About this database
The main database for each season is the one with the star next to it.
Here you will find all of the items listed below plus more. You can
select which fields you want to see, the order to present them for,
as well as limit it to your favorite conference. Please note that
this is all unofficial and that I'm sure there are mistakes in the
data. If you just want to see the RPI, you can grab the brief version,
also a conference RPI rating s available. Note that this
is calculated by averaging the individual team ratings.
There is also a Z-Rating brief available if you want to check my power ratings.
Description of Calculated Data
All records are for games against Division I opponents only
- RPI
- Ratings Percentage Index. This is the one everyone talks
about with regard to making the tournament. It is calculated as
.25*WIN%+.5*(Opponents WIN%)+.25*(Opponents Oppenents Win%)
- RRPI
- Road RPI. Same as RPI, but calculated only using the teams
road record, your opponents home record, and the opponents opponents
road record. Note that the data used has no provision for neutral site games
which will show up randomly as home/away games which taints this calculation.
- HRPI
- Home RPI. See previous, but in reverse
- WRPI
- Weighted Rating Percentage Index. This uses the same formula
as the RPI, except that you are awarded your opponents winning percentage
for a win, and 1-oppwin% for a loss. This puts more weight into which
teams you beat and which ones you lose to, so losing to Tennessee isn't
nearly as damaging as a loss to a 3-8 team. On the other hand, beating
Tennessee is worth more that beating the 3-8 team.
- DWRPI
- Dominance Weighted RPI. This is similar to WRPI, except the
win/loss value is scaled statistically based on the margin of victory/loss
of the game compared to the other teams which have played your opponent.
The idea is, ok, you played that 3-23 team, but you beat them by 60 points,
while their average margin of defeat was 10 points, so you get more points
for beating them than everyone else. It's not scaled by how much, but
compared to everyone else.
- RPI2
- Second Order RPI. This calculation essentially reruns the
RPI again, except that the teams RPI is substituted for their winning
percentage.
- Z
- Z-Rating. This calculation does a statistical analysis using
standard deviation of points scored, points allowed, and margin of victory
along with a weighted winning percentage (see WRPI). Z scores are also
presented for each teams defensive(DEFZ), offensive (OFFZ), and margin
of victory (MZ). The unscaled (rawz) score is the sum of DEFZ,OFFZ, and MZ
and has no concept of winning percentage for the team in question or
the opponent (i.e. no strength of schedule). It is purely a measure of a
team's dominance over their opponents. The unscaled Z settles out much more
slowly than the other measurements. Sorting by MZ can also provide interesting
results.
- STD and P-Ratings
- These are similar in concept to the Z. STD generates
a standardized score for each game a tam plays in for their offensive and defensive efforts (points scored and allowed) based on the opponents other games weighted by the opponents winning percentage. The standard scores are averaged per game for an offensive and defensive rating which are summed for the final power rating. The P system is similar, except the offensive and defensive rankings are assigned based upon (points scored)/(opponents average points allowed) and
(points allowed)/( opponents average points scored). The offensive and defensive ranking are averaged per game and summed (off+.75*def) for the final rating. The defensive P rating is scaled down as it tends to inflate rankings. This
was originally an arbitrary factor chosen just by eyeballing, but the offensive
rankings break pass over 1.000 near team# 150, while the defensive ratings
pass 1.000 near team #200, 150/200=.750 :-).
- T-Factor
- Tournament-Factor. This is a value I came up from analyzing the RPI data versus the teams receiving bids. It is defined as .25*win%/RPI. Studying the RPI data
over the last several years, it appears that having a "t-factor" less than
1 is a bad omen for getting an at-large bid. Only one team has gotten
a bid at all (automatic or at-large) in the past four years. Florida,
with a t-factor of .980, made it in for the 98-99 tournament. A casual
study indicates that it should be around 1.333. Below is a graph of the last
4 years (97,98,99,2000) showing the tfactor (x-axis) vs obtaining a tournament
berth. It's a bit difficult to see the actual data, but the cutoff at
0 is clear (note: the graph is actually 1-tfactor).
In addition, data is presented for each teams current win/loss streak,
record in the last 10 games, record versus the RPI Top 10,25,26-50,51-100,
101-150, and teams below RPI index 150. Also, the percentage of non-conference
opponents in the RPI Top 50 (PT50) and RPI bottom 50 (PB150) is listed. These
along with overall wins and losses to the top 50 and bottom 150 are believed
to be the basis of the "rewards" and "penalties" the NCAA applies to each
schools base RPI. Teams playing 50% or more of their non-conference schedule
versus top 50 are rewarded, while schools playing 50% or more against the
bottom 150 are penalized. (See Jerry Palm's RPI FAQ)
Some of the calculations are also done for non-conference (NC) and conference
(C) opponents. The non-conference calculations eliminates bias introduced
by teams which play in weak conferences, and also highlights teams which
play in strong conferences but play weak non-conference schedules. Some
people believe that the selection committee puts more emphasis on your
non-conference opponents than on you conference opponents (i.e. who you
choose to dance with rather than who you have to dance with).