APBRmetrics

booee, great to see you posting here...and great to know how to pronounce your last name!

Knicks issue might be the market chasing recent reality. Knicks has just missed three straight home spreads by a bunch.

NY (+1.5) lost to Orlando 102-93 (7.5 point miss)
NY (-8) lost to Phoenix 91-88 (11 point miss)
NY (-4.5) lost to Milwaukee 100-86 (18.5 point miss)

Also, both Denver and NY were on night two of a b2b, and there may have been some thought that Denver's depth was better able to deal with that than NY's lack of depth.

NY now 3-13 ATS on the year...meaning market has misread them pretty badly.

I'm probably going to stop with the eyeball PR estimates for a bit because so many starters are out. Duncan sat last night. Nowitzki will miss a week. Tough to have "full strength" numbers for everyone, then "-1 guy" numbers for everyone when we also have so many goofy schedule sequences in the mix mucking things up. Fun to watch the market scramble though...

_________________
Blogging basketball at http://www.statintelligence.blogspot.com/

I don't know if it's equally true for basketball, but for football ATS record can be all over the place. The Saints, for example, went 12-4 ATS this year despite being a popular team that did well. The Seahawks were 11-5 but had an average record. At the other end, the Bucs were 4-12 ATS and the Rams 3-12-1. I think it's more a sample size issue than a misreading issue. In 16 NBA games, I don't think it's surprising that a team has done that poorly, and I assume there's also a team that's done equally well.

"NY now 3-13 ATS on the year...meaning market has misread them pretty badly."

And "market" here means precisely what? The house or the bettors or both?

Did the house want betting in favor of NY to lose?

Does anyone here know or think they know how the betting broke to the two sides in terms of volume and what the financial outcome of this betting proposition was to the "house" / line?

Xbonk, it's not a surprise that somebody's done poorly ATS...but it doesn't mean it's just a sign of randomness within a short sample size that the particular team has done poorly. Knicks haven't played to expectations, whether you're talking about market, media, their own, etc...

*Market had them favored to win 11 of their first 16 games
*I think most neutral observers would have been in that neighborhood
*They've won only 6 of their first 16 games

For those interested, a listing of extremes thus far...

Teams at -4 or worse in ATS W/L differential
New York 3-13
Detroit 5-12
Washington 5-11
Sacramento 6-11
Milwaukee 5-9

Teams at +4 or better
Philadelphia 11-4-1
Denver 12-5
Dallas 11-6
Chicago 11-7
Minnesota 10-6

Everyone else is within +/- 3 games of .500, unless I missed one scanning through.

Will respond to crow momentarily. But, in short...pointspreads define expectations to a pretty good degree...so if a team isn't playing to expectations, it shows up in ATS records. Don't think it's out of line to call it a market misread when a team is that far away from early expectations...

_________________
Blogging basketball at http://www.statintelligence.blogspot.com/

Crow...I think the market is generally defined as "opening line plus action" until a number settles into a "widely available" line as they call it. Oddsmakers put up an opener. In the NBA, it's mostly professional wagerers ("sharps") who are betting at it. The public just doesn't bet much NBA, particularly before the playoffs. I've heard that from quite a few behind-the-line sources. And, there sure wasn't much interest in the NBA when I was visiting Vegas a couple of weeks ago.

So, the "market" in regular season pro hoops ends up being very well informed...a battle between oddsmakers and sharps with very little dumb money getting into the mix. It's not like the NFL where the public lines up on certain powers and the sportsbooks have to defend against that money by shading the line up front against it.

When I was referring to the "market misread," I was talking about how the settled numbers were still way too optimistic about New York's caliber of play. The market was pricing New York as if they were at 11-5 type team expected to earn a top four seed in the East. The Knicks are instead 6-10 despite playing more home games than road games.

Hope that helps clarify what I meant. Any other questions I can help with on the market? John Coltrane would be able to help here as well with a much more experienced voice in terms of the current market in flux. I believe he would know much more about what's going on offshore. My sources are generally in Vegas...and it can be hard to separate who's influencing what in the NBA (many Vegas stores copy offshore numbers for example, rather than doing their own work to determine a number).

For me, the "market" is the sum of the money...oddsmaker's openers plus bets.

I don't currently know if there's been a difference between how oddsmakers and sharps have been reading the Knicks. I can probably go back to the game-by-game line charts at covers to figure it out if you wish. Or, maybe somebody has a handy "open/close" thing for Knicks games this year and we can go off of that...

_________________
Blogging basketball at http://www.statintelligence.blogspot.com/

I assume / understand that the books go for equally divided action normally but how often do they end up somewhat skewed? How close to 50/50 do they aim to get and do they get? Do they ever aim to end up skewed? Or is that not worth the risk to a big book / get them into a different business too much like that of their clintele? I hear what you say about moving the line in the face of lopsided fan support to achieve more balanced action, especially in football. How often do they cut off bets entirely if the action runs quickly in one direction? I am not that educated about this business.

50/50 almost never happens. Books generally have a position on the game because of the nature of betting action.

Football: Square action (the public) is extremely heavy on favorites to a degree that's hard for outsiders to believe...but has been consistent over time. Square action dwarfs sharp action in this sport (particularly the NFL, because EVERYONE wants to bet the NFL). Oddsmakers tend to shade their openers against what they know will be action on the favorites. So, sportsbooks are getting 11/10 in their favor...AND a half a point or a point the best of it. Over a larger sum of games, that really pays off. Sharps (pro wagerers) are generally on underdogs as a result, because that's where the point value is.

In terms of days of the week...sharps will bet football favorites early if they like them, knowing that public money usually doesn't hit until the weekend.

It's hard to make all of this split out 50/50 because it's a mountain of public money and a hill of sharp money. So, sportsbooks generally have positions on the underdog in football.

In the NBA: the public just doesn't bet much. So, it's oddsmakers vs. sharps, with oddsmakers trying to guess what sharps will do...and any mistakes end up leading to sportsbooks having positions against the sharps. The public just isn't there.

Let's say Favorite opens -3.5 and gets bet up to -5. Sharps are on the favorite at -3.5, -4, and -4.5. Maybe some bought back at +5, maybe not. It's very unlikely the action balanced. Sportsbooks are rooting for the dog at +3.5, +4, and +4.5.

Hope that helps explain it some. The 50/50 thing is largely a myth. It's a good goal to strive for because then the losing bets pay the winning bets and sportsbooks pocket the 10% vigorish. It's a goal that's very difficult to reach though because of how the money falls. Square money is a gorilla in the room in the NFL, and isn't even in the room in the NBA...

_________________
Blogging basketball at http://www.statintelligence.blogspot.com/

Thanks for the replies Jeff.

Actually, I think that one has a simpler explanation: Popovich sat Duncan. My rankings, like yours, predicted the line to be about even (0.5 points in favor of Houston). The final line ended up at 3.5 points in favor of Houston, which tells us that the market thinks Tim Duncan is worth about 3 net points over the course of a game. Using my offense and defense rankings (derived from the over/unders), we can even get more specific: My rankings predicted an over/under of 197 for that game. The actual over/under was 193. Using a little algebra, that means that the market value of Tim Duncan is +3.5 points on offense and -0.5 points on defense. Given that the rankings have an average error of 1.5 points, I wouldn't read too much into that -0.5 point. Basically, the market thinks Tim is worth 3 points, and those three points are due to his offense, rather than defense.

I did a similar analysis on Carmelo Anthony last week (he also appears to be worth about 3 points, according to the market).

_________________
Betting Market Analytics
NBA Team Rankings (updated daily)

Ah, thanks. I didn't know Duncan was out.

_________________
The City: A Golden State Warriors-Centric NBA Blog
NBA WOWY: Find statistics of your favorite team with any arbitrary combination of players on or off the court.

Note also that Spurs were in a b2b 4th game in 5 nights, while Houston was off Wednesday and Friday Nights. I'm skeptical that Duncan is worth 3 points at this stage in his career. Much more likely to be a "rest vs. fatigue" adjustment, and the market taking it as a "red flag" that the Spurs were taking the night off given Coach Pop's longterm history of taking nights off on night two of b2b's....

Almost always at least 1 point in the mix right now for a rested team at home vs. a b2b. Spurs were also in a 4-in-5 spot.

Splitter had a huge game off the bench though...25 points in 31 minutes...so Spurs got the money anyway...

_________________
Blogging basketball at http://www.statintelligence.blogspot.com/

That's a good point on rest. When I have the time, I was going to do a study on how rest differences between teams affects point spreads and over/unders.

_________________
Betting Market Analytics
NBA Team Rankings (updated daily)

I tried to work in higher weight for recent games into ridge regressed power ratings, but so far all signs point to this not being a fruitful exercise and this confuses me.

Here's what I do:
I pick a date(1). For that date I want to forecast all games that happened on that date.
I may use data from games that happened on dates before (1) to forecast games on (1).
Usually I'd now fill the result vector with (homecourt adjusted) point differential (HAPD) of those older games
Instead of doing that, I fill the result vector with
HAPD*(factor)^(difference in days between (1) and the date of the game we're currently looking at - 1)

Say I want to forecast games on Jan.5 and our factor is 0.99. Games on Jan.4. get the weight of 0.99^0 = 1. Games on Jan3 get the weight 0.99^1 = 0.99 etc.

To get RMSE I started forecasting all remaining games of the season on the 4th of December of each of the last 10 seasons, so sample sized should not be an issue. I use data from the running season only, which is ~35 days worth of games on Dec4th.

Now the problem is that I get lowest RMSE with a factor of exactly 1. Basically it wants me to weigh all games equally. Factors higher than 1 and lower than 1 give me a higher RMSE. This puzzles me a little and I was wondering if anyone had some ideas why we see that effect or new ideas on weighing schemes.

_________________
http://stats-for-the-nba.appspot.com/

I don't know what's wrong with the math, but I assure you something is. I went through the exercise of creating Bayesian Power Ratings (maximizing predictive power) and there should be a significant drop off over time.

So you're using HAPD*Weight for each game... Are you (at some point) dividing back out the weight properly (i.e. creating weighted average of HAPD)?

_________________
APBRmetrics Forum Administrator
GodismyJudgeOK.com/DStats/
Twitter.com/DSMok1

I don't understand this part. Can you elaborate a little more? Do you happen to have RMSE for the last 10 (or more/less) years using your model (using whatever starting date you want)?

_________________
http://stats-for-the-nba.appspot.com/