Hello There, Guest! (LoginRegister)

Post Reply 
How predictive are the December OOC stats & rankings?
Author Message
jedclampett Offline
All American
*

Posts: 3,542
Joined: Jul 2019
Reputation: 149
I Root For: Temple
Location:
Post: #1
MyBB How predictive are the December OOC stats & rankings?
There is perhaps universal agreement that, due to the extraordinary discrepancies between even the most well recognized ranking systems, it is extremely difficult to predict final season rankings for most D1 teams on the basis of early season rankings.

This observation also applies to the wide range of variations that are common throughout OOC schedules and at times into early conference play within many/most individual rating systems.

However, from a "30,000 feet perspective," broad examinations of MBB rankings data from prior seasons suggest that, while there is tremendous variation in the consistency of team rankings even within ranking systems through the course of a season, some team ranking systems, such as the Massey composite, through the first 8 games (1/4 of season) may actually have more predictive value than many may suppose.

If the effectiveness of a team's play and strength of schedule remains consistent enough through a full season, the "first 8 game" rankings can predict final season rankings fairly well, as is most frequently seen with teams that spend an entire season ranked among the top 15 or bottom 50 teams.

--------------------------------------------------------------------------------

On the one hand, rankings 1/4 of the way through the season tend to be very imprecise and inconsistent predictors of late season rankings - especially when the widely discrepant assessments by specific ranking systems - due to the many factors that contribute to improvements (team "gelling", lower SOS conference play) and declines in team play (e.g., player injuries, higher SOS conference play).

On the other hand, rankings and W-L records through the first 8 games do identify teams that may have high, intermediate, or low potential to be NCAA contenders.

--Consensus or average top 10-20 teams through the first 1/4 of a season are much more likely to make the NCAA than those ranked below #125th in the nation.

--Teams with average rankings in the top 50, after the first 8 games, tend to be more likely to make the NCAA field than those ranked below #150th in the nation.

--Very few teams ranked among the bottom 150, by most ranking systems, after 8 games end up being ranked in the top 75 or top 50 by the end of the regular season.

--Overall, perhaps 3 out of every 4 D1 teams/programs fall into 4 broad tiers in most ranking systems after the first 8 games of any given season:

1) The elite: Typically, these are teams from programs that have made multiple runs beyond the first weekend of the NCAA tournament within the past 5-10 years, have the deepest and most talented rotations, are considered perennial top 25-35 programs, and/or have the best recent recruiting classes, and the highest tier of D1 basketball Head Coaches.

2) The contenders: These are teams from perennial top 68 or up-and-coming programs that have a deep rotation of experienced and talented players, very strong if not elite recruiting classes, very highly regarded Head Coaches, and/or have made 2+ successful NCAA visits and/or NIT Championship round trips in the past few seasons.

3) The upwardly-striving: These are teams from perennial top 130-150 programs that have had some post-season tournament experience (NCAA, NIT, CBI, etc.) during the past 5-10 years, or up-and-coming programs that have a relatively deep rotation of experienced and moderately talented players, 2-3 solid recruiting classes, and/or highly-regarded Head Coaches.

4) The rest: These teams tend to be from perennial bottom 150-170 programs that either haven't made a trip to a NCAA, NIT, CBI, or other post-season tournament in the past 5-10 years, or have a relative shortage of skilled players with experience or talented younger players (due to injuries, ineligible players, and/or poor recruiting), are going through a major rebuilding year, and/or have a Head Coach who has had relatively little success in recent years, is approach the end of his contract, or is "on the hot seat."

Q: Do teams and programs tend to move between the above four categories frequently from season to season?

A: There are certainly semi-frequent mid-season shifts between the groups of elite and contenders teams, between the contenders and upwardly-striving teams, and between some of the better teams among "the rest" group and the struggling teams in the upwardly-striving group.

However, as a long review of the teams ranked among the lower half of 160 least highly ranked teams will tend to confirm, there are 7-8 scores of D1 teams that have little chance to make their way into the third "upwardly striving" tier, and 5 scores of teams that have a very low likelihood of dropping into the bottom 160.

--From this standpoint, rankings after the first 8 games may well tend to be fairly accurate predictors of where most teams will end up: Among the top 160 vs. the bottom 160 teams at the end of the season.

--If this is broadly correct, then the more accurate ranking systems in any given season may tend, at least, to identify teams with top 125 or top 150 potential, as opposed to the more likely middle or lower-tier teams.

============================================

It is important to note, in contrast to the general imprecision of 1st quarter basketball rankings, that the mathematical data that underlie and determine the rankings and bracketologies after the 1st quarter of a season are weighted equally and contribute just as strongly to the final regular season statistics as the data from the 2nd, 3rd, or 4th quarter of the season.

--The implications of this for the relative stability of team statistics will be the subject of the next post in this thread.

.============================================
(This post was last modified: 12-05-2019 07:01 PM by jedclampett.)
12-05-2019 03:42 AM
Find all posts by this user Quote this message in a reply
Advertisement


jedclampett Offline
All American
*

Posts: 3,542
Joined: Jul 2019
Reputation: 149
I Root For: Temple
Location:
Post: #2
RE: Are OOC statistics or rankings - - after 8 games - - meaningful or meaningless?
.
Part II:Game 1-8 statistics as predictors of full 32-game statistics:

.
--Importantly, the data (W-L records, offensive and defensive statistics and quality of opponents), themselves, which determine the computer rankings and bracketologies during the first 1/4 of the season matter just as much as they matter during the second 1/4, the third 1/4, and the final 1/4 of the season.

*As a result of the equal weighting of data from each quarter of the season, "first 8-game" statistics tend to be at least somewhat predictive - - ranging from broadly to moderately to strongly predictive - - of final regular season statistics, even if the "first 8-game" team statistic should turn out to increase or decrease steadily with respect to the national average by a net 40%, 60%, or as much as 100%.

--The ability of first quarter statistics to predict final regular season statistics with a relatively high, intermediate, or broad/modest degree of accuracy - - as opposed to rankings - - is of particular interest because statistical increases or decreases of 40% or higher are not uncommon, as when a team's strength of schedule (SOS) increases or decreases considerably between early season OOC and late season conference play.

--This is true, whether the statistic of interest has a national mean of 0 (e.g., average scoring margin or Win-Loss ratio), or non-zero (e.g. assists per game), with the proviso that, when the national mean is greater than zero, the team statistic must be computed in terms of deviation from the national mean.*(see part 4 addendum below)

Examples:

--If the average scoring margin after 1/4 season is 5.0 pts, followed by 4.0, 4.5, and 6.5 pt vacillations in scoring margins in the 2nd, 3rd, and 4th quarters of the season, the final regular season scoring margin will be 5.0 (5.0+4.0+4.5+6.5=20.0; 20/4=5.0).

----Thus, the "first 8 game" average scoring margin turns out to be a very strong predictor of its 32-game average scoring margin in those cases when the average scoring margins vacillate within a relatively narrow range across the 32 game schedule.

--If a team's average scoring margins for the 4 quarters are 5.0, 4.25, 3.75, and 3.0, the 32-game average scoring margin is 4.0 (5+4.25+3.75+3.0=16; 16/4=4.0).

--If the average scoring margins for the 4 quarters are 5.0, 5.5, 6.5, and 7.0, the team's 32-game scoring margin is 6.0 (5+5.5+6.5+7.0=24; 24/4=6.0).

----In these cases, the "first 8 game" average scoring margin remains quite strongly correlated with the full 32-game average scoring margin, even when a team's average scoring margin decreases (or increases) by +/- 40% between the first and last 8 games of the regular season.

----Thus, a team's +5 ppg scoring margin during its first 8 games turns out, when the full set of regular season data are available, to predict that - - even when the quarterly average scoring margins increase or decrease steadily by a net +/- 40% - - the team's final regular season scoring margin will fall within quite a narrow range, specifically between +4 and +6 ppg.


These three examples illustrate the point that, whether a team's scoring margin in the 2nd through 4th quarters of a season vacillates within a 2.5 ppg (points per game) range, declines steadily by 2.0 ppg[b], or increases steadily by 2 ppg, the team's final regular season scoring average would only vary from the 1st quarter scoring margin within a narrow (2 ppg) range (i.e., between 4.0 and 6.0 ppg).


.
The same types of computations show that a team's first quarter scoring margin remains a fairly good predictor of the full 32-game average scoring margin, even if the team's "first 8 game" to "final 8 game" average scoring margins steadily increase or decrease by 60%. (see part 3, below)
(This post was last modified: 12-05-2019 04:00 AM by jedclampett.)
12-05-2019 03:48 AM
Find all posts by this user Quote this message in a reply
jedclampett Offline
All American
*

Posts: 3,542
Joined: Jul 2019
Reputation: 149
I Root For: Temple
Location:
Post: #3
RE: Are OOC statistics or rankings - - after 8 games - - meaningful or meaningless?
.
Part 3: What about 60% and 100% scoring margin increase/decreases?

.
Specifically, despite a net +/- 60% 1st to 4th quarter increase or decrease (e.g., a 5.0 to 6.0 to 7.0 to 8.0 ppg increase or a 5.0 to 4.0 to 3.0 to 2.0 ppg decrease) in average scoring margins, the "first 8 game" scoring margin will nevertheless - as it turns out - predict that the team's final regular season average scoring margin will fall within a surprisingly narrow range, specifically between +3.5 ppg and +6.5 ppg.

.
Remarkably, even if there should be a net +/- 100% change in 1st to 4th quarter average scoring margins (a net 5.0 to 6.67 to 8.33 to 10.0 ppg increase or a net 5.0 to 3.33 to 1.67 to 0.0 ppg decrease in average scoring margins), the "first 8 game" scoring margin will predict that the team's final regular season scoring margin will nevertheless retain at least some broad predictive capacity.

--While a 5 ppg "first 8 game" average scoring margin, in the event of a steady 100% increase by the 4th quarter, is only modestly correlated with the final season average, [b]the correlation is, nevertheless, of sufficient magnitude to predict accurately that the final regular season scoring margin will fall within an unexpectedly narrow range, specifically between +2.5 ppg and +7.5 ppg.


----To make this crystal clear, a team with an average 5.0 ppg scoring margin at the end of the 1st 8 games in a 32 game season, can be predicted accurately to end the season with a final season average scoring margin somewhere between 2.5 and 7.5 points per game, even if their average scoring margin surges to 10.0 ppg or drops to 0.0 ppg during the last 8 games of the regular season.
(This post was last modified: 12-05-2019 04:05 AM by jedclampett.)
12-05-2019 03:51 AM
Find all posts by this user Quote this message in a reply
jedclampett Offline
All American
*

Posts: 3,542
Joined: Jul 2019
Reputation: 149
I Root For: Temple
Location:
Post: #4
RE: Are OOC statistics or rankings - - after 8 games - - meaningful or meaningless?
.
Part 4: Addendum

NOTE: The analyses reported above were all conducted using a statistic "average scoring margin," that has an approximate or theoretical mean of 0 among all D1 teams, nationally.

Other statistics with a mean of 0 would include, for example, average assist, rebounding, steal and blocks margins, as well as W-L record and W-L %.

In order to conduct similar analyses with those statistics that have a non-zero mean among all D1 teams (e.g., FG%, FT%, etc.), it is necessary to compute (i.e., transform) these variables from their raw data form into the form of "deviations from the national mean" for any given period of time within any given season.

Once the raw data are transformed into the form of deviations from the national mean, these computed deviations can be entered, simply, in the place of the "average scoring margin" statistics in the examples above. The remaining computations would follow the same procedures outlined above for average scoring margins.

----The interpretation of such analyses would have to be qualified to some extent, due to the fact they would be based on a statistical assumption that the national mean for any given variable would remain relatively constant across the measurement periods being examined and compared.

------However, one could augment a report by computing and comparing the national means for the two periods being compared. If the means turn out to be nearly constant, then the qualification would be of minimal importance for the purpose of these analyses.

============================================

NOTE: The analyses above require using published data from a complete regular season, for any given set of teams, such as all AAC teams, all P5 conferences, etc.

As such, these analyses are considered retrospective, "post-facto," or "post-hoc" analyses.

The terms "predict," "predictive," or "predicted" cannot be used to make definitive and direct inferences about the ability of specific variables to predict future events or performances.

These terms can be applied "after the fact" only because they are based on correlational analyses. In conducting such analyses, the strength of a correlation is used not to infer causality of any kind, since "correlation does not imply causation." Instead, the strength of a correlation ® can be transformed by simply squaring r to r-squared.

r-squared is a statistical estimate of the % of variance in an outcome variable (e.g., stability or change in the average scoring margin) that is attributable to the variance in the independent variable (the baseline statistic, in this case, the average scoring margin during the first 8 games).

Any indirect inference that the independent variable functions as a predictor in a data set collected in a previous season obviously can't predict data in any other data set, due to variations between data sets, also known as "cohort effects."

However, it is possible to test the hypothesis that the observed predictive association may generalize to a newer or ongoing data set, such as a specific season, recognizing the limitations of cohort effects, etc. when interpreting the findings.

In addition, the hypotheses would have to be precise enough to state clearly that they are based on specific assumptions of the magnitude of the hypothesized range of increases or decreases in the actual statistic being examined (e.g., scoring margins in this case).

Example:

"I hypothesize that statistic "X" will be a strong predictor of statistic "Y" if the observed (actual) change in statistic "Y" amounts to minor vacillations, as opposed to 30% or 50%, or any other % of actual change that you wish to compare the minor vacillations with.

One option would be to follow the examples presented above, of an actual change of zero, of 40%, of 60%, or of 100%.

.
All that would be required is the time and interest to do such analyses, the available of a computer with some basic calculating, spreadsheet, or statistical features, and access to published full-season data (see for example the websites ending with the terms "sports-reference").

Good luck if you wish to attempt this! It should be a fun little project.



If anyone would like to examine
(This post was last modified: 12-05-2019 04:43 AM by jedclampett.)
12-05-2019 04:22 AM
Find all posts by this user Quote this message in a reply
sierrajip Offline
Heisman
*

Posts: 5,697
Joined: May 2011
Reputation: 187
I Root For: UCF
Location:
Post: #5
RE: Are OOC statistics or rankings - - after 8 games - - meaningful or meaningless?
Trying to sell your product?
12-05-2019 06:11 AM
Find all posts by this user Quote this message in a reply
Advertisement


Def Berkkat Offline
1st String
*

Posts: 2,185
Joined: Dec 2017
Reputation: 219
I Root For: UC
Location:
Post: #6
RE: Are OOC statistics or rankings - - after 8 games - - meaningful or meaningless?
If this thread was a class...

I'd have given it 10 minutes then dropped it.
12-05-2019 07:20 AM
Find all posts by this user Quote this message in a reply
Tiger1983 Offline
BBA
*

Posts: 35,184
Joined: Apr 2006
Reputation: 2033
I Root For: Tigers - GTG!
Location: The enemy’s lair

DonatorsDonatorsDonators
Post: #7
RE: Are OOC statistics or rankings - - after 8 games - - meaningful or meaningless?
I did not venture into the briar patch, but the general consensus I've read over the years is metrics like KenPom are meaningful starting around January.

It is also dangerous to draw conclusions about a team's strengths and weaknesses based on stats in early December. For example, it is reasonable to assume Memphis will climb the learning curve quicker than a veteran squad and improve upon addition of Wisemen and LQ.
12-05-2019 10:54 AM
Find all posts by this user Quote this message in a reply
Stickboy46 Offline
Heisman
*

Posts: 8,915
Joined: Dec 2016
Reputation: 406
I Root For: Wichita State
Location:
Post: #8
RE: Are OOC statistics or rankings - - after 8 games - - meaningful or meaningless?
When you underline so many random things it loses its impact.
12-05-2019 11:42 AM
Find all posts by this user Quote this message in a reply
jedclampett Offline
All American
*

Posts: 3,542
Joined: Jul 2019
Reputation: 149
I Root For: Temple
Location:
Post: #9
RE: Are OOC statistics or rankings - - after 8 games - - meaningful or meaningless?
.
The main point of it is that what we learn about our teams' early performance in their expanded box scores, to date, is more informative after 8 games than we may imagine, and give us a better idea about the team's prospects than the various rankings do after only 8 games.

--Also more informative than the rankings at this point are the W-L records, scoring margins, and indicators of the SOS of our team's upcoming opponents.

8 games is a small sample, and everything can change in conference play, but the data we have at this point tells us a lot more than we may realize, for anyone who is able to review pages like this one in enough detail:

https://www.sports-reference.com/cbb/sch.../2020.html

.
The other main point is that, in almost every aspect of life, early data are often more representative than many of us realize.

That's why first impressions are often correct.
(This post was last modified: 12-05-2019 05:15 PM by jedclampett.)
12-05-2019 05:01 PM
Find all posts by this user Quote this message in a reply
Post Reply 




User(s) browsing this thread: 1 Guest(s)


Copyright © 2002-2024 Collegiate Sports Nation Bulletin Board System (CSNbbs), All Rights Reserved.
CSNbbs is an independent fan site and is in no way affiliated to the NCAA or any of the schools and conferences it represents.
This site monetizes links. FTC Disclosure.
We allow third-party companies to serve ads and/or collect certain anonymous information when you visit our web site. These companies may use non-personally identifiable information (e.g., click stream information, browser type, time and date, subject of advertisements clicked or scrolled over) during your visits to this and other Web sites in order to provide advertisements about goods and services likely to be of greater interest to you. These companies typically use a cookie or third party web beacon to collect this information. To learn more about this behavioral advertising practice or to opt-out of this type of advertising, you can visit http://www.networkadvertising.org.
Powered By MyBB, © 2002-2024 MyBB Group.