Opponent-Adjusted Four Factors

This is something I’ve been thinking about for some time, and I finally gave it a shot yesterday.  I should point out right up front that I haven’t done any kind of testing to see if the way I’ve implemented it makes sense, nor checked to see if the adjusted numbers are more predictive than raw Four Factors data.  But I wanted to get a rough draft of the system done before today’s games are underway – I’ll test and tweak in the off season.

My method was basically to reproduce the Pomeroy ratings, except using (for example) offensive and defensive OR% instead of offensive and defensive efficiency.  I tested the method first on efficiency, just to make sure I was able to reproduce Pomeroy’s numbers.  He actually weights the numbers based on how recent the games are, and I didn’t try to mimic that, but my numbers still essentially aligned with his:

That’s the easy part; I’ve done that before.  The trickier part was deciding what value to use for home field advantage for the various factors.  Pomeroy uses 1.4% per team for each number - he adds 1.4% to the home team’s offense and visitor’s defense, and subtracts 1.4% from the home defense and visiting offense.  I don’t know where he got that number, but I was able to roughly reproduce it (1.25%) by looking at only home-and-home games from this season, finding the average overall efficiency, and the average home efficiency, and then taking the square root of the ratio between the two:

( [Ave Home Eff] / [Ave Overall Eff] ) ^ 1/2 = (103.6/101.0)^0.5 = 1.0125

Since this makes sense, and seems to reproduce Pomeroy’s HFA number, I used this method to come up with the HFA adjustments for the Four Factors:

• eFG%: 1.007
• TO%: 0.980
• OR%: 1.007
• FTR: 1.031

Once I had these, all I had to do was re-do the opponent adjustments using the Four Factors and their HFA’s, instead of the efficiency values.  Then I looked to see if the outputs passed the smell test.  I’d say they look fine; here is a comparison of the raw and adjusted Four Factors rankings for the Final Four participants (click image for larger version):

I highlighted a few of the significant changes.  For the most part, all the teams look better in all the categories, but there are a few exceptions.  Here are a few notes:

• The only instances where the adjustment moved a team’s ranking down more than a couple spots were Butler’s eFG% and Michigan State’s FTR.
• Duke and West Virginia’s shooting numbers are not as bad as they seem – the adjustments pull them up to basically even with Butler and MSU.
• Michigan State’s turnover problem is as bad as it seems – especially with the adjustment improving Butler’s defensive TO%.
• The best three offensive rebounding teams in the nation are still alive.
• Butler doesn’t have as much of a defensive rebounding advantage as I thought.
• The FTR adjustment for MSU is interesting – I’m wondering if that is a result of Big Ten officials “letting them play.”

Of course, now that I have these numbers, the obvious temptation is to use them to predict the Four Factors numbers for tonight’s games.  I of course gave in, but right now I have no faith whatsoever in the accuracy of these.  Just for fun, though:

Eyeballing those, it looks like Butler wins by forcing Michigan State turnovers, and Duke wins by forcing West Virginia into a poor shooting night.  Sounds about right to me.

Similarity Preview: Duke Vs. West Virginia

[NOTE: Cross posted at UFR. For more info on similarity scores and predictions, see here and here.]

Saturday’s headlining game features the offenses ranked 1^st (Duke) and 12^th (West Virginia) in adjusted offensive efficiency, compared with 25^th (Michigan State) and 46^th (Butler). So it obviously features the better shooting teams, right? Nope:

Duke and WVU made it this far not because they make the most of their opportunities, but because they get so many more opportunities than their opponents. They’re both great rebounding teams, and also come out on the plus side of the turnover balance sheet. In fact, when you add together their rebounding and scoring margins, they are the top two major conference teams in the country, in terms of creating extra chances for themselves, with both averaging over +6 in combined Reb+TO margin. Obviously they can’t both keep this up on Saturday, and whoever wins the “extra chances” battle will have a very good chance of extending their season.

Now let’s get more specific, and look at what to expect when Duke has the ball:

Duke’s not going to be able to get a lot going from inside the arc, as they’re below average in 2P% while WVU is on the plus side. And not shown here is the fact that the same is true for Block%, meaning the interior defense of Wellington Smith and John Flowers may end up playing a key role. Flowers had 3 blocks in 23 minutes against Kentucky, and I’d expect the two players to combine for at least that many against Duke.

But as indicated above, Duke’s emphasis is on using offensive rebounding and a low turnover rate to ensure they get a lot of looks at the basket – a good shooting night would just be icing on the cake. West Virginia is above average at defensive rebounding and forcing turnovers, but just barely. Here is how they fared this season against major conference teams in the top 30 in OReb%:

It’s going to be tough to beat Duke if they allow another 40% offensive rebounding night. They were able to overcome that against Kentucky by holding them to a low shooting percentage (37.3 eFG%) and forcing a lot of turnovers (22.6%). They might be able to hold Duke to similar shooting numbers, but likely won’t be able to force them into that many turnovers. Duke should be able to take care of the ball, and get a lot of first and second shot opportunities, but their best looks will be long jumpers. How efficiency they are will ultimately come down to whether they are hitting those jumpers.

You may have noticed above that West Virginia plays at a slow pace, and are wondering whether that will frustrate Duke. No, not at all – the Blue Devils have played 12 straight games at a sub-70-possession pace, and for the year they are 17-1 in games with 65 or less possessions. Shouldn’t be an issue.

Now how about when West Virginia has the ball:

This is going to read a lot like the section above, so I’ll keep it relatively short. West Virginia should also have trouble getting open first looks on offense, but Duke defends the perimeter (1^st in 3P% against, 10^th in fewest 3PA/FGA) better than the interior (39^th in 2P%), so the Mountaineers won’t have as many open long jumpers as Duke will (nor as many as they had in the first half against Kentucky). However, WVU is even less dependent on good shooting than Duke is, as they’re even better at offensive rebounding (and because they shoot more 3’s, there are a few more rebounds for them to grab). They also get to the line slightly more, but it’s a small advantage.

So if everything holds true to form, West Virginia’s offense will look largely the same as Duke’s. They’ll turn it over a bit more, and probably shoot a bit worse (especially from long range). But they’ll draw a few more fouls, and grab a few more rebounds. The core strategy will be the same, though – throw it towards the hoop, then go and get it in the likely event that it misses.

Let’s look at the Similarity Prediction:

Vegas has the Blue Devils favored by 2.5, and Pomeroy’s numbers have them even stronger. The similarity analysis disagrees, showing that both teams have performed nearly the same in games against similar opponents – which makes perfect sense, given that both teams use the same volume-over-efficiency strategy on offense, and both rely on good FG% defense (as opposed to turnovers or rebounding) at the other end. Duke and West Virginia usually miss a lot of shots, but make up for it with excellent offensive rebounding. If one of them starts hitting their first chance shots, or if one can dominate the boards, that team will probably win. Duke is better at shooting and defending the three point line, and that advantage is what separates the two teams statistically. But given West Virginia’s first half against Kentucky, it should be clear that anything can happen in a span of 30 possessions. Duke’s perimeter advantage will likely be less important than the advantage that one of these teams gains through luck – or clutchness, or heart, or whatever you want to call it. So I feel comfortable with the similarity prediction – Duke as a slight favorite, but really anybody’s game.

Something’s Gotta Give: Extra Chances

I noted in a pre-tourney post that Duke makes up for their poor interior shooting by corralling 40% of their misses, which gives them more cracks at the basket.  That’s really only part of the larger story, which is that this Blue Devils team is one of the best in the country at getting themselves more scoring chances than their opponents.  Let me explain what I mean.

Given the national averages listed on Pomeroy’s site for TO% (20%), 3PA/FGA (33%), 3P% (34%), 2P% (48%), OReb% (33%), and FTA/FGA (38%), I estimate that about 2/3 of all possessions are simple one-chance affairs, where a team gets exactly one shot attempt (or one trip to the free throw line).  The actual number’s not important – the point is that one-and-done is the natural state of things.

There are two things a team can do that will disrupt this natural state – grabbing offensive rebounds and forcing turnovers.  We can count how many times a team does these things, and how many times they let their opponents do them, and the difference will represent the “extra chances” that a team has over the course of the game.

Using Duke as an example (numbers pulled from TeamRankings.com):

ExtraChances/G = (OR/G – OppOR/G) + (OppTO/G – TO/G)

ExtraChances/G = (12.9 – 9.7) + (14.2 – 10.8) = 6.6

This can be further adjusted for pace:

EC/100 = 100 * (ExtraChances/G) / (Poss/G)

For the teams in the Final Four, this doesn’t make much difference, so I’ll be ignoring pace for now. (Plus, you could argue that a higher pace for these teams is a good thing, and we don’t want to toss that info.)  I pulled the stats for every 2010 team and found their ExtraChances/G.  Here’s a Google Spreadsheet that shows them all – I’ve highlighted the Final Four teams in blue:



Saturday’s main event will be a battle between the best two major conference teams in the country, in terms of creating extra chances for themselves.  These teams win by exploiting volume, as opposed to efficiency, and obviously they can’t both come out ahead.  If one of them does, there’s your likely winner.

Similarity Predictions: Elite 8

Figured I should get these up before the games start this afternoon.

As I mentioned in the KSU-Butler preview, I think the prediction for that game is skewed by the early season numbers.  Butler’s playing better than they were then.  If I look at only games from this calendar year, then the Similarity prediction pretty much agrees with Pomeroy and Vegas.

I should get a Duke-Baylor preview done this afternoon, and get it posted here and at UFR.

Similarity Predictions: 2nd Round (Sunday)

I'll start with the prediction chart for all of Sunday's games, then get to the previews for the Big 12 games, which are also cross posted over at Upon Further Review.  If you are unfamiliar with my similarity predictions, read about the methods here.  Also, sorry about the ugly tables - I tried a new method of posting them, and it didn't work so well.  You live, you learn.

SIMILARITY vs. POMEROY
GAME			POMEROY	SIMILARITY
Syracuse	-	Gonzaga	Syr +8.5 (78%)	Syr +11 (84%)
Ohio St	-	Georgia Tech	OSU +4.5 (68%)	OSU +7.5 (77%)
Maryland	-	Michigan St	MD +3.5 (64%)	MD +9 (81%)
West Virginia	-	Missouri	WVU +2.5 (60%)	MU +3 (62%)
Wisconsin	-	Cornell	Wisc +8.5 (82%)	Corn +2.5 (61%)
Pittsburgh	-	Xavier	Xav +1.5 (56%)	Xav +2 (57%)
Purdue	-	Texas A&M	Pur +0.5 (51%)	Pur +4.5 (69%)
Purdue w/o Hummel	-	Texas A&M	Pur +0.5 (51%)	A&M +3.5 (79%)
Duke	-	California	Duke +8 (78%)	Duke +9.5 (82%)

Final Four Characteristics

Just a short post to share something I noticed when digging through data last night…

In the Pomeroy Era (2004-09), there have been 24 Final Four teams.  All of them except 2006 George Mason and 2006 LSU fit the following critera:

• 30th or better in Adjusted Offensive Efficiency
• 30th or better in Adjusted Defensive Efficiency
• 15th or better in Pythagorean Rating
• 150th or better in defensive eFG%
• 150th or better in offensive 2PFG%
• 150th or better in defensive 2PFG%

Here are the teams in each region that fit these criteria:

• EAST: Kentucky, West Virginia, Wisconsin
• WEST: Syracuse, Kansas State, BYU
• MIDWEST: Kansas, Ohio State
• SOUTH:

Yes, I left the South blank on purpose.  Duke is disqualified due to their terrible shooting percentage within the arc (46.9%, 205th nationally).  However, they make up for this by being a fantastic offensive rebounding team (40%, 10th nationally), so given the lack of alternatives, I’d slot them in as the best fit in the South by these criteria.

Tickets Punched: #18 - #26

This series is turning out to be more writing than I have time for, so most of today’s teams will be posted without comments.  The charts are all still here, though, and those are really the important part.

Similarity Predictions: 3/13

Yesterday’s experiment went fine, so I’m going to post a few more of these predictions, for today’s major conference championship games – without commentary this time.

Similarity Scores: Title Contenders

This is part 3 (or 5, depending on if you count the posts at UFR) of the Similarity Scores series. The post on how the scores are calculated is here, in case you missed it.  Tonight I'll be taking a look at which teams from the past are most similar to the current AP Top 10, and how those teams fared in the NCAA tournament.

[Data is from games through March 1, 2008.]

The Big 12 post ended up checking in at a Posnanskian length, which is usually a bad thing for anyone other than Joe, so I'm going to keep the commentary to a minimum this time.  Also, there will be a couple changes to the lists themselves.  First, I'm only including teams that made the NCAA tournament.  If they weren't good enough to make it, they can't be THAT good of a comp.  And second, I'm adding two new columns to these graphs: NCAA seed, and PASE (Performance Against Seed Expectation).  This tells to what extent each team exceeded or fell short of expectations, relative to their seed in the big dance.

2010 Syracuse - Historical Comps

SCORE	YR	TEAM	SEED	W's	PASE
90	2009	Syracuse	3	2	0.1
88	2005	Syracuse	4	0	-1.5
88	2008	Kansas	1	6	2.6
88	2006	Kansas	4	0	-1.5
87	2005	North Carolina	1	6	2.6
87	2007	Kansas	1	3	-0.4
87	2006	Florida	3	6	4.1
86	2007	Georgetown	2	4	1.6
86	2004	Providence	5	0	-1.1
86	2008	Georgetown	2	1	-1.4
		Average	2.6	2.8	0.5

Pretty all-or-nothing here - 3 champs, and 3 first round upsets.  But notice that the upsets are all 4/5 seeds, meaning they may have had good numbers, but they apparently didn't take care of business as well as this year's Orangemen.  Limit it to seeds 1 through 3, and you're looking at an average of 4 wins, and a +1.3 PASE.  Or, looking at just 1 seeds, where 'Cuse expects to end up this year, we see 5 wins and a +1.6 PASE.