Quote:
|
Originally Posted by sw293
I would not use these rankings for scouting for three reasons:
1. They are based on incomplete data.
2. An objective formula is a poor substitute for an analysis of a team's performace.
3. You want to rank based on compatibility more than performance when scouting robots, because if you have to pick, you don't want three robots who like to start with the tetra in autonomous or three robots who spend a lot of time in the human loading zone. The rankings look at performance, rather than compatibility.
|
As one of the developers of this system, I have far better reasons to not use this for scouting ...
<RANT>First off, this system needs no justification or rationalization, it is nothing more than mere "Relative Power Index" power rankings (similar to the BCS). I do this for the NFL teams (
http://relativepowerindex.com) and thought a fun port over to US FIRST could prove interesting. it was never intended to be used as an "I am better than you" or "pick me, I'm on the top of the list" ... Statistically, if you look at the NFL rankings for the past 30 years, and compare the rankings week with the following week winners throughout the season, the program is between 30 and 40% accurate at selecting the winner ... lets analyze this a bit more, this means, statistically you have a better chance flipping a coin to determine the winner than using the power ranking ... lets even go a bit further, the home team has a 35% chance of winning (solely on home field advantage), which means if you just select the home team you will be right more often than using the rankings ... and these numbers are consistent for over 30 years of data .. I often mused over the fact, if it's only 40% accurate, selecting the opposite of the "higher ranked" team to win each week would be 60% accurate .. which statistically is significant over this data set ... implying the system is really flipped up-side-down, but I digress
So, to put an end to this banter .. this power ranking is for fun, as a way to measure your team with teams you have not faced, and a way to play with numbers ... but as part of the TechTigers, we will NOT BE GOING SOLELY ON THESE RESULTS for anything ... they help indicate who we'd like to scout further, and who we should talk with as potential partners. It helped us select an alliance at Palmetto, and regardless of the rankings, we would have selected the same teams to go to Finals with. We collect other data throughout the competition ... we'd encourage all teams to stop by and see their "Power Rankings". We will be monitoring and charting the progress during the day Friday and Saturday, but would strongly recommend other methods of scouting to determine alliance partners ... this data is statistics, and by their very nature they are flawed in more ways than this thread can discuss ... but that doesn't stop it from being a fun way to look at data
</RANT>
please use the data for what it is, and don't make it out to be what it's not
thanks
-Allen