Haslametrics, an alternative to KenPom

[_kai_]

http://haslametrics.com/

Really cool website I stumbled on Twitter today. At the bottom there is an All-play estimate where you select a team, then select neutral, home, or away and then it'll show you the likely outcome between the selected team and the rest of D-1. Give it a try and see how many teams it predicts to beat the Shockers. A little spoiler: not many.

[Shocker1976]

Predicts seven losses on neutral court; no blowouts and several coin flips:

Duke - 69.90 vs WSU - 69.09

Iowa - 66.89 vs WSU - 66.53

Kansas - 67.34 vs WSU - 65.32

Villanova - 61.55 vs WSU - 58.38

Kentucky - 67.67 vs WSU - 62.37

Michigan St - 66.12 vs WSU - 61.44

Virginia - 58.55 vs WSU - 53.13

[Shocker1976]

Predicts three home court losses and sheds light on why P5 schools are reluctant to schedule home and home games with WSU :

Michigan St - 64.23 vs WSU - 63.32

Kentucky - 65.79 vs WSU - 64.28

Virginia - 56.80 vs WSU - 54.83

[Jamar Howard 4 President]

It predicts only 7 losses on neutral courts because it has WSU as #8 overall. It is the high rank that is driving those predictions.

[Shocker1976]

It predicts only 7 losses on neutral courts because it has WSU as #8 overall. It is the high rank that is driving those predictions.

Yes and the metrics in his table drive the rankings.

[Jamar Howard 4 President]

Any metric that currently ranks WSU as #8 nationally seems to be pushing the bounds of plausibility IMO. Heck, I love KenPom, and even I think his formula has WSU ranked a little too high at #13.

[proshox]

I am guessing his tendency to truncate games inflates our rating.

[Shocker1976]

I have no skills regards analysis via metrics; for me this is one man's opinion just as KenPom is one man's opinion.

[OregonShocker]

Any metric that currently ranks WSU as #8 nationally seems to be pushing the bounds of plausibility IMO. Heck, I love KenPom, and even I think his formula has WSU ranked a little too high at #13.

It may be the weight KenPom places on margin of victory; difficult to extrapolate into other games without a fair amount of subjectivity.

[Shocker1976]

I am guessing his tendency to truncate games inflates our rating.

From his website regards his data:

"First of all, I have sworn off several of the more popular methods endorsed by hoops stats enthusiasts to rate teams. This includes, most notably, Dean Oliver's "Four Factors of Basketball Success" (effective field goal percentage, turnover percentage, offensive rebounding percentage, and free throw rate). Together, these variables are frequently utilized to construct a regression equation to predict offensive and defensive efficiency. After experimenting with linear and logistic regression for years, I have chosen to reject that methodology in my predictive analysis based on the opinion that sample sizes for one particular team in one single season are just too small. I believe there just isn't enough data available to properly formulate a reliable equation that way.

Second, I have narrowed my analysis down to the bare necessities.....specifically, how often teams shoot, how close to the basket each shot is, how well teams shoot from different locations on the floor, and how often steals and offensive rebounds affect the shot selection and success. One must also factor in those same traits from a defensive perspective (e.g. how often, how well, and from where a team's opponent shoots). Using this shooting data alongside an estimate of the number of trips upcourt a team and its opponent will make, I can scientifically make a prediction for the result of any contest.

Third, based on play-by-play logs that I have collected and parsed, I only utilize data for a particular game where the outcome of said contest is still in question. Using a formula to determine when a game is essentially "over," I can truncate data that is likely to be contaminated by bench players ("scrubs") getting time on the floor when a lead is out of reach.

I should mention that, while play-by-play data is available for over 90% of college basketball games, it is not available for all games. Despite this fact, I still get more informative "bang-for-the-buck" statistics from using play-by-play data vs. the box score data typically used by other stats enthusiasts. In cases where play-by-play data is not available, I attempt to extrapolate as much data as possible from the box scores.

The algorithm I utilize knows nothing about each team's history and, therefore, treats all teams as absolute equals on Day 1 of the season. As a result, you may see some unfamiliar names near the top of the rankings in the first month of the season as the algorithm continues to build a larger and larger sample set. Over time, however, the data settles in and the ratings become more and more accurate."

[_kai_]

I was talking with Erik earlier on Twitter. If you guys have specific questions for him, he said he'd answer them.

[Shocker1976]

I was talking with Eric earlier on Twitter. If you guys have specific questions for him, he said he'd answer them.

Erik is quick to respond and informative IMO!

[OregonShocker]

The best metric is the one that most accurately parallels NCAA invites; cause, in the long run, isn't that what we all want to know.

[Shocker1976]

Any metric that currently ranks WSU as #8 nationally seems to be pushing the bounds of plausibility IMO. Heck, I love KenPom, and even I think his formula has WSU ranked a little too high at #13.

Erik provided the following answer to your question -
"It's driven by the All-PlayPercentage (AP%).....which is different from the Pythag method that somany in sports analytics use.

For every possible matchup, there has to be awinner and a loser. So the All-Play looks at all 61,425 possible matchupsin Division I and asks the question, "If X played Y on a neutral court,who would win based on this season's historical data?" There's always awinner and a loser. So the algorithm solves for all 61,425 matchups and picks awinner for each depending on how that team has performed this season. Virginiais currently 350-0 in its matchups with all possible opponents on a neutralcourt.

Expectedly, Kentucky had an AP% of 1.000 goinginto the postseason last year. Now, does that mean that Kentucky was sure todefeat every single one of the 350 possible opponents last year and the samefor Virginia this year? No, obviously not. (My Badgers made sure of that lastspring.) But if you had to pick a winner for every single possible matchup, myalgorithm does just that. There are also some overlaps here and there (A beatsB, B beats C, C beats A). In essence, the AP% and the predicted outcomes at thebottom of the page are setting a spread. Though Virginia (my current #1) wouldalmost certainly lose a few games if they played all 350 opponents on a neutralcourt, it's not surprising that the #1 team would be favored in each of thosecontests."

[ripemupshocks]

http://haslametrics.com/

Really cool website I stumbled on Twitter today. At the bottom there is an All-play estimate where you select a team, then select neutral, home, or away and then it'll show you the likely outcome between the selected team and the rest of D-1. Give it a try and see how many teams it predicts to beat the Shockers. A little spoiler: not many.

I like it.