- Unsupervised Learning
- Posts
- The GT Rating System
The GT Rating System
It is difficult to predict a given person’s success. Whether in a work setting, or just generally in life, the problem with prediction is determining which characteristics are predictive.
Happily, there has been progress in this area. A pair of scientists, Duckworth and Dunn, have been studying a concept called grit for a number of years now.
The scientists found that a simple rating of grit for multiple types of people was more predictive of success than any other measurement available, including wealth, IQ, SAT scores, etc.
Talent is more familiar to us. It basically represents one’s ability to be exceptional in a particular area.
The GT System is combines these two attributes into a composite score in order to predict an individual’s success.
Each person starts as a 5/5, meaning they have a five score in Grit, and a five score in Talent. From there, scores are adjusted based on the each line in each section.
Legend
Here are some descriptions for the various rating criteria:
ProjectsA major project that has over 100,000 direct users per year. Examples might include a book, a website, a software program, a mobile app, a band, etc.
Elite TesterThis is a peer rating, whereby most everyone on the team accepts that the person is an elite tester.
Minor CreatorThis is someone that has put out a lot of small things, e.g. little scripts, tutorials, talks, small applications, etc. They produce, but not on a large scale with big projects.
Major CreatorThis is someone that creates larger projects, complete with a large userbase, an announcement, a website, a talk, etc. Major projects, artistic creations, etc.
ExceptionalPerhaps the most subjective on the list right now, this is a peer rating whereby everyone who interacts with the person thinks they’re phenomenal in one or more ways. It seems more subjective than it is. When people have this, it’s obvious. The key is to ask people who are themselves major players.
[ NOTE: Attributes do not stack, so you only get one bonus for Twitter, one reduction for weight, etc. ]
Once scores have been calculated, you take the scores for each and rank them on the grade scale like so:
You then take the lowest score of the two, so if someone ends up with a B in Grit, but a D in talent, they are given a D ranking.
This is consistent with my experience and the experience of others I’ve spoken. The idea is that you are limited—over long periods and many different types of projects—not by your strengths, but by your weaknesses.
So if you’re extremely disciplined with a great work ethic, but not too creative or talented, that lack of creativity holds you back. Conversely, if someone is brilliant but can’t seem to ship anything, that person is limited by being unable to finish things.
The result, which should be both intuitive and shown in the data, is that true A and B players have high marks in both talent and grit.
The goal of this system is to be able to quickly rank people who you may want to work on something with. This could be at your main job, or a side project, or any number of things.
The idea is that you could, when time is a factor, refer to someone as an “A Player”, and have that actually mean something. And if you say someone else is a “C Player”, with low Grit, for example, the person you’re talking to can know exactly what they’re getting into by taking them on.
Again, I cannot stress enough that the system must be shown to be effective by comparing ratings to actual performance. There are a thousand rating systems out there, with arbitrary, biased, and fetishized ranking criteria, that fail to predict anything.
The goal of the GT Ranking system is to be truly predictive, as bourne out by data. This requires doing the hard work of comparing performance to ratings, as well as working hard on the list of rating criteria.
It could turn out that the system doesn’t work for whatever reason, and that would be fine. I’m willing to try, however. Let me know if you’d like to help.
Any system of this type is bound to be imperfect. The question is whether it can be shown to have predictive value with respect to performance in a particular role. This requires comparing these ratings beforehand to performance attained later, which is an ongoing project.
I would love input on more attributes to add, both for Grit and Talent, and in both directions (positive and negative).
This attribute list is currently tuned for my use; you’d have to adjust it accordingly.