In 2007, I came up with a bracket contest where you only pick one team from each seed and get seed*round scoring.
In 2009, we started doing Calcutta. Still doing it because it's so great.
In 2010, we tried the Eliminator.
Let's try something new.
It's your classic bracket contest but with custom scoring that rewards underdogs.
But how to do it...
I thought of 5 ways off the top of my head:
Seed X Round
Seed X increasing value per round
Seed X Round + Constant Points
Seed X Round + Increasing Points
Seed + Increasing Points
Warning, this about to get a little hairy.
- - -
To do this I'm going to look at 1st round games, Sweet 16 games, and the Championship. (1st, 3rd, 6th round)
Hypothesis #1: Picking all 15s to win and getting one right should be worth more than picking all 2s to win and getting 3 right.
So these are first round games. So round multiplier is 1. Getting one 15 right at the very least gets you 15 points.
Getting your three of the 2s right, could get you 6. 15 to 6 seems like a not bad ratio for this simple matchup. Let's come back to this one.
Hypothesis #2: Picking a 6-seed to win a Sweet 16 game means knocking off a potential 3 and 2. Lets compare the value of picking all 6s to win vs picking all 2s.
This is a 3rd round game. If we're just doing seed x round, (and only one comes through) that's 18 points for that. If you picked 2s (and got three right) it's also 18 points.
Let's come back.
Hypothesis #3: Correctly picking a 3 to win it all should be worth more than picking a 1. But picking a 1 correctly to win it all should be worth about as much as correctly picking a 15-over-2 first round game.
This is where seed*round falls apart. 1*6 isn't enough reward for getting the championship right.
- - -
I found this guy saying he uses Seed X Round + Increasing Points , with the Fibonacci sequence to determine the increasing points as such: 2-3-5-8-13-21.
Let's try this with our numbers.
Getting one 15 gets you 17 points. Getting three 2s gets you 12 points.
Getting one 6 gets you 23 points. Getting three 2s gets you 33 points.
Getting the 3 in the title gets you 39 points. Getting a 1 in the title gets you 27 points.
I'm not in love with this either. The championship is fine, but it rewards chalk too much in earlier rounds.
- - -
What if it was just Seed + Fibonacci points?
Getting one 15 gets you 17 points. Getting three 2s gets you 12 points.
Getting one 6 gets you 11 points. Getting three 2s gets you 21 points.
Getting the 3 in the title gets you 24 points. Getting a 1 in the title gets you 22 points.
- - -
Let's try Seed X Value per round. For starters, I'll use 1-2-3-5-10-20.
Getting one 15 gets you 15 points. Getting three 2s gets you 6 points.
Getting one 6 gets you 18 points. Getting three 2s gets you 18 points.
Getting the 3 in the title gets you 60 points. Getting a 1 in the title gets you 20 points.
That's pretty good. It certainly incentivizes people not to pick chalk...let's keep going. I'll call this Model A.
- - -
Let's go back to seed x round + increasing points, but change the sequence to 1-1-2-4-8-16.
Getting one 15 gets you 15 points. Getting three 2s gets you 6 points.
Getting one 6 gets you 20 points. Getting three 2s gets you 24 points.
Getting the 3 in the title gets you 34 points. Getting a 1 in the title gets you 22 points.
That's not bad either. I'll call this Model B.
- - -
Let's try out the system that Drew Magary proposed yesterday, with a new wrinkle (seed differential)
Getting one 15 gets you 14 points. Getting three 2s gets you 3 points.
Getting one 6 gets you 10 points. Getting three 2s gets you 18 points.
Getting the 3 in the title gets you 17 points. Getting a 1 in the title gets you 15 points.
This system really reward first-round upsets. Which is interesting because going for them could compromise your later picks. Though a 12 over a 5, would become probably too valuable. Also, if you pick a 12 to advance to the sweet 16, but they beat a 13 seed in the second round, you don't get any bonus for that. I think seed scoring is better than seed differential.
- - -
So I have Model A and Model B. Let's introduce two new tests.
Test #1. Comparing the relative value of getting a 13 to win a 2nd round game.
Model A: The 13 seed winning their second game gets you 26 points. (If you had correctly picked three 4s to advance you'd get 24 points.)
Model B: 27 points for the 13 seed. Three 4s gets you 27 points.
Test #1 outcome is basically the same.
Test #2. And getting a 4 to win a 4th round game.
Model A: Getting a 4 to win their 4th game gets you 20 points.
Model B: Getting a 4 to win their 4th game gets you 24 points.
- - -
Conclusions
All things considered, Model A and Model B are both good.
But I prefer Model A. It's simpler which is more important than you might think. And it encourages even more boldness.
Of course, we'll have to try it out. And then refine if we like it.
No comments:
Post a Comment