FAs or by inclusion in his guidebook. Janicepts, Colosseum Corner and myriad others. This is where you find the fundamental "recipe" for Ewbank grading, not in graid. I haven't gone through the Ewbank era routes in detail but some spot checking indicates that graid only ever upgrades them. Maybe there should be some give and take (JE was only human) but it seems clear that graid is SYSTEMATICALLY getting it wrong relative to THE defining routes.

And, as I've also pointed out previously, Graid also clearly has major flaws at higher grades, as amply illustrated by the violin plot in the paper which has a mean/median which clearly deviates upwards away from Ewbank by multiple grades. I.e. graid clearly fails to conform to Ewbank at both the original "low" grades and also at newer high grades.

So currently, to rely on graid is to effectively redefine the Ewbank scale.

All that being said, if graid can be brought into conformity with the Ewbank scale on both these issues (and, frankly, numerous others) then it might have potential to be an interesting tool.

I continue to be baffled with how one human edit of one grade of one route can get people up in arms ... but no one has a problem with an inaccurate system's outputs being published onto tens of thousands of routes.

replied 25 days ago.

Finding reference routes is part of the effort to fine tune the algorithm.

Monty Curtis replied 24 days ago.

I don't think we will find reference points by randomly asking people on the internet.

replied 24 days ago.

This comment has been removed.

Lewis replied 24 days ago.

I’ve noticed recently that literally all jamming crack climbs have overestimed grades. For example from my recent climbs I’ve seen any climb relying on jamming has the AI grade 4 grades harder:

Where angels fear to tread 17 (Graid 21)

Lift Girls Lament 22 (Graid 26)

Harlequin 18 (Graid 22)

Beowulf 18 (Graid 22)

Where angels fear to tread, Harlequin and Beowulf are all ‘classic’ jamming crack climbs of the grade and in my opinion think they should be added as grade references.

This type of overgrading is likely made when strong climbers have poor jamming technique and fail to send a jamming climb, which is likely very often, sending the AI grade up.

Not sure if there is the capacity in the system but it would be a major upgrade to search for keywords in the climbing comments for each particular climb eg, jamming, fingers, hands, fist, offwidth. (crack is a poor choice as it doesn’t indicate jamming) so the system can flag that the style of climb includes jamming, you could then separately rate jamming climbs in the AI grade knowing that some people are stronger in different techniques.

This may only be possible to implement if each climber is profiled into a dynamic system where they have different ability levels for different styles eg, slab, overhangs, face, jamming. If you're looking for an upgrade, this would be a good one.

rickau replied 24 days ago.

Maybe the ewbank grading system isn't fit for purpose for sport routes. Correct me if I'm wrong but wasn't Ewbank very much anti bolting? Did he put up any sports routes as a point of reference?

Maybe the Ewbank system is more appropriate for trad and mixed routes while sports routes should use the French system.

Taner replied 13 days ago.

Lewis Yes! Same thing with slab climbs. Classic stereotype comes to mind of a gym climber that can campus hard but struggles on techy/slab climbs. Very anecdotal example but I've seen a climber capable of climbing burly mid 20s retreat on a bolted 15 slab at mt buffalo.

Main suggestion that I think is technically not so hard to implement with GrAId would be to factor in rock type as a weighting. If someone hasn't climbed much granite in awhile, and falls on a slab climb or 'Where Angels Fear to Tread', then their ascent shouldn't affect GrAId as much, or there should be a buffer between their current CPR or whatever is used to measure against.

I also like Lewis's idea of profiling climbers.

Maybe each "grade reference" climb can be manually reviewed and identified as a certain type of climb, for the purpose of considering different climbers' strengths and weaknesses.

replied 13 days ago.

Given the 'Hueco' Grading system, or V-Grade System - I'd suggest that Hueco Tanks is a good place to start for benchmark grades for bouldering.

V2 - Nobody Gets Out of Here Alive V4 - Moonshine Roof V5 - Dragonfly

V10 - Full Service - This is THE original Hueco V10 and should be considered the gold standard for V10.

Also, do you guys have a link to Dean's full academic write up on how the algorithm works - I'd love to learn more about the inputs to the model here.

replied 12 days ago.

Doug replied 11 days ago.

From the VL article it looks like the algorithm needs more work:

Loop the Loop (25), Shipley Upper , NSW Rubber Lover (25), Centennial Glen, NSW

are listed as hard ticks. I would have thought general consensus over a long period of time from people with a wide variety of experiences is that this is about benchmark.

Issues with the algo will be current thecrag user bias to accept upgrades. From personal experience I rarely see anyone log a route softer on thecrag; they may comment to such but no actual downgrade. Thecrag maintainers also seem eager to upgrade based on personal experience. This can quickly introduce a softer grade standard.

This is where guidebooks are at an advantage. Grades that cannot easily drift over a short period of time based on a wide enough selection of critical opinion from an experienced user-base across multiple eras, styles and areas. It will be difficult to capture this quality of opinion, from what may already be a biased thecrag user-base?

Showing all 20 messages.

You are not part of this discussion.