Runboard.com
You're welcome.

~TALK SMACK IN EXILE~
Revolutionizing the way we... uhm, talk smack.


runboard.com       Sign up (learn about it) | Sign in (lost password?)

Page:  1  2  3 

 
NINJA Terp Profile
Live feed
Blog
Friends
Miscellaneous info

Administrator

Registered: 07-2007
Posts: 1183
Karma: 55 (+85/-30)
Reply | Quote
Re: First Weekend Predictions


Thanks for the clarification on the #1 to #18, Dave.

To further clarify, I never said that chances of Pitt NOT winning their group was small. My statement was Pitt AND UGA not winning their groups. In my model results, the chance of NEITHER Pitt or UGA winning their pool is 1% [(1-0.162)*(1-0.988)].

Now, mind you, my gut tells me that Pitt has a better than 16% chance to win their pod which makes said 1% even smaller (I believe that UGA's 98% chance is pretty accurate).


Small note - these percentages may change as I run through the tourney as a whole 20k times and update this post.

Last edited by NINJA Terp, 5/Aug/15, 3:13 pm
5/Aug/15, 3:12 pm Link to this post Send Email to NINJA Terp   Send PM to NINJA Terp
 
NoVADeac Profile
Live feed
Blog
Friends
Miscellaneous info

Registered user

Registered: 08-2009
Posts: 72
Karma: 2 (+5/-3)
Reply | Quote
Re: First Weekend Predictions


Would be interesting to go back through the last three years and see how many 1 seeds failed to advance to the second weekend (or conversely, how many 3/4 seeds advanced).

Also of those 3/4 seeds that advanced, how many were historically strong teams that tanked the regular season vs. upstart teams or true surprises (obviously more subjective than quantitative).

Last edited by NoVADeac, 5/Aug/15, 3:25 pm
5/Aug/15, 3:24 pm Link to this post Send Email to NoVADeac   Send PM to NoVADeac
 
CAN Archives Profile
Live feed
Blog
Friends
Miscellaneous info

Registered user

Registered: 07-2007
Posts: 85
Karma: 16 (+17/-1)
Reply | Quote
Re: First Weekend Predictions


quote:

NoVADeac wrote:

Would be interesting to go back through the last three years and see how many 1 seeds failed to advance to the second weekend (or conversely, how many 3/4 seeds advanced).

Also of those 3/4 seeds that advanced, how many were historically strong teams that tanked the regular season vs. upstart teams or true surprises (obviously more subjective than quantitative).



Coming soon!

5/Aug/15, 3:28 pm Link to this post Send Email to CAN Archives   Send PM to CAN Archives
 
WilliamTheChild Profile
Live feed
Blog
Friends
Miscellaneous info

Registered user

Registered: 08-2007
Posts: 47
Karma: 0 (+10/-10)
Reply | Quote
Re: First Weekend Predictions


Thank you ... that makes more sense ...

---
---billyk
5/Aug/15, 4:31 pm Link to this post Send Email to WilliamTheChild   Send PM to WilliamTheChild
 
NINJA Terp Profile
Live feed
Blog
Friends
Miscellaneous info

Administrator

Registered: 07-2007
Posts: 1183
Karma: 55 (+85/-30)
Reply | Quote
Re: First Weekend Predictions


I was able to update some code the last few days (with some minimal motivation at work) to simulate the entire tourney from start to finish 20,000 times (this is partly for you, CAN Archives!). Now, you will find numbers and probabilities different from the previous group stage-only analysis because of random numbers... the numbers are random, so they rolled different values, resulting in different runs per inning, resulting in different results.

A few quick notes:

First, formatting. I used Excel's handy-dandy conditional formatting and you'll see values ranging from GREEN (100%) to RED (0%). Now, the RED's bottom out at 0.00005%... if it's ACTUALLY 0%, you'll see the cell with grey italicized text. So, consider UCLA - near the bottom of the list - they advanced out of pool play as a group winner once and as a runner up 3 times. The percentages round to 0%, though they are not actually zero. However, they never advanced to the Round of 16, which is why that 0% is different. Clear as mud? Good.

Second - a clarification on assumptions. I did not update each team's component offensive and defensive performances in each run. Certainly a Cinderella-type team would benefit from this, and would likely have resulted in some medium-tier teams getting a bit more play, but THAT code was a bit more than I could handle this week. Maybe next year?

Third - I realize that there's the first round of the elimination bracket that is missing. My code ignored this - I didn't think that knowing you played (and lost) in the first round would be of any use.

Fourth - I would've liked to put in something to capture all the knockout game combinations (and in which round). I would've loved to see how often certain matchups occured. (*cough* UGA-JMU *cough*)

Fifth - The table is ordered by chance of winning, followed by chance of getting to the title game, followed by chance of getting to the semis, etc.

Image

And... some thoughts:

* Welp. I checked on this last night when it was about 5,000 iterations in and Florida was, by far, the favorite (~22%). Way to go, Random Number Generator - you just screwed over Maryland. I'm not coming to the tourney anymore.

* Florida and VT are third- and fourth-choices to win the title... and they're in the same group. WOOF. Sorry, LSU and Virginia.

* Pitt's chances of gettig to the quarterfinal are nearly twice as much as winning the group (a sample path - defeating WVU, defeating NOVA).

* There are several teams with percentages that drop 60+% between advancing and getting to the Round of 16 (which would require the winning of t least 1 eliminatino game): Penn State (64.2%), Navy (70.6%), Georgetown (67.4%), Cornell (74.9%), NC State (62.5%). A lot of this, I believe, is due to matchups in that second round versus matchups in the - for example, Penn State as a group winner would be re-seeded in the 14-17 range and as a runner-up as a 30+. It's tough to consistently beat some of the better teams.

* Looking at the next jump - from the Round of 16 to the Quarterfinals - here are teams with a larger than 30% drop: Georgia (34.7%), JMU (35.5%), Tufts (36.8%), UNC (32.5%), NOVA (32%), American (33.4%), Illinois (36.7%), Massachusetts (34.5%), Vanderbilt (39.4%). UGA and JMU is easy to explain - they played each other in the Round of 16 a LOT (I wish I had numbers) and they are very close statistically.

* Most random title game that I saw? I happened to check on this before turning in for the night and saw Central Florida over Navy. Not to disparage either team, but the casual CAN fan would likely think this as unlikely as the numbers suggest.

BTW, the last run? Wisco was the Cinderella falling to Florida in the semis. JMU won that run going through - PSU, UGA, Ohio, MD, and Florida. Eesh.

6/Aug/15, 7:54 am Link to this post Send Email to NINJA Terp   Send PM to NINJA Terp
 
NINJA Terp Profile
Live feed
Blog
Friends
Miscellaneous info

Administrator

Registered: 07-2007
Posts: 1183
Karma: 55 (+85/-30)
Reply | Quote
Re: First Weekend Predictions


Also - can someone verify that the image works? I can't access tinypic URLs from my office (I uploaded the screenshot from home). Thanks!
6/Aug/15, 7:54 am Link to this post Send Email to NINJA Terp   Send PM to NINJA Terp
 
DCKnights Profile
Live feed
Blog
Friends
Miscellaneous info

Registered user

Registered: 07-2015
Posts: 33
Karma: 0 (+0/-0)
Reply | Quote
Re: First Weekend Predictions


Image works, Still digesting... but looks awesome!

Last edited by DCKnights, 6/Aug/15, 7:58 am
6/Aug/15, 7:57 am Link to this post Send Email to DCKnights   Send PM to DCKnights
 
CAN Archives Profile
Live feed
Blog
Friends
Miscellaneous info

Registered user

Registered: 07-2007
Posts: 85
Karma: 16 (+17/-1)
Reply | Quote
Re: First Weekend Predictions



Screenshot looks good -- thanks!

0.3%... So you're saying there's a chance...
6/Aug/15, 8:06 am Link to this post Send Email to CAN Archives   Send PM to CAN Archives
 
NINJA Terp Profile
Live feed
Blog
Friends
Miscellaneous info

Administrator

Registered: 07-2007
Posts: 1183
Karma: 55 (+85/-30)
Reply | Quote
Re: First Weekend Predictions


Appreciate the confirmation guys.

6/Aug/15, 8:07 am Link to this post Send Email to NINJA Terp   Send PM to NINJA Terp
 
longshanks78 Profile
Live feed
Blog
Friends
Miscellaneous info

Registered user

Registered: 08-2009
Posts: 281
Karma: -5 (+9/-14)
Reply | Quote
Re: First Weekend Predictions


quote:

NINJA Terp wrote:

Also - can someone verify that the image works? I can't access tinypic URLs from my office (I uploaded the screenshot from home). Thanks!



Not sure why, but I can't see the image. Images from Imgur show up, but not from Tinypic.

I was able to see it on my iphone... strange
6/Aug/15, 8:20 am Link to this post Send Email to longshanks78   Send PM to longshanks78
 


Add a reply

Page:  1  2  3 





You are not logged in (login)
_uacct = "UA-1981555-2"; urchinTracker();