paper: 2015 FRC Championship History

[Adam Freeman]

Originally Posted by Andrew Schreiber

I recall he explained it because it meant that things that happened more than a student cycle ago didn't count for much.

Edit: That's why I've stolen that number for a lot of things I do too.

Yep, says that exact thing right in the notes tab of the spreadsheet:

"An annual derating factor was applied to all past data. This causes the data from past years to have a diminishing effect on the current team standings. A weighing factor of 66.66% was used. This means that the current season carries 1/3 of the weight of a team's entire calculated history over long periods. This factor was determined somewhat subjectively, and is largely based on the fact that a team's student population will turn over completely every 4 years. Thus we do not want events from over 4 years ago having a significant impact on the current cumulative rankings."

[Bennett548]

I understand the reasoning behind the diminishing effect, I was just curious about the value chosen. With the current number ~80% of a team's ranking comes from the most recent 4 years.

I don't think there is a number everyone can agree upon. I would use a bigger weight that 66%, so that older years have more weight. It seems to me that mentors and sponsors are a big part of a teams success, and that they have a slower turnover rate.

[GeeTwo]

Originally Posted by Bennett548

I understand the reasoning behind the diminishing effect, I was just curious about the value chosen. With the current number ~80% of a team's ranking comes from the most recent 4 years.

I don't think there is a number everyone can agree upon. I would use a bigger weight that 66%, so that older years have more weight. It seems to me that mentors and sponsors are a big part of a teams success, and that they have a slower turnover rate.

Also, not all teams are four years. There are many 3 year high schools, and schools with FTC teams as well. As I recall from The New Cool, 1717 members only compete one year. Others, such as the Girls of Steel, have an 8th grade apprentice program and are essentially a five+ year program. Probably most teams which mentor a "farm system" of FLL/FTC would tend to have more than a four year decorrelation time. All that said, -33% per year sounds reasonable to me.

[IKE]

Originally Posted by Bennett548

...snip...
I would use a bigger weight that 66%, so that older years have more weight. ....

Making it larger makes it significantly harder for "up and comers" to get their value to a reasonable level. Follow 1678 to see a good example (or 3476, or 1640. It takes several years of high performance to get there. another perspective, look at the "total history chart" tab, and you will get a feel what it looks like if you do not use any historical derate.

It is a neat exercise to do, and I would highly recommend playing around with it. I remember talking to Jim about this several years ago when he started publishing it. He played around with a few different values and 2/3 seemed to be "about right".

[Peter Matteson]

I was just browsing and noticed an error or inconsistency in the data.
177 is shown on the 15 year tab as division finalist in 2001 not division winner, however on the 2001 tab the data is correct showing the Galileo win.

The point values also look correct in the 15 year data sheet.

Also as I was about to submit this I noticed 125, 365, 71 & 294 are listed as WF instead of World Champ for 2001, also with the correct point value for world champ. 111 is however correctly listed.