Ah, but you can't fool the algorithms! | Page 2 | The Boneyard

Ah, but you can't fool the algorithms!

jake14mw said:
Those rankings are just a tool. It's an algorithm a computer is programmed with to try and rank teams. I'm guessing they adjust them more and more over time, when it is learned what situations it doesn't handle well? No reason to get upset with algorithms.
Expand ...
I never get upset with algorithms, my favorite is when he used to say "lockbox." There's a funny SNL skit on that.
 
CL82 said:
I have to admit it is really counterintuitive that winning a game would make you drop. I mean, I guess, I get it if you're playing a really horrible team, but nova isn't that. I suppose, it is a function of our cumulative SOS and nova cumulative SOS and the fact that it was a one point victory. But still...
Expand ...
We didn't go down, Auburn jumped us. They beat a then 15-2 team by 23 points.
 
auror said:
We didn't go down, Auburn jumped us. They beat a then 15-2 team by 23 points.
Expand ...
I hear you, and yet we did go down. The number 13 team beat the number 22 team, what's that got to do with us.

( I understand the math of it, I'm just talking about how it feels intuitively.)
 
auror said:
Depends on the metric. Some focus on being predictive, some are descriptive and reflect resume strength.

Most of us believe UConn is underrated in the predictive models due to our injuries this year. Most people seem to agree which is why the Vegas lines have been moving towards us after opening lately. Injuries are very hard for models to correct for, because the sample sizes are so small and basketball makes it hard to separate individual players abilities from their teammates. So most models don't really account for it, and sometimes the effect is actually pretty small.

Most people also believe Alabama is overrated because of their disparity in play against higher and lower level opponents. Usually teams that kill bad teams are actually pretty good, and when they play better teams the scoring margin holds to some degree. Some teams are exceptions, though. Alabama certainly appears to be one.
Expand ...
In addition to injuries, are officiating and crowds factored into algorithms? Playing in an arena with 21,000 fans is different than playing an away game in which no one shows up for the home team.

Is it possible the Alabama coaching staff, recognizing their limitations, are playing the algorithm game to get the best possible seeding as opposed to using the weaker opponents to improve overall development? Coaches of mediocre teams that have no realistic opportunity to win their conferences or national championships could run up scores and not put any effort into overcoming their weaknesses in the games against weaker opponents. I guess I'm asking if job security is factored into the algorithms.
 
Sphinx said:
I would only say this to those who defend the algorithms. Reread my examples.. Would you really care to bet money on the algorithm's current projections? Would you really prefer to take Houston or Alabama or BYU (or even UNC) against UConn on a neutral court? Good luck.
Expand ...
You listed 4 teams, 3 are rated lower than UConn.
 
.-.
ctchamps said:
In addition to injuries, are officiating and crowds factored into algorithms? Playing in an arena with 21,000 fans is different than playing an away game in which no one shows up for the home team.
Expand ...
KenPom does give each team a bespoke home court advantage calculated from the last few years worth of data. Not sure if he separates it by specific arena for teams with multiple home courts.

Officials are not considered by regular metrics. Gamblers do consider them in their models.
 
djct1999 said:
Not in one game. But in a 7 game series, or 20 game simulation or better yet, 100, I will go with the computer every time, regardless of teams. Even then, something unforeseen can happen, but the larger the sample set the more likely the computers get it right.
Expand ...
This. Eye tests are absolutely relevant, and in a particular instance might be better than a computer analysis. But over a course of many 7 gamer series, where do you think the smart money is? Computer metrics or one person’s eye test?
 
superjohn said:
Nah, Vegas is in the business of making money.
Expand ...
Houston is REALLY liked by the metrics. We're not close to them there right now. They're considered 2021 Baylor/Gonzaga level right now. They went up since B12 play as well. We would be given a ton of consideration for injuries. It wouldn't make up the difference between them and us. Straight off the metrics (both KenPom and Evan Miya) they'd be 7 point favorites over us. Vegas would list it a bit closer (maybe 4 or so), and then I expect the public would bet it even closer and then Vegas would live with being on Houston's side.
 
auror said:
Houston is REALLY liked by the metrics. We're not close to them there right now. They're considered 2021 Baylor/Gonzaga level right now. They went up since B12 play as well. We would be given a ton of consideration for injuries. It wouldn't make up the difference between them and us. Straight off the metrics (both KenPom and Evan Miya) they'd be 7 point favorites over us. Vegas would list it a bit closer (maybe 4 or so), and then I expect the public would bet it even closer and then Vegas would live with being on Houston's side.
Expand ...
You think healthy Houston would be 4 or so point favorites over healthy UConn on a neutral floor...

Sorry brother but that's nuts.
 
auror said:
Our rating went up (well actually mostly stayed the same), our ranking went down.
Expand ...
What happened with Kansas and Duke yesterday (ratings and rank)?
 
.-.
superjohn said:
Would you take Houston, Bama, BYU, and UNC over UConn on a neutral court?

You think Vegas would make us dogs to those teams on a neutral court?
Expand ...
We’d be favored against 3 of those I would guess.
 
7774 said:
So given they’re terrible why do you think Vegas opening lines align with Kenpom projected scores so closely?
Expand ...
Because they're lazy and use Kenpom?
I share the same doubts when I look at who Houston's played and their record and look back to our win over uNC.
 
Looked at Massey. They have UConn as # 2 behind Purdue with Houston a close #3. They also have something called "power" rating which is per them a measure of potential vs. actual past wins and losses (not sure what that means or what they take into consideration). In order from #1 to #7 power is Houston, Purdue, Arizona, Alabama, Tenn., UConn, Gonzaga.
In team vs. team calculation have UConn losing by 1 to 3 points to all rated higher except Tenn. which is tie and Gonzaga which is tie.
Two things stand out in Massey ratings:
1. Houston has by far the worst SOS in top 10 except for Baylor but seems to not be dinged in terms of "how good their performance has been/will be".
2. Top 15 or so teams are mostly 1 or 2 points favored or not which is going to make the NCAA's wild. As an example has Purdue favored over # 25 St. Mary's by 5 if they played on neutral court.
 
yungdowork said:
We failed to cover the spread last night which doesn't normally help the ratings. Some of the computer models are heavily based on how you perform relative to expectations vs whether or not you won. The saying "good teams win, great teams cover" exists because great teams outperform expectations consistently. I think our metrics are hurt a lot by failing to cover vs some of the cupcakes early on as well. We're only 11-6 vs the spread this year. I think a lot of the algorithms will be more favorable to us as Clingan eases back in and our defensive efficiency climbs back up, but there is room for improvement.

Houston might very well be overrated but their 2 losses are true conference road games against ranked opponents by a combined 5 points. Losing by 15 vs Seton Hall in a game we were picked to win by 9ish hurt us a lot.
Expand ...
You're right if history has no influence on the algorithms.

I point to Houston's high seeding and poor performance in the tournament as exhibit #1 and rest my case.
 
.-.
7774 said:
lol well you certainly summarized why Vegas is insanely successful at making money off sports betters.
Expand ...
So you would bet Houston, Alabama, UNC and BYU against UConn straight up on a neutral court?
 
Scrutiner said:
You're right if history has no influence on the algorithms.

I point to Houston's high seeding and poor performance in the tournament as exhibit #1 and rest my case.
Expand ...
Using previous years' tournament results to try to judge the resume of this year's teams seems like maybe not the best way to do it, but I have been wrong before. Houston's best player wasn't even on the team last year, and they've been pretty successful in the tournament compared to 98% of programs in the country.

The algorithms are just measuring resumes, they're never going to be perfect for determining who would beat who, or who is going to be more successful in the postseason. They're fun little metrics you can use to say your team is better than somebody else and should be ranked higher
 
Sphinx said:
BL's suggestion that I stop worrying about this has me wondering if I should start worrying about this.

Who would have thought that the proposition that UConn is better than the models say would elicit so much support for the models? On a UConn forum at that! Is it because KenPom had us in the Top 5 most of last year when the pundits were ignoring us?

This causes me to recall that, besides KenPom, Jay Bilas was very high on us last March, picking us to win the whole thing.

Now I'm worried -- because besides KenPom rating us outside the Top Five, we are not Jay Bilas's current favorite, either. He's about the only pundit at ESPN who is not on the UConn bandwagon.

Darn you, BL, now I'm worried.
Expand ...
I’m a lawyer. I’m already damned, thank you very much, without needing anyone’s help.
 
Scrutiner said:
Because they're lazy and use Kenpom?
I share the same doubts when I look at who Houston's played and their record and look back to our win over uNC.
Expand ...
If your theory is that large public companies are lazy, and not setting spreads to equalize the money bet on each team so that they maximize profits … well, let’s just say you need a new theory.
 
I think there's a inherent flaw to trusting metrics. It's based on pure numbers not taking into account so many nuances that take place such as:

Teams don't play the same common teams.
Some home courts are tougher than others.
Some neutral courts aren't always neutral (MSG).
Injuries and outlier foul trouble.
Teams that back off often that have big leads which impact statistical values.
The occasional bad game or unexpected blow out.
Some conferences are tougher than others regardless to how talented or not the teams might be. (an actual top team could be in a weak conference but be real good, or in a really tough conference where they lose some tough road games.)

The eye test is important that won't show up in the numbers, such as:

Has the team showed the ability to win and compete in various ways and against different opponent styles.
Can they handle pressure situstions, come back from late deficits, put away teams late in close games.
Do the have players who can make shots when highly contested.
I'm sure we could come up with others.

How often have we seen some top ACC and SEC teams get bounced early when facing a team that plays relentless D, or some deliberate offensive team unable to handle the quickness when they don't have any or many teams that play fast and/or physical in their conference.

I like how last year's and this year's UConn teams can beat you in many ways. I don't see a team out there that I'd be afraid of unless the refs would allow our opponent to impede our off the ball motion, hack our handlers and/or call a lot of incidental contact. I don't think there are metrics for that.
 
DogMania said:
I think there's a inherent flaw to trusting metrics. It's based on pure numbers not taking into account so many nuances that take place such as:

Teams don't play the same common teams.
Some home courts are tougher than others.
Some neutral courts aren't always neutral (MSG).
Injuries and outlier foul trouble.
Teams that back off often that have big leads which impact statistical values.
The occasional bad game or unexpected blow out.
Some conferences are tougher than others regardless to how talented or not the teams might be. (an actual top team could be in a weak conference but be real good, or in a really tough conference where they lose some tough road games.)
Expand ...

Modern metrics account for most of those things...
 
.-.
businesslawyer said:
This. Eye tests are absolutely relevant, and in a particular instance might be better than a computer analysis. But over a course of many 7 gamer series, where do you think the smart money is? Computer metrics or one person’s eye test?
Expand ...
Better yet, that one person being a biased fan.
 
Let's just focus on the eye test. Works every time.
 
Sphinx said:
More and more these days we are passing the eye test with pundits, but it appears that you can't fool the algorithms. If you thought that last night's win at Villanova might fool them, silly you.

This morning we're 6th in KenPom, 7th in Torvik and 8th in the NET, down a notch, I believe, in all three rankings.

The NET, which is said to greatly value road victories, has us lower now, after wins at Butler, Xavier and Villanova, than it did when we had no road wins at all. Mr. Pomeroy, this morning, has us 7 points down to Houston on a neutral court. Mr. Torvik values Alabama's 14-6 record, 2-5 against Quad 1, higher than our 17-2, 6-2. The NET believes UNC would take us on a neutral court, that is, inless you believe the lying evidence of our already having beaten them by 10 on a neutral court.

And don't even get me started on BYU.

Yes, I know, the algorithms are still catching up. Our KenPom defensive rating in now down to 29. But it's a curious kind of catching up that sees us dropping in the rankings. We are the defending national champions. We are 17-2 with 7 straight wins, 3 of them in tough road gyms. How can it be that other teams are passing us in the rankings?
Expand ...
nerds GIF
 
If you don't like those numbers, take a look at Strength of Record. An evaluation of how a team has done against their schedule to this point. Purdue 1, UConn 2.

www.espn.com

2025-26 Resume Men's College Basketball Power Index - ESPN

View the 2025-26 Resume Men's College Basketball power index on ESPN. The BPI is the best predictor of a team's performance of the season.
www.espn.com www.espn.com
 
.-.
You must log in or register to reply here.

Forum statistics

Threads
168,304
Messages
4,562,279
Members
10,454
Latest member
caw2


Top Bottom