search

Cammac7 / bracket-filler

Using ML to win March Madness Bracket
0 stars 1 forks source link

Picking first round upsets #1

Open Cammac7 opened 7 years ago

Cammac7 commented 7 years ago

The average number of upsets in the first round (see first_round_upsets.py) is 8. all-time min is 3, and we've got 2. Should we try to force the algo to pick some upsets?

jasontrigg0 commented 7 years ago

I view it like this: Suppose once a year you bet on 100 coin flips, each of them 60/40 in favor of heads. Then you'd always bet heads for each one. Each year there would be ~40 tails (=upsets), but you still wouldn't want to randomly bet 40 of them to be tails (=upsets), because you don't know which ones will be the upsets.

Cammac7 commented 7 years ago

Ok so my counterpoint to that is that we know overall there will be ~40 tails, but not every coin is 60/40. If two teams are incredibly close in our elo chart (like, #20 and #21) that's more likely to be an upset then #20 vs #2. So would it be smart to pick some upsets, whichever ones we think are most likely?