Greater Toronto Regional West 2013

[Gregor]

Beat ya , should keep it to that thread.

[2185Bilal]

Originally Posted by Gregor

Beat ya , should keep it to that thread.

well that was for last year, so idk LOL

[Gregor]

Originally Posted by 2185Bilal

well that was for last year, so idk LOL

This was last year's thread.

[2185Bilal]

Originally Posted by Gregor

This was last year's thread.

ohhh okay my bad

[omgnocomms]

It appears to be working now.

http://www2.usfirst.org/2013comp/Eve...chresults.html
http://www2.usfirst.org/2013comp/Eve.../rankings.html

[shuhao]

So I've been playing around with the data and watching the competition all day and wrote a quick script to "predict" winners of each match. So far it is working quite well, maybe it will be of use to people (Note: this is not my team's analysis, I'm not with them right now as I've graduated high school):

https://docs.google.com/spreadsheet/...UE&usp=sharing

Edit: Does anyone know which matches had robots tipping over?

Edit 2: Simulated rankings are now available.

[Racer26]

Originally Posted by shuhao

So I've been playing around with the data and watching the competition all day and wrote a quick script to "predict" winners of each match. So far it is working quite well, maybe it will be of use to people (Note: this is not my team's analysis, I'm not with them right now as I've graduated high school):

https://docs.google.com/spreadsheet/...UE&usp=sharing

Edit: Does anyone know which matches had robots tipping over?

Edit 2: Simulated rankings are now available.

What's it making predictions based on?

[shuhao]

Originally Posted by Racer26

What's it making predictions based on?

Public available data for the regional so far. The current rankings and standings and some "heuristics" (really bad guesses, maybe not that bad, i don't really know).

[Racer26]

Well, that didn't answer my question at all.

I'm interested in your 'heuristics'. OPR is a popular choice for such things.

[shuhao]

Originally Posted by Racer26

Well, that didn't answer my question at all.

I'm interested in your 'heuristics'. OPR is a popular choice for such things.

No where near as complicated as OPR (I just did something simple as I created the spreadsheet in ... 15 minutes?). Used all local data (just from this competition, not scouting data... so everything is binned).

Very simple heuristics: Pretty much just added up a normalized version of each team's points as listed in their standings and then added the teams' together.

So this is probably wrong, but then my cross-validation set did fairly well.. (although i don't have a lot of that either) so I'm rolling with it for now.

I wish I had our scouting data.. but then I'm 500km away on a computer... so..

[Racer26]

Do you mean to say you took a team's mean score, and summed those for teams in a match to come up with the predictions?

If so... seems... wrong, but its predicting with reasonably good accuracy.

[shuhao]

Originally Posted by Racer26

Do you mean to say you took a team's mean score, and summed those for teams in a match to come up with the predictions?

If so... seems... wrong, but its predicting with reasonably good accuracy.

Pretty much and IDK if the second statement is a statement or a question.?

It's predicting with reasonably good accuracy so maybe something is right. Proceed with caution. I'm looking around for more data... but I don't want to dig through historical records because this is more of a fun project for me to procrastinate on.

The predictions on tomorrow's matches seems to be following what intuitions (mine, some people I know) say.. so maybe there's some merits to this.

We shall find out tomorrow.

[nicholsjj]

It did perdict the 2056 loss so you have something going here.

[shuhao]

Originally Posted by nicholsjj

It did perdict the 2056 loss so you have something going here.

Against 1114

[Jash_J]

Originally Posted by nicholsjj

It did perdict the 2056 loss so you have something going here.

I think he's talking about match 48. Where 1310 was predicted to win against 2056 by the system and actually did.