My Math Forum  

Go Back   My Math Forum > College Math Forum > Advanced Statistics

Advanced Statistics Advanced Probability and Statistics Math Forum


Reply
 
LinkBack Thread Tools Display Modes
May 18th, 2009, 10:19 AM   #1
Newbie
 
Joined: May 2009

Posts: 1
Thanks: 0

How to regress actual data towards projected data.

Hey all,

So I am currently working on a system that will predict the outcome of MLB games.

I need a method to regress a player's actual stats with his predicted stats. By the end of the season, I want player A's predicted stats for a single game to be 50% projections 50% actual.

For example, let's say player A is projected to play 50 games and hit at an average of .300. In actuality, player A has played 25 games and is hitting .200. So what can we predict player A's average to be for his 26th game?

Originally I went with a formula like this
w = weight = Games played actual/Games played predicted = .5

Average(26th game) = (Projected Average)*(1-(w/2)) + (Actual Average)*(w/2)
Average(26th game) = (.300)*(.75) + (.200)*(.25) = .275

This looks all well and good, but things start getting messy if we let the actual games played get larger than 2x the predicted. For example...
Average(201st game) = (.300)*(1-(4/2))+(.200*(4/2))
Average(201st game) = -.300 + .400 = .100

Now obviously this result is not correct. In this scenario, a players predicted average for a single game should NEVER drop below his actual average for the entire season. Instead, as the actual games played approaches infinity, the predicted average for a single game should approach the actual average. I'm not sure how to calculate this.
BigLRIP is offline  
 
May 18th, 2009, 11:01 AM   #2
Global Moderator
 
CRGreathouse's Avatar
 
Joined: Nov 2006
From: UTC -5

Posts: 16,046
Thanks: 937

Math Focus: Number theory, computational mathematics, combinatorics, FOM, symbolic logic, TCS, algorithms
Re: How to regress actual data towards projected data.

You're weighting the actual games as
((actual games)/(predicted games))/2
but it seems that you should weight them as
(actual games)/(predicted games + actual games).

This would probably solve your issue.
CRGreathouse is offline  
Reply

  My Math Forum > College Math Forum > Advanced Statistics

Tags
actual, data, projected, regress



Thread Tools
Display Modes


Similar Threads
Thread Thread Starter Forum Replies Last Post
Rolling 3 months YOY data to Monthly YOY data lumpa Real Analysis 0 October 19th, 2012 09:07 AM
Data Management Help Please gaz Algebra 4 November 25th, 2010 12:11 AM
Analyzing a Set of Data minicoop503 Algebra 1 September 8th, 2009 04:06 PM
FFT on sample data ramyy Applied Math 1 June 29th, 2009 05:49 PM
A Data Problem EddSjd Algebra 2 September 8th, 2007 11:06 AM





Copyright © 2017 My Math Forum. All rights reserved.