KenPom Dropped

#27
#27
Unconditional Surrender said:
That 88.4 for AdjD looks optimistic
Click to expand...
At the same time the offensive outlook of 110.0 appears like it could be much better, 137.3 against Lenoir-Rhyne. Would’ve been 2nd highest output last year, highest of 2020 and highest of 2019. Obviously 137.3 isn’t realistic, but could certainly see above 110.
 
  • Like
Reactions: MinisterofDef#92, berryvol, Unconditional Surrender and 1 other person
#30
#30
Chris4Vols22 said:
21-9 sounds realistic to me. Wow, all of the losses are 1-posession.
Click to expand...
Yea, games 47%-53%…
N Villanova
A Colorado
N Texas Tech
N Memphis
A LSU

He has us going 3-2 in those…would seem preseason like those are kind of the swing games…all the others he gives at least a 61% chance of winning, and then at least a 59% chance of losing.
 
#34
#34
dtmtn said:
I only see 6 total losses in his projections. Where does the 9 losses come from?
Click to expand...
That’s the analytics side of it…basically saying a cumulative group of games which he has us winning but by small margin we are unlikely to to go undefeated in.

Example being, he has wins against Colorado, Memphis, LSU, Mississippi State, Kentucky and Arkansas…but none as more than 60% likely. So the percent of going 6-0 in that stretch is very small, make sense?
 
  • Like
Reactions: TrumpedUpVol
#35
#35
bleedingTNorange said:
That’s the analytics side of it…basically saying a cumulative group of games which he has us winning but by small margin we are unlikely to to go undefeated in.

Example being, he has wins against Colorado, Memphis, LSU, Mississippi State, Kentucky and Arkansas…but none as more than 60% likely. So the percent of going 6-0 in that stretch is very small, make sense?
Click to expand...
It's just basic probabilities
 
#37
#37
Every year I aggregate human and computer rankings/ratings to create a power rating, and use that to simulate the season. The 5 computer rankings I use are Ken Pom, Team Rankings, ESPN BPI, Bartorvik, and Haslametrics. There are a bunch of different computer ratings published in the preseason but these 5 seem to have the best track record and they happen to be as precise as possible in incorporating player-based adjustments. Even as KenPom has pointed out, any one system won't be consistently more accurate than an aggregation of systems because each has different features and bugs, and aggregating helps to smooth through that. The two human rankings I use are the only two which rank all 358 teams (CBS & SI). Because the rankings all operate on different scales, I normalize them to have the same average and standard deviation. The computer rankings are given a 75% weight and the human rankings 25%.
ranking.jpg
That power rating is then applied to all 5,561 scheduled NCAA games (I also simulate every early season tournament), and calculate records.
schedule.jpg
Win totals.jpg
 
  • Like
Reactions: The Dog, Volprofch05 and Vol49er
#41
#41
bleedingTNorange said:
Why’s Tennessee 21-8 but plays 31 games?
Click to expand...

Probably since the 2nd round tournament opponent hadn’t been determined yet.

Edit: But the SEC record doesn’t align. 15-3? 12-6? I guess the difference is considering the W/L probabilities of each game in aggregate. At LSU is only a slight favorite to win, so it’s really worth about a half loss and a half win although it counts as a full win in the projected game by game results. Math is hard.
 
Last edited:
  • Like
Reactions: golfballs
#42
#42
bleedingTNorange said:
Why’s Tennessee 21-8 but plays 31 games?
Click to expand...
Only 30 known opponents right now. I simulate the early season tournaments separately from the schedule regular schedule since there’s some combinatorics required. Those wins are in the column in the table above.
 
Last edited:
#43
#43
Thunder Good-Oil said:
Probably since the 2nd round tournament opponent hadn’t been determined yet.

Edit: But the SEC record doesn’t align. 15-3? 12-6? I guess the difference is considering the W/L probabilities of each game in aggregate.
Click to expand...
Win probabilities. It’s why I show the decimal in the standings
 
#45
#45
golfballs said:
Only 30 known opponents right now. I simulate the early season tournaments separately from the schedule regular schedule since there’s some combinatorics required. Those wins are in the column in the table above.
Click to expand...
21+8 is only 29 though…30 opponents are known.
 
New posts
You must log in or register to reply here.

VN Store



Back
Top