We aim to predict the difficulty grade (V0 - V17) of climbs when given only a representation of the route. These routes will be sourced from a number of climbing boards, these boards are fixed width and height walls that have identical holds.

(For Kilter Board) Users are able to create routes by selecting certain holds from the board. Green holds represent starting holds, which climbers must start with their hands on. Orange holds are footholds only. Blue holds are the intermediate positions, available to both hand and foot. Pink holds are the ending holds, where climbers must have two hands on to finish the climb. Our goal is to “grade” or classify given training board routes on their difficulty (V0 - V17) given the data above.

Each route is represented by a string. For example, p1083r15p1117r15p1164r12p1185r12p1233r13p1282r13p1303r13p1372r13p1392r14p1505r15 represents the climb above. This string can be divided into each hold starting with p.

The p{4 digit number} represents a hold ID which corresponds to the holds x and y coordinates, and the r{2 digit number} represents a color (green for start, blue for hand, etc.). The string-based route can be represented as a sparse matrix.

Given the string-based route, we plan to use a similar approach as Recurrent Neural Network for MoonBoard Climbing Route Classification and Generation and preprocess our data to capture the nuances of a route's difficulty. This would proposedly involve calculating a difficulty feature vector for each climb based on a number of factors, including the difficulty of the route’s path, the difficulty of the route’s holds, foot placement along the climb, wall angle, distance between holds, and more.

Week 5:

We have a working representation of data, with the ability to sort and parse/visualize our climbs.

Goal for this week is to find a more friendly way to represent data. Im thinking like a single value for EACH hold.

May 6th:

added difficulty grades hard coded to coordinate values. I need to double check with andrew about confirming the difficulty bc these are random. Also, double check regarding def normalizeDifficulty(difficulty).

Also figure out how to preview routes on your own Malachi

Week 6:

Normalize values for hold difficulty, 0-1. - Calculating difficulty = (distance from previous move [will likely need some sort of traversal algorithm] * hold difficulty const * board angle const) + colors ? - Consider average ape index to decide distance difficulty. Add features. - Total moves, spatial factors (distance) Create dataloader. - Sparse matrix and features, label