How I Leveraged AI to Assist in Re-Rating the NHL/AHL in Pivot

[xECK29x]

Hello EHM Community,

I wanted to create a post to document the completion of a multi-month project where I leveraged AI to help combine analytic 60+ analytic datasets and create models to help plot attribute values across the NHL and AHL. For years I've been doing EHM database attribute updates by hand, the ratings were subjective and best effort and not evenly grounded to data. As the team keeps automating more data collection and processing to keep the Pivot/TBL databases updated in a way that's sustainable for a small team I used ChatGPT to assist in modeling analytic data.

In the documentation stack linked below you will see how I accomplished this monumental task, one that I hope is sustainable for the future as we can easily update and change or modify datasets and can be applied and scaled for other leagues with robust analytic data such as the SHL and DEL.I've been working with AI technology for a few years now, both personally and professionally, it is a tool that can produce amazing outcomes, but it's very flawed, and is very time consuming to get those good outputs. It will lie to you, hit computational boundaries and degradation points that are not clear to the novice user or make assumptions. I am not a coder, that is why I tasked AI with this, but I do work heavily with code and reverse engineer it for a living.

I wanted to provide this post stack mainly for transparency into the project and how each Attribute was modeled and the data used and to foster continued discussion here, on Reddit and Discord. I am very much open to and looking for feedback from the community. I do have plans to create a version of this for some of the European leagues where good data exists (AHLTracker.com equivalent or better). However I would need help from the community for European re-rates using the same process.

Documentation stack: https://1drv.ms/o/c/3b942c96172742cf/Ig ... Q?e=o98Obh

I am happy to privately provide the code, I will eventually make it fully public on GitHub when comfortable.