Preseason Publications A10 Predictions

[Iggy1979]

BlueRibbon picked Dayton first in conference and 25th in country not knowing Dyshawn Pierre would miss at least first semester. Wonder what they would've done if they'd known.

[Seawrightspostgame]

Advanced Earth? or Flat Metrics? This is beyond me.

[ace]

Or, we could all just take a deep breath and realize that Ken himself says about a month's worth of data is needed until his information is meaningful. He chooses to release his starting point, even knowing that it may be misunderstood, probably just because everyone else does. Plus, it helps to track how accurate these starting points are with how things develop, which strengthens his process.

Like I said, I like Ken Pom's stuff. But if he himself acknowledges you need a month's worth of data to draw even remotely meaningful conclusions about who is worth a shit and who isn't, then why the hell is he releasing preseason rankings? It is absurd on its face, considering his rankings don't attempt to do any subjective qualitative analysis and are dependent upon metrics that do not exist to any degree for the upcoming season.

His qualification is that a few weeks of play helps to support his model, not that the initial prediction is meaningless. By definition, no preseason poll exists based on play from the upcoming season, and I'm not sure the other polls are better informed. Is a poll reflective of human subjectivity better? There's merit to both.

His metrics, as a predictive source, do exist for the upcoming season. It starts with the best predictor of a team's offense, which is last season's offensive efficiency followed by the previous season's offensive efficiency followed by last season's defensive efficiency (flip for defense). The model takes those past stats and adjusts for returning players. It also determines which player's returning minutes were most valuable compared to his teammates and adjusts for a player's class. One thing not considered is transfers or recruits outside the top 100. He's looked back at transfers after the fact and so far has determined that it's a little bit all over the place. Take a guy like Iverson- I'm not sure anybody has a foolproof way to predict what he'll give them this year. For McGlynn, we know he's a shooter, but he's moving up a conference, which may have a defensive infuence, both his and the defense played on him.

It's not an either/or situation, and Pomeroy is clear on that, as are most stats people.

In general, I understand if people find stats talk tedious, but the people doing this work love the games and sports they cover. Who else but a rabid fan would put this much time and effort into figuring it out?

[TruePoint]

I'm not sure if we are talking past each other or what, but I totally understand what Pomeroy is doing and, as I've said a number of times now, I have a tremendous appreciation for stats-based analysis globally and Pomeroy's ranking in particular. However, as a predictive model, I don't think there is much value in this type of analysis. In general, I think this kind of analytics is interesting as a means of analyzing what has already happened but less useful as a means of predicting what is going to happen in the future, because it does not account enough for non-quantifiable variables that are virtually guaranteed to impact future events. I believe this even in the context of a season in progress, but especially for a season that hasn't even started yet. In layman's terms: I believe last year means shit.

No matter how good your model is, I do not think it can be meaningfully accurate in this kind of prediction. We could put it to the test: what percentage of teams do you think will finish within five spots in either direction of Pomeroy's prediction? I'd say between 5% and 10%, which may actually be impressive given the degree of difficulty, but also makes the prediction essentially meaningless.

I'm not saying Pomeroy isn't a good fan or whatever. I think it is obvious he is a great college basketball person and contributes greatly to others' enjoyment. But I think his preseason rankings are worthless. (In fairness, as I said before, I think all preseason rankings are worthless, but before the ball is tipped I'd actually put more stock in subjective predictions; during the season that shifts dramatically to the analytics-based rankings as more and more data is available.)

[ace]

In general, I think this kind of analytics is interesting as a means of analyzing what has already happened but less useful as a means of predicting what is going to happen in the future, because it does not account enough for non-quantifiable variables that are virtually guaranteed to impact future events. I believe this even in the context of a season in progress, but especially for a season that hasn't even started yet. In layman's terms: I believe last year means shit.

This seems to be what we disagree on. Stats analyzes what has already happened to predict what is likely to happen based on all available and relevant data points. I trust those numbers, with the given suggested probabilities, because it's already happened. What's predictive is applying it to a new guy, who is similar statistically. If the numbers, in countless examples that have already happened, suggest that a player who performed X in one season will perform Y in the next, I think that has merit. Of course, it's not foolproof. There's chance and luck and all that good human stuff, but variability is accounted for in statistical models. Previous behavior and performance are meaningful to future behavior- in life, in basketball, by the numbers, by human experience, and I say that as both a psychologist and a stats geek.

[SGreenwell]

I just did a quick search, and the first link that popped up was an NC State fan site that had the preseason rankings from 2014-15. Predicted ACC finish, with the Pomeroy ranking:

1. Duke (1)
2. Louisville (3)
3. Virginia (8)
4. Syracuse (11)
5. North Carolina (20)
6. Pittsburgh (23)
7. Florida State (41)
8. Miami (57)
9. NC State (60)
10. Notre Dame (69)
11. Clemson (75)
12. Georgia Tech (96)
13. Wake Forest (104)
14. Boston College (134)
15. Virginia Tech (168)

The actual finish in the final 2015 Pomeroy rankings:

Duke (4, -3)
Virginia (5, +3)
Notre Dame (9, +60)
North Carolina (11, +9)
Louisville (16, -13)
North Carolina State (31, -11)
Miami (46, +11)
Syracuse (59, -48)
Pittsburgh (78, -55)
Clemson (80, -5)
Georgia Tech (83, -13)
Florida State (97, -56)
Wake Forest (124, -20)
Boston College (106, +28)
Virginia Tech (184, -16)

Obviously, it’s just one year of data and one conference; I didn’t really have time to look at more, since I’m between meetings. The biggest difference was 60 spots, while the smallest was 3 on two teams. He was off by an average of 23.4 spots, between preseason and year-end rankings.

If you want to project that to URI, based on their 2015-16 preseason ranking and using the average, that would give them an outcome range from #27 to #73/#74. That seems about right to me. As is, they’re probably projected to be tournament team or on the bubble. If they exceed expectations or get a bit lucky, they’re solidly in the field and in the Top 25. If they unexpectedly struggle or have some injuries, then they’re probably on the outer bubble or repeat last year’s success. They’re probably not cratering out though.

[bressler3south]

Just win.

[TruePoint]

In general, I think this kind of analytics is interesting as a means of analyzing what has already happened but less useful as a means of predicting what is going to happen in the future, because it does not account enough for non-quantifiable variables that are virtually guaranteed to impact future events. I believe this even in the context of a season in progress, but especially for a season that hasn't even started yet. In layman's terms: I believe last year means shit.

This seems to be what we disagree on. Stats analyzes what has already happened to predict what is likely to happen based on all available and relevant data points. I trust those numbers, with the given suggested probabilities, because it's already happened. What's predictive is applying it to a new guy, who is similar statistically. If the numbers, in countless examples that have already happened, suggest that a player who performed X in one season will perform Y in the next, I think that has merit. Of course, it's not foolproof. There's chance and luck and all that good human stuff, but variability is accounted for in statistical models. Previous behavior and performance are meaningful to future behavior- in life, in basketball, by the numbers, by human experience, and I say that as both a psychologist and a stats geek.

Ironically, I am full-fledged stat geek when it comes to baseball. In basketball (and even more so in football), I think there are too many factors to isolate and properly assign value to - injuries, roles, impact of having different players with different skill sets on the floor with you, the ability to add or dramatically improve an entire skill set from year to year, even things as nebulous as "chemistry" probably do not matter at all in baseball but can impact basketball and football teams dramatically. I also think, and this is maybe a little bit of a different argument, that in basketball and football the result of any given game can be so matchup-dependent that the transitive third property barely applies at all - and how that is somewhat relevant to this discussion is I'm not sure how that can be accounted for in a purely mathematical predictive model.

One last point: I think overall statistical projections are much better when looked at in the macro than they are likely to be for any individual team, which is another reason why I am interested in this kind of analysis but do not put as much stock into it as many others seem to.

[Rhody74]

At this point I don't care about any predictions. I just want the season to begin already!

[ace]

At this point I don't care about any predictions. I just want the season to begin already!

Soon...

There's Madness, then a secret scrimmage, then an exhibition, and then a real, actual game.

[Rhody74]

I'm not known for my patience, Ace. Ask my wife ....

[Seawrightspostgame]

I feel like it's low hanging fruit to poke holes in the "advanced metric" world. It's obviously not perfect and nothing is. I get it. I understand people like it. I just don't like being force fed the advanced metric narrative.

The # that goes on the paper rarely tells the actual story. Everyone that cheers against the pats this year sees their team as a few plays away from winning. I see a patriots team that's a few plays away from blowing everyone out.

[rjsuperfly66]

I think advanced metrics are more prevalent in college sports, because it's more about efficiency, production, and results. Who did you play, where did you play them, how did you do?

In professional sports, it is all about the wins and losses. You need wins in college basketball as well, but a team that goes 17-14 can still make the tournament with a strong SOS, compared to a team that goes 25-8 with a weak SOS.

I agree with those who think KenPom is mostly ineffective at this time of the year, but I think it needs to be used as an average baseline as opposed to gospel. If Team X performs on average, they can go _-_ and be favored in these games.

As the sample size grows, expectations change, and the numbers change as well. A team that plays well will have higher expecations. A team that plays poorly will have lower expectations, and a team that plays average will stay right around neutral.

All metrics, even RPI, have to start from somewhere. Otherwise, all teams would be the same. Beating Texas would be the same as beating Columbia. Beating Kentucky would be the same as beating Eastern Michigan, and on and on.

[SmartyBarrett]

This seems relevant: http://kenpom.com/blog/index.php/weblog ... on_ratings

[Rhody15]

All in all, I would have liked to see us ranked a bit higher than 50.

[luke]

It won't matter where they are ranked in preseason this year. They play enough good teams
to determine their own destiny. If they beat good teams like Maryland and Valpo they will be
highly ranked, and if they don't they won't be.

[BFC]

Atlantic 10

Team to beat

Dayton Flyers

This is going to be a battle between Dayton, Rhode Island and Davidson. I expect great competition in this league.

Player of the year

DeAndre Bembry, Saint Joseph's Hawks

Bembry gets the nod over E.C. Matthews of Rhode Island and Jack Gibbs of Davidson.

http://espn.go.com/mens-college-basketb ... ayers-year

[Rhody15]

EC ranked as the 9th best off guard in the country.

http://collegebasketball.nbcsports.com/ ... asketball/

[Rhody15]

Dicky V predicts us, Dayton and Davidson to make the big dance.

http://espn.go.com/espn/dickvitale/stor ... t-field-68

[ramster]

Dicky picks Valpo and Iona to win their Leagues - two of our scheduled opponents.
Only 4 teams from the BE?

[ace]

Not too much new, but there's a clip of one of my favorite Hassan dunks. I always loved the bench's reaction.


http://www.ncaa.com/news/basketball-men ... sf14956818

[Ramulous]

Let's hope the A-10 ushers in the "Dan Hurley Era"......

[ramster]

Maryland Number 1

http://espn.go.com/mens-college-basketb ... r-rankings

[Rhody74]

ESPN's Insider had this to say about Rhody (I don't recall seeing this anywhere on this board):

Sleeper

Rhode Island Rams

I hesitate to hang a sleeper label on the A-10's No. 1 defense (much less on a team that went 13-5 in conference play last season), but after being passed over for an NCAA tournament bid, Rhode Island still has an under-the-radar feel. If Dan Hurley's team can cut down on turnovers, the Rams can push Davidson for the A-10 title. While we're on the subject of sleepers, no one's paying much attention to George Washington or St. Bonaventure, even though the Colonials and the Bonnies each return three starters from teams that scored about as many points as they allowed in conference play last season. La Salle also broke even in that department, more or less, and the Explorers are bringing back easily the league's most prolific per-possession shot-launcher in Jordan Price.

They pick Davidson to win the A-10 and Dayton to disappoint.

The most surprising prediction is they left EC off the All-Conference team, but make Hassan one of the three forwards on it. The only guard is Jack Gibbs of Davidson.

If you have an Insider acct. the URL is http://espn.go.com/mens-college-basketb ... tlantic-10

[Rhody15]

Highest seed I think I've seen a site have us. I'll take it!

http://collegebasketball.nbcsports.com/ ... -november/

[Shaolin Swat]

I would sign up for that seed and placement in a heartbeat

[bigappleram]

Why, the dreaded 8/9 seed....i would rather be 11-12 in that case.

[Shaolin Swat]

I like the way that bracket unfolded, I'm not as high on UNC this season as most are. Would love to see us beat Shaka in the first round and then pull an upset on UNC in the round of 32.