.What is this?
Like any sport competition climbing is full of variables, but in climbing it felt like it all gets boilded down to number of medals, or number of times an athlete has reached the final.
This website is an attempt to scavenge as much data as possible to hunt down these variables and bring them to light.
Not only do we then get to have more interesting conversations about the nuisance of the sport, but we can celebrate athletes for more than just podium places.
- Are certain athletes super consistent?
- Do certain athletes perform better at home?
- Who actually has the strongest fingers?
.What is the whole ELO thing?
Currently all we have to go on is the athletes position (or world ranking score) after a competition.
We like to say that athletes aren't competiting against each other (they are competing against the problems), but then we slot them into a ranked list at the end of the day.
This is not only counterintuitive, but it also hides so much information about the performance. We can't tell how close the competition was, or if the problems were way harder than normal, etc. etc.
My proposed solution was ranking the athletes actual performance, indepentent of the end position.
ELO is commonly used in 1v1 competitions like chess. Given player A's ELO and player B's ELO, you can calculate the expeceted probability of A or B winning.
Obviously climbers aren't really going head-to-head in the same way... but... they are going head-to-head with a problem.
If we can work out the ELO of the problem, then we are cooking.
By carefully selecting an initial ELO for each problem, we can then match the problem against each athlete's attempts, like a chess match, each time the athlete tops a problem they gain ELO and the problem loses ELO.
The opposite happens when the athlete fails to top the problem, they lose ELO and the problem gains ELO.
Even if an athlete doesn't place where they "should", as long as they are topping the boulders they "should" their ELO will reflect that.
From here we can infer a whole world of detail.
- Graph athlete's performance over time.
- Track times when the athlete is over and under performing.
- By capturing the style of problems, we can build a reflection of the athletes abilities.
- And, by looking at the otherside of the match up, we can identify events which have had harder and easier setting.
.How does it work?
The current process is a bit manual. There are a number of different sources available for competition results, which I use to get the raw data in various formats.
It needs a bit of preprocessing before I can feed it into my tooling for generating the ELO ratings and other stats.
For predictions, start lists are thin on the ground, and there isn't any problems to analyse until the competition starts.
I will try to populate predictions with a startlist if I can find one pre-competition. The problem data can then be fend in as the competition progresses.
However, due to this being entirely manual, visual analysis, I need to have the free time to watch the competition live and the problem analysis is biased by the whims of
my attention. So, as with all of this, it is all for a bit of fun.
If you do have any suggestions for improvments, or stats you want to see, message me on Instagram _frnt3_.
