Runboard.com
Слава Україні!

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


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

 
NINJA Terp Profile
Live feed
Blog
Friends
Miscellaneous info

Administrator

Registered: 07-2007
Posts: 1183
Karma: 55 (+85/-30)
Reply | Quote
Statistics Primer


A quick primer for the statistics and numbers that I post here.

NINJAMetrics
This is the collective name given to the statistics of Offensive Relativity (OR) and Defensive Relativity (DR). These stats were developed in the 2007 CAN Softball season as a measure of how teams perform relative to their opponents. While this is meant to simply encapsulate past performances, it ended up being able to predict the result of future games with 70% accuracy during the regular season and was able to get within one standard deviation of the final score 10% of the time. On a completely unrelated note, I'll be going to Vegas at some point.

The thought process is as follows. Consider that Maryland, for example, defeats Virginia 20-10. Is Maryland's offense that good or is Virginia's defense that bad? Conversely, is Virginia's offense that bad or is Maryland's defense that good? The numbers look a bit different if Maryland normally scores 40 runs a game - Virginia actually didn't play that badly as they held the Terps to 20 less runs than they are used to scoring. To cite another thought, consider if Virginia normally holds teams to 5 runs a game... well, giving up 15 more than they normally do says something about Maryland's ability. It is through these lenses that OR and DR take shape.

Predicting games takes into account the effect of both team's OR, DR, runs (points) scored per game, and runs (points) allowed per game. There is no bias towards particular schools and prior seasons are thrown out.

Other numbers that are in the NINJAMetrics chart are SOS, RPI, RF(PF), and RA(PA). SOS is strength of schedule and represents 2/3 of the winning percentage of a team's opponents in games not including that team and 1/3 of the opponents' opponents' winning percentage. RPI is the commonly known Ratings Percentage Index and is comprised of 1/4 team's winning percentage plus 1/2 of a team's opponents' winning percentages plus 1/4 of a team's opponents' opponents' winning percentage. A variant is used by the NCAA to help develop their Division I basketball tourneys. The aggregate percentages used are averages. RF(PF) refers to the runs for (or points for) that a team scores. Likewise, RA(PA) refers to the runs against (or points against) a team. The rankings of each team in each metric are listed next to the metric itself.

Also, the 'metrics will not acknowledge games that are won or lost by forfeit. While I can understand the argument that it should count as a loss for RPI/SOS purposes, I don't feel that it tells anything about the team itself and its on the field performance.

COLLEY MATRIX
The Colley Matrix is one of the computer rankings that is used by the NCAA to determine its football national championship matchup at the end of the season. This is being leveraged for personal amusement only and more details can be found at Mr. Colley's website - www.colleyrankings.com. It should be noted that there is a significant amount of noise in the rankings at the beginning of the season. The noise usually levels out a few games into the season - in week 2 of the 2007 college football season, South Florida was ranked #8 and Maryland was ranked one notch below Southern California. However, by Week 4, things started to look more 'normal'.

Also, like the 'metrics above, the Colley rankings will not include games that are won or lost by forfeit for the same reasons.

Last edited by NINJA Terp, 3/Oct/07, 9:16 am
25/Sep/07, 9:55 am Link to this post Send Email to NINJA Terp   Send PM to NINJA Terp
 
neutron man Profile
Live feed
Blog
Friends
Miscellaneous info

Registered user

Registered: 07-2007
Location: Fairfax, VA
Posts: 131
Karma: -6 (+9/-15)
Reply | Quote
Re: Statistics Primer


Will there be predictions as for softball?
25/Sep/07, 4:27 pm Link to this post Send Email to neutron man   Send PM to neutron man
 
NINJA Terp Profile
Live feed
Blog
Friends
Miscellaneous info

Administrator

Registered: 07-2007
Posts: 1183
Karma: 55 (+85/-30)
Reply | Quote
Re: Statistics Primer


To some extent, yeah - I may either put the predictions straight up in this thread or place spreads in the Coaches' Poll post. I haven't decided, but am leaning towards the former.
26/Sep/07, 8:18 am Link to this post Send Email to NINJA Terp   Send PM to NINJA Terp
 


Add a reply





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