538 (Nate Silver) NCAA Tourney Projection

Submitted by Gameboy on March 17th, 2014 at 1:22 PM

Nate Silver has come down from his stat geek heaven to grace us with his own tourney projections.

http://grantland.com/features/fivethirtyeights-tournament-predictions/

Based on uber-stat-geek, Louisville has the best chance to win the tourney, followed by Florida and Arizona.

Michigan has one of the best chances to getting to Sweet 16, but it gets progressively tough from there with 3% chance to win it all (I believe that is better than what we had last year).

If you are looking to fill out your bracket and need help, this is probably good as any.

Comments

French West Indian

March 17th, 2014 at 1:58 PM ^

Considering that even though there are often more than a dozen political parties involved, the "powers that be" only feature two choices and thus, at worst, it's a 50/50 guess.  Add in a few other factors (strenght of ecomony, peace/war, incumbent seeking re-election, etc) and even an idiot like me can accurately predict most Presidential elections.

SituationSoap

March 17th, 2014 at 4:25 PM ^

Er, Silver tends to correctly predict elections down to the district level. It's not that he's correctly predicting the winner, it's that he's correctly predicting exactly how all the congressional races and districts are going to fall. 

 

And he's almost always right...three months before the election. Often times within just a percentage point or two of the final result.

MichiganTeacher

March 17th, 2014 at 7:57 PM ^

There hasn't been a real surprise in a Presidential election since Dewey vs Truman, and that one was predicted by Jimmy the Greek working out of a grocery store parking lot. I'm not impressed.

If Nate Silver could predict what was going to happen on the stock market, or which country is going to have the next violent or nonviolent revolution, or when cancer would be cured, or any other sort of black swan or singularity, then I'd be impressed. That's hard. James Simons, him I'm impressed with. Nate Silver, not so much.

alum96

March 17th, 2014 at 2:55 PM ^

He became famous for his political calculations (nailing every district I believe) not his work on a baseball site.  And again calculating odds of political outcomes is different than sports.  If it was just "math and stuff" the stats guys in engineering / finance / accounting would win every NCAA tourney office pool out there as they go up against those "idiots" in human resources, sales, or the warehouse.   There are reams of data for both but its usefullness is very different.

Just like calculating odds at a casino is different than the stock market.  There is a degree of math and stats to it all.... but they are all different animals.  

Gameboy

March 17th, 2014 at 5:00 PM ^

I don't think you understand probabilities.

Statistics gives you probabilities of an event happening. That is same for elections, investments, sports, gambling, etc. They are all treated the same way.

You may not win a single hand of a black jack based on probabilities, but over hundreds of hand you know what percentages of time you are going to win. By counting cards you can figure out when you have the best probabilities of winning. This is how a bunch of kids from MIT won big bucks in Vegas.

Math is applicable in all aspects if life. Some more predictive than others. But the principle is always the same.

LSAClassOf2000

March 17th, 2014 at 1:42 PM ^

If you sort the table by region, the result is rather interesting. Looked at this way, it show that the East Region bracket would be predicted to be a three-way fight between Virginia, Michigan State and Villanova, whereas all the others have much more clear favorites to proceed to the Championship game, at least in this model. I admit, I sat here thinking either the Midwest or the South might have some good fights in them, but looking at it by region, this would disagree with that perception some. 

 

bronxblue

March 17th, 2014 at 1:44 PM ^

Sounds about right, though the MSU love still seems a bit unfounded.  Putting Iowa State as an 8.2% chance of getting out of that bracket seems way too low considering they've beaten some really good teams all year and just knocked out KU.  I alos think it gives WAY too much credit to Louisville, a team that is playing well now but isn't some juggernaut.  And this probably isn't really the fault of the system, but it basically comes up chalk-ish save for that MSU bracket.  Yes, if you are a top-4 team you are the prohibitive favorite to advance.  Congrats - they had to run a lot of calculations to spit that out.

mistersuits

March 17th, 2014 at 2:12 PM ^

People are confusing 2013 Big East Louisville and 2014 AAC Louisville.

2013 Strength of schedule: 9.
2013 Non-conference SOS: 100.

2014: Strength of schedule 108.
2014: Non-conference SOS: 299.

Of any team seeded 11 or higher (44 teams) in the tournament Louisville's SOS ranks 43rd. The only team with a worse strength of schedule with a tournament seed higher than 11 is.... Wichita State.

Louisville has pushed to the maximum the margin-aware methods of efficiency ratings by absolutely murdering inferior competition. Here's their record against NCAA tournament teams this year:

Non-conference:
(6) UNC L 84-93
(14) LA-Lafayette W 113-74
(8) Kentucky L 66-73
 
Conference:
(8) Memphis L 67-73
(7) UConn W 76-64
(5) Cincinnati L 66-69
(5) Cincinnati W 58-57
(8) Memphis L 66-72
(7) UConn W 81-48
(7) UConn W 71-61
 
That's a 5-5 record against tournament teams, none higher than a 5-seed whom they split the season series with. They beat the same 7-seed three times and lost to a 6-seed and three games to 8-seeds.
 
Think twice before following the herd on Louisville 2014.

HipsterCat

March 17th, 2014 at 2:22 PM ^

but they also have many of the same players back from last season when they won the championship and are a favorite in many of the polls and #2 on kenpom, which are basically all the data points used for these calculations. those conference losses are all quite close and blowing out UConn 81-48 is pretty crazy, so it is what it is.

as they say though "ANYTHING IS POSSIBLEEEEEEEEE" especially in a single elimination tourney

Rather be on BA

March 17th, 2014 at 2:19 PM ^

Does this statement make any sense to people more mathematically inclined than me:

If an underdog has a greater % chance of winning than the % of people picking them to win the game, it is a good value to pick that upset.

For example, Silver gives Harvard a 42% chance to win the game, while according to CBS only 38% of people are choosing them to win.  Thus it would be a good pick from a value standpoint.

Am I just making things up, or does this make sense?

EDIT: This is obviously not an end-all evaluation, but another factor to take into consideration when choosing picks/upsets.

HipsterCat

March 17th, 2014 at 2:28 PM ^

it makes sense becasue the potential value outcome is greater in the event of you having the correct underdog pick and less people also having the correct pick. if everybody is picking against harvard and they lose then everybody gets those points but if they win its just the 38% who gets the points. the odds of them winning and you getting the points is just less likely, so it is a higher risk/higher reward kinda deal since 6/10 times you lose the bet and get no points in the bracket while others benefit and 4/10 you benefit and everybody else suffers.

Rather be on BA

March 17th, 2014 at 3:09 PM ^

Would it make the most sense to follow this thought process pretty strictly through the entire bracket, or deviate based on pure odds, especially in the later rounds?

For example.  46% of people on CBS have MSU getting to the final four, while 538 only gives them a 24% chance to be there.  Meanwhile, only 9% of people are picking Nova to make it to the final four, while 538 gives them a 21% chance of being there.  Thus, Nova would be the better pick from this train of thought

However, most would agree that MSU still has the better chance of making it out of the region.  This is where ti would get tricky, trying to balance value and straight odds.

EDIT: By making a bracket solely based off of value as it is defined here, I have answered my own question..  The bracket looks ridiculous.  

Basically, choosing the underdog almost always is more "valuable".  Seems like the better way to use this info is to use "value" as a way to choose upsets, and a way to choose final four teams that less people are choosing, but statistically have a relatively good chance to make a run.

Muttley

March 17th, 2014 at 4:20 PM ^

Say you're playing in a 100 entrant pool.  Before the brackets are filled out, each entrant has a 1% chance of winning.

Now, for simplicity, let's call this the Bandwagon Pool.  Under most scoring systems, the winner will have picked the National Champion.  So let's say nobody else in your pool takes Villanova.  If you take Villanova, according to Nate Silver, you have traded your 1% chance of winning for a 4% chance of winning.

Conversely, let's say 20 entrants each take Louisville, Florida, Arizona, and MSU to win the National Championship, respectively.  Those teams have the best chance, correct?  Well, those entrants have traded their 1% initial starting point for a 15%, 14%, 13%, and 6% likelihood of picking the National Champion.  But, they have to share that favorable outcome with 19 other entrants, and the pool winner will be decided by who did best through the semi-finals among those 20.  So in total, those entrants have traded their 1% initial starting chance for 15% / 20, 14% / 20, 13% / 20, and 6% /20, or a 0.75%, 0.70%, 0.65%. and 0.30% chance of winning the pool.

So I usually try to find that slightly off the bandwagon, undervalued pick.  Note, also, that where you're pool is run has a big bias.  When I lived in ACC country, half of the picks would usually be split between UNC and Dook.

Of course, going with what you think will be an undervalued pick can backfire.  In 2008, I thought going with 2 seed Georgetown would be an under-the-radar pick.  I was horrified when our ~70 entrant  pool came out and there were 8 other Georgetown picks.  So instead of increasing my chances, I misidentified what I thought would be a shunned pick, and was stuck with a way-overpopular, overvalued Georgetown.  Worse, Georgetown was upset by Davidson in the 2nd round.  That really makes for a miserable pool experience.  ("Paying" too much for a team you're not crazy about straight-up.)

On the other hand, if your pool has only 10 entrants, you are trading a starting chance of 10% for what you hand in.  Going w/ Villanova would REDUCE your chances from 10% to 4% (if the champion is picked by someone in the pool).

 

dahblue

March 17th, 2014 at 2:32 PM ^

I still think the regions are a mess.  Swap Nova and Duke and you've got a MUST more equitable arrangement.  Frankly, they don't even deserve a 3-seed but so be it.  1-3 against top 25 (with the only win against Iowa).  Mark Hollis, you won.

alum96

March 17th, 2014 at 3:07 PM ^

Agreed its all a bit odd but looking back what the committee did the RPI seems to be the most important measure after of course W-L.  Nova seems to be the #2 everyone is down on but Nova is 5 in RPI and Louisville is 19 in RPI.   Duke is 9.   SOS (80) doomed Louisville.

MSU and Louisville are the two 4s a lot of people are scratching heads at (Wiscy and MSU are essentially the same on paper to me yet one is a 2 and one is a 4)  Why? Wiscy has the UF win and a RPI 6... MSU has RPI 18. 

New Mexico is the one oddity in terms of high RPI (12!) but a 7 seed....

RPI will be something I'll be watching a lot closer next year since this is where the committee is worshiping.  Looks like a lot of the fan world has moved on to kenpom but the committee has not (UMass 52 in Kenpom, 23 in RPI - explaining a 6 seed)

http://sports.yahoo.com/ncaa/basketball/polls?poll=5

jermrs

March 17th, 2014 at 6:50 PM ^

Ken Pomeroy was just on the Dan Lebetard show and was asked who was the most over rated and the most under rated.

Ken was quick to call MSU as the most over hyped team.

He then noted Syracuse and NDSU as teams "that aren't getting the hype they should".