Ah, but you can't fool the algorithms!

[nelsonmuntz]

Teams in the Big Xii and SEC are playing a tougher game in, game out schedule than we are, and we played SIX games OOC against really God awful basketball teams. Not one or two. Six. It doesn’t take a computer genius to wonder if teams that don’t play so many so awful teams might get a boost for that.

We’re playing really well, are #1 in the polls and are in a good place to win our first conference championship in 18 years. If the computer models bother you, why don’t you just take advantage of them by betting heavily on the Huskies every game and view it as an opportunity instead of something to worry about?

Kenpom has the Big East ahead of the SEC. Are you saying you disagree with KenPom or you agree with KenPom?

 

[cordeodato]

It' sports. Brian Cashman gets deceived by algorithms every year.I have a feeling the the Yankees have more money to spend
on programmers and data miners than Ken Pom. Too many variables and mutations galore.

 

[nelsonmuntz]

Including Margin of Victory beyond some simple, minor level was always a bad idea, because it would lead to teams like this year's BYU, that are OK but are pounding teams when they get up so they look very good in the analytics.

I think there should be an analytics of the analytics. I would like to see an analysis of what each team's win prediction was in the middle of December compared to where each team ended up, calculated by analytics' system, measured by actuals standard deviation to mid-December prediction.

The RPI was fine and simple for measuring the quality of a teams' record, and easy for coaches and fans to understand. Now we have a bunch of black boxes that no one really has any idea if they produce good information or not. There may not be another field in the world where a model that is widely used is not audited, peer reviewed or available for assessment to the people who use it.

 

[Mots010]

Who were metric darlings in late January last season? It sure as hell wasn’t UConn after we lost the whole month. We became metric darlings as we started stacking win after win, and then started winning with blowouts in March and April.

 

[WhereistheDove?]

The best know predictor of outcome is the scoreboard at the end of the game. Just take care of business, and improve. We are not nearly good enough when Clingan isn’t on the floor to dominate.

 

[auror]

Who were metric darlings in late January last season? It sure as hell wasn’t UConn after we lost the whole month. We became metric darlings as we started stacking win after win, and then started winning with blowouts in March and April.

There weren't any to the same degree last year. Every team in the KenPom top 10 on Jan 23rd last year was in the top 18 of WAB resume metric too. Tennessee was #18 in WAB and #2 in KenPom, though, I guess that is a significant difference. They had a key injury at the end of the year and still made S16 as a 4 seed. Creighton was probably the closest thing to a scoring margin warrior. They were 13th in KenPom, 17th in Torvik, and 60th in WAB. They almost made the F4. But 13th hardly counts for "darling" status.

This year, Alabama at #8 is 32nd in WAB. BYU at #10 is 30th. Iowa St at #12 is 33rd. The top 7 though, are all in the top 12 of WAB. It's a pretty good top 7 (Houston, Purdue, Arizona, UConn, UNC, Auburn, Tennessee). I would be surprised if the champion didn't come from those 7 teams.

 

[3rdbass]

This is a lot easier than some are making it. UConn is essentially two different teams, based on if Donovan Clingan takes the court or not, combined into one ranking.

When UConn has Clingan, it is at least a top 3 team and the computers reflect that. With a healthy roster, UConn is in the clear top tier with Houston (I think Houston’s defensive extremes kind of break the algorithms but that’s a topic for another day) and Purdue, and UConn would easily be favored over every other team mentioned.

Without Donovan, UConn has not performed like a top 3 team and is probably more of a top 15-20 type team. Without him, we would not be favored against teams like UNC, Arizona or Tennessee and very few of us would (realistically) pick us to beat those teams on a neutral court based on what we saw without Donovan available.

Right now, we’ve had him for 2/3 of our game and not for 1/3 and the rankings reflect that split. The more games Donovan plays, the more our rankings will reflect that full strength version.

I would pick us to beat 2 of those 3. Arizona will be tough but I was impressed how we looked with Donavan out.

 

[twiceahusky]

Lol that's a common myth, but the sports books are definitely not setting spreads to equalize the money bet on each team. And even if they could do that, that is not how they would maximize their profits. That would certainly eliminate all the risk, but the books are willing to put their capital at risk for the sake of bigger profits.

Sports books absolutely take positions on the games, and they do it with a level of research and analysis that is far more refined than most bettors'. They are betting against the public, and winning.

A better way to think of it: the sports books want to solicit bets in a proportion that optimizes the outcome of their risk/reward analysis, which is not necessarily an equal amount bet on each side.

Are there any books or verified sources on this? It makes sense, but I always assumed they just balanced the money by sliding the line, and took the vig.

 

[Hans Sprungfeld]

I don't know.

If you can find a place to make a bet after the game's over, you gotta be able to do better than the fools who take their chances and bet before they know. Think about it.

Maybe it's as easy as just finding the right betting site that's a little more flexible with an abstract concept like time and not so picky about rules & such. Maybe back off on being a semantics-wielding buzz kill flexing superiority with bold formatting.

 

[Leebo]

Who were metric darlings in late January last season? It sure as hell wasn’t UConn after we lost the whole month. We became metric darlings as we started stacking win after win, and then started winning with blowouts in March and April.

On Torvik, UConn entered the skid #4 and left it #7. Were you thinking it would have been more dramatic?

 

[uh-no]

Who were metric darlings in late January last season? It sure as hell wasn’t UConn after we lost the whole month. We became metric darlings as we started stacking win after win, and then started winning with blowouts in March and April.

lol uconn never dropped below 6 kenpom during the skid last year. they were metric darlings for almost the entire year. it was the pundits who ruled them out.

 

[Mots010]

lol uconn never dropped below 6 kenpom during the skid last year. they were metric darlings for almost the entire year. it was the pundits who ruled them out.

I did think was more dramatic, but we looked like we forgot how to play basketball during that January stretch. No one anywhere thought we were a threat after losing that 6th game. Hell, after nearly blowing a huge lead to providence then losing to Marquette in the BET, no one predicted we’d blowout every game in the NCAA’s.

Honest question, are the metrics telling us we’re not as good as we think? We look so damn tough this year.

 

[pepband99]

It's not rocket science. We played about as expected vs creighton and villanova (worse on O, better on D). You don't move up in KP by playing as predicted.

This is still the simplest and most correct answer to the OP. Kenpom, in particular, is meant to be predictive. It's amazing a thread like this starts right after a game KenPom predicted as a 2 point win, and we ended up winning by one, almost on the numbers, too.

I always have the same answer for people who shun the numbers - make a better one. You'll be rich beyond your wildest dreams, if successful.

The alternative - a poll (which is usually filled out by half-interested journalists, and the 5th assistant for the coaches poll) isn't a great measure of anything.

 

[navery12]

I did think was more dramatic, but we looked like we forgot how to play basketball during that January stretch. No one anywhere thought we were a threat after losing that 6th game. Hell, after nearly blowing a huge lead to providence then losing to Marquette in the BET, no one predicted we’d blowout every game in the NCAA’s.

Honest question, are the metrics telling us we’re not as good as we think? We look so damn tough this year.

No, the metrics are saying we're one of a handful of teams with a chance to win the championship this year. Just like they did in January last year. Just because you overreacted to a bad month last year doesn't mean everyone else here did the same thing

 

[Mr. Wonderful]

I don't know how many people will understand this nuance, but...

Calling models like Kenpom and Torvik "predictive" is wrong. The term you're looking for is "indicative," because those models are not even structured in a way that predictions can be made without adding further logic to the data.

And no, this is not pedantic. If Kenpom was analyzed as a results based predictive model, it would fail horribly. It's frequently incorrect on that granular level.

 

[giddyup]

lol. Good stuff. That was my really my point. There is no such thing as an error free predictor. If there was there would be no gambling.

 

[The Yakk]

This year, Alabama at #8 is 32nd in WAB. BYU at #10 is 30th. Iowa St at #12 is 33rd. The top 7 though, are all in the top 12 of WAB. It's a pretty good top 7 (Houston, Purdue, Arizona, UConn, UNC, Auburn, Tennessee). I would be surprised if the champion didn't come from those 7 teams.

Add Kentucky and I like your list. Kentucky is no joke.

 

[The Yakk]

Are there any books or verified sources on this? It makes sense, but I always assumed they just balanced the money by sliding the line, and took the vig.

The total Vegas handle on most games can be found, and it is often quite different than 50/50.

For example, close to 80% of the public money was on UConn to roll Arkansas in the sweet 16 last year. That's one that Vegas got way wrong, which was beautiful.

Let's say you are running a sports book. If your goal is to have a 50/50 spread on each game so that you can just earn the vig risk-free, then you have two problems:

1. Ensuring a 50/50 handle is not as easy as it sounds. There is a lot of guesswork in that, and if you get it wrong you are exposed to a potential loss. And once you have that exposure, then you need capital to back it up.

2. The vig alone is simply not enough profit for the sports books. Since they are putting capital at risk (because of 1.) then they want to earn a hell of a lot more profit than just the vig.

All the sports books' research and analytics are not for the purpose of setting a line with a 50/50 handle. Rather, it is for predicting a range of outcomes for the game--something like a mean result with a variance in either direction. The book then tries to set the line to entice the public to bet on the losing side of that range.

They don't get it right every time, but they do often enough to make a nice return on the risk they are taking.

I think stuff like Ken Pom is a great help to bettors, and is reducing a lot of the information advantage that the books have always enjoyed. But the bettor has to use that information correctly.

Edit: As far as "sliding the line":

Yes, the books do that if they misjudged the public and they have too big a position on one side of the bet. But they are not necessarily trying to slide their position back to 50/50.

And there are limits to how much and how fast they can move the line, because they might create "middles" that can be exploited.

 

[CL82]

lol uconn never dropped below 6 kenpom during the skid last year. they were metric darlings for almost the entire year. it was the pundits who ruled them out.

Post/handle

 

[auror]

Add Kentucky and I like your list. Kentucky is no joke.

Maybe. They don't really play defense.

 

[martyn]

So given they’re terrible why do you think Vegas opening lines align with Kenpom projected scores so closely?

Because Vegas likes algorithms as much as Kenpom does. But after watching that struggle against a not great Nova, I wonder..,

 

[RedStickHusky]

We've got to stop playing so many crap teams in the beginning of the season.

Why? Our metrics are setting us up just fine for the post season and will likely improve as other top teams take losses. The net matters for tourney seeding the rest are there to feed our trash talk (or betting for those inclined). You don't get trophies for January kenpom rankings.

 

[cordeodato]

I find it hard to believe Vegas books are directional players. It has been my experience that directional players have to lose in the long run.
Also if these guys that run books are so good at picking winners why bother with running a book at all. Why not sit on a beach in the Bahamas and make your bets from there, Why split your winnings with anyone?

 

[The Yakk]

I find it hard to believe Vegas books are directional players. It has been my experience that directional players have to lose in the long run.
Also if these guys that run books are so good at picking winners why bother with running a book at all. Why not sit on a beach in the Bahamas and make your bets from there, Why split your winnings with anyone?

You might ask the same question about wall street traders, or reinsurance underwriters. Lots of reasons, here are the main two:

1. The individual on the beach in the Bahamas has a limited bankroll; not nearly enough capital to ensure a win in the long run, unless he is happy making $20 bets everywhere. The books don't win every bet; they make their money in the long run by having the best of it on most bets. No different than the blackjack table. But the book needs a lot of capital to sustain the losses that will inevitably happen.

2. The casino is taking the bets, so they have the advantage of the vig on every bet. The genius on the beach is making the bets, and has to overcome the vig before he can make a profit.

I'm not sure what "directional players" are, but if you mean the average gambler then yes they lose in the long run because of 1) the vig, and 2) a deficit of analysis/information/discipline. To be a long-term winner, you need a hell of a lot of 2) in order to overcome 1).

It is possible, by the way. The books call those bettors "sharps", and they try to avoid taking bets from them because they prefer to take bets from people who are at an information disadvantage. The sharps' research is as good or better than the books', which is a problem.

If the business were simply a matter of getting a 50/50 handle, then why the hell should the books care about the sharps?

Still, it's not surprising you find this hard to believe. The myth of the 50/50 vig-earning sportsbook has been ingrained in the gambling public for years. I think the books like it that way.

 

[MU in MKE]

3 points.

1. Most of the analytics models do not account for wins and losses of games, they are looking at every single possession. That is what the numbers are based on, not the final score. This is why NET removed the margin of victory portion of their model (but in the case of NET, they did keep some portion for the game outcome, unlike Kenpom/Torvik/etc)

This means you can play like crap and win, but still fall in efficiency numbers (and thus analytics ranking) or LOSE a game but have played really well and thus see ranking actually go UP after a loss.


2. The data is not taken inside a bubble.

• Most of these analytics take into account the level of the team you've played, and how they've done before and since you've played them.
• Additionally, and perhaps more importantly on a day to day basis, other teams doing well/poorly can jump or fall below your team, even if your team hasn't played on a given day.

3. If you're concerned about outcomes of NET, remember that UConn's NET ranking doesn't matter to UConn. Don't worry what the NET is for this team, just focus on the teams the Huskies have played. The selection committee doesn't seed teams based on their NET, they use the NET of their opponents as ONE OF the tools to judge quality of a resume.