Monte Carlo simulation using ratings for NCAAB

[uva3021]

I'm trying to figure out the best way to do this. Usually the ideal simulation involves some expected or accepted winning percentage based on an adjusted true measure of score, like a Pythagorean score, and then converted to a Winning %. I ran several for the conference tournaments using average spreads and conference efficiency with numbers from basketball state, and the results were pretty satisfying. I had Minnesota, Houston, and Rhode Island as the only teams with real value compared to sportsbooks.

Now I want to do one using a rating system, preferably Basketball State's Location based performance ratings, which have proven to be the best predictor of NCAA tournament basketball. So I need to find a way to convert a raw rating system based on around an aggregated 120 pt scale, and appropriate each team's rating to a percent chance of winning against one another. But I'm struggling with the most efficient way to achieve this.

Maybe instead of Pythagorean % I can incorporate a Normal Distribution, using the Standard Deviation and Mean of the rating system.

Any ideas?

[goterps74]

Are u posting this from an internet cafe? :)

[Romanowski]

uva can you expand on the method you used for conference tournaments

perhaps show an example

[The Judge]

Uva, I am not familiar with the Pythagorean score or percentage that you speak of. Of course, the Pythagorean theorem states that in any right triangle, the area of the square whose side is the hypotenuse (the side opposite the right angle) is equal to the sum of the areas of the squares whose sides are the two legs (the two sides that meet at a right angle) but I am not sure how this relates to sports wagering.



............

I'm with Rome, please clarify.

[uva3021]

A Pythagorean Winning % is often synonymous with Baseball, if you survey baseball-reference you will see a team's Pyth W% in a column on the team page. How is tihs calculated?

Pyth W% = (Runs scored ^2) / (Runs scored ^2 + Runs allowed ^2)

For basketball the exponent is 14. Basically its formulated by taking 50-100 years of data, and using the exponent that best fits the league average over that course in time, so its rather arbitrary, but there is reason to the formulation. Bill James explained all of this in his Baseball Abstract

So for basketball, I use the efficiency differential to come up with an expected score (using basketball state or kenpom, and adjusted every thing to average tempo [66.7]), and subsequently expected winning % versus an accepted winning %. This is all based on the foundation of probability factor, where teams at length will regress towards the mean.

To calculated a team's chances of winning versus another team, you can do a linear calculation

Chances of Winning = Team A% - Team B% + .500

Or a more precise measure, a logarithmic relationship between the two teams:

W% = ( Team A - Team A * Team B ) / (Team A + Team B - 2 * Team A * Team B )

I have extracted the data, now I need a way to convert a rating scale to a percent chance of winning.

[uva3021]

I finished constructing the data set for all teams through the sweet 16, the "IF" statements are becoming more and more complicated and its really hindering my efforts, any if/then experts out there?

Almost finished just need the "IF" statements to run it through a simulator

[Romanowski]

judge here is a full explanation on pythagorean for expectations, baseball and basketball

http://en.wikipedia.org/wiki/Pythagorean_expectation

[The Judge]

Cool, thanks Rome and congrats on the mod position. Very well deserved, in my opinion.

[GFootsXtreme]

Originally Posted by uva3021

I finished constructing the data set for all teams through the sweet 16, the "IF" statements are becoming more and more complicated and its really hindering my efforts, any if/then experts out there?

Almost finished just need the "IF" statements to run it through a simulator

i can help you construct your IF THEN if you tell me in plain terms i can help you convert it to the forumla

[GFootsXtreme]

usually i find that the hangups occur where you need to use AND or OR

i have an automated spreadsheet i've been developing for college football, it's automated to the point where it tells me exactly what to pick after the data is entered

all based upon IF THEN OR AND COUNT formulas

it does get tricky but it's way cool when all you need to do is feed it the data and it spits out the answers!

[uva3021]

I stopped doing this, the ratings weren't practical in terms of predicting head to head, just using the average spreads and efficiency now, and running a chalk stipulation in order to look for value

a simulation is not really a predictor, its an arbitrarily statistical appropriation of randomness, effort not worth the reward IMO