Data rules
36 views
Skip to first unread message
Danny
Feb 25, 2012, 4:20:58 PM2/25/12
to Machine March Madness
This thread is for discussing what data is and is not allowed.
Danny
Feb 25, 2012, 4:36:03 PM2/25/12
to Machine March Madness
The official statement of the rule is as follows (we reserve the right
to update this with clarifications as necessary):
Any data is allowed, so long as *no human judgement went into
producing it* (I know this is unclear, because human judgement goes
into any model, but we're going to try to enforce the spirit of the
rule as best as possible). So for example, all of the score and box
score data we provide, RPI, Sagarin, or Pomeroy-based scores are
allowed. Seeding, the results of AP polls, or other human-produced
rankings are not allowed. When in doubt, the rule is that the burden
of proof lies with you to show that no human entered into the loop,
but we can discuss specific cases.
to update this with clarifications as necessary):
Any data is allowed, so long as *no human judgement went into
producing it* (I know this is unclear, because human judgement goes
into any model, but we're going to try to enforce the spirit of the
rule as best as possible). So for example, all of the score and box
score data we provide, RPI, Sagarin, or Pomeroy-based scores are
allowed. Seeding, the results of AP polls, or other human-produced
rankings are not allowed. When in doubt, the rule is that the burden
of proof lies with you to show that no human entered into the loop,
but we can discuss specific cases.
Danny
Feb 25, 2012, 5:27:29 PM2/25/12
to Machine March Madness
Scott writes...
"I think you might want to make exceptions for at least seeding --
since you might well have models which are trained on past year's
seedings & results, etc., and seeding is really fundamental to the
tournament."
My response:
I tend to lean towards the hardline on seeding data and saying "no",
but I realize that I'm probably atypical on this issue, so I'm willing
to compromise if the majority wants to include it, because--as you
say--it is pretty fundamental to the tournament.
Thoughts from others?
"I think you might want to make exceptions for at least seeding --
since you might well have models which are trained on past year's
seedings & results, etc., and seeding is really fundamental to the
tournament."
My response:
I tend to lean towards the hardline on seeding data and saying "no",
but I realize that I'm probably atypical on this issue, so I'm willing
to compromise if the majority wants to include it, because--as you
say--it is pretty fundamental to the tournament.
Thoughts from others?
Reply all
Reply to author
Forward
0 new messages