UFC 158 Picks and Bets using a Mathematical Model

MMAPlayground.com » MMA General » UFC Forum » UFC 158 Picks and Bets using a Mathematical Model
fightpredictor
3/13/13 8:38:08PM
How funny is Nick Diaz? I realize a lot of people hate him, but he's very entertaining. Just swearing like mad in the pre-fight press conference.

Anyway here's my picks and bets based on my statistical model:

St Pierre over Diaz. 1 unit at 1.23 to 1 (-450)
Hendricks over Condit. 1 unit at 1.77 to 1 (-132)
Ellenberger over Marquardt. 1 unit at 1.62 to 1 (-160)
Ring over Camozzi. 1 unit at 1.89 to 1 (-112)
Elkins over Carvalho. 1 unit at 1.5 to 1 (-201)
Miller over Mein. 0.5 units at 3.52 to 1 (+352)

Who do you guys see winning this weekend?
prozacnation1978
3/13/13 9:50:38PM
I got three of your 6

Rings wins two should have been losses
Camozzi got momentum he looked good against Luis cane

Condit vs Hendricks is pretty much a toss up
I doubt johnny finishes condit but I.can see condit finishing him.

My only question is why u have a model of 6 fights only
There are 12 fights on the card
fightpredictor
3/14/13 11:46:53AM
Unfortunately, my prediction model is only set up for when fighters have at least 3 fights recorded on fightmetric.com. This is to acoount for things like a fighter debuting in the UFC and dominating their first fight so they become overrated based on their fightmetric stats.

It allows for a better sample of data(3 fights).
airkerma
3/14/13 2:22:40PM
Can you better explain your model? Or provide your algorithm? I enjoy maths as much as the next guy, but it looks like your model just picks the betting favorites.
edit: on that topic, happy Pi day!
jay98107
3/14/13 5:51:42PM
I agree with all of them except Hendricks over Condit. Not bad.
fightpredictor
3/14/13 8:34:52PM

Posted by airkerma

Can you better explain your model? Or provide your algorithm? I enjoy maths as much as the next guy, but it looks like your model just picks the betting favorites.
edit: on that topic, happy Pi day!



Thanks haha. And sure, the basic underlying model is a linear regression based on a whole bunch of data (takes about 2 hours to collect for each event). When I run it through a test where I randomly remove observations and then test the accuracy it shows up as around 80% accurate historically.

Obviously, 80% historically doesn't mean 80% in the future, but it is a good indication that in the long run the model will be making a profit. If you check out my website, you will see that the model has been successful at picking underdogs as well. It is 10-3 in the last two events.
prozacnation1978
3/15/13 5:35:19PM
That's cool that u made a mathematical model
I'm into calculation and stats myself for hobbie only