clock menu more-arrow no yes mobile

Filed under:

2014-2015 Tempo-Free Stats Primer

Stats stuff for stats geeks and future stats geeks alike!

Tony Bennett and his Virginia Cavaliers have turned a generation of Wahoo fans into advanced stats proponents. Using commonly reported basic "counting" statistics  leads to a lower level of analysis, especially when using these numbers for predictive purposes. And UVA's slow pace of play exacerbates this issue, rendering comparisons between our players and their opponents meaningless if adjustments aren't made.

Thankfully, "tempo-free" statistics are here to save the day.  Essentially, their purpose is to move the basis of comparison from the game to the possession. Here's an example of the issue: UVA scored 66 points per game last season, while UNC averaged 77 ppg.  Using these "counting stats," which were widely reported by the media, it appears that UNC had a substantially more high-powered offense than the Cavaliers did; after all, they scored almost 20% more points each time on the floor!  However, Virginia averaged 60.6 possessions per game last season, 346th in the nation, while UNC clocked in at 71 possessions, good for 19th.  So, Virginia actually scored more points per possession (PPP) than the Heels did (~1.09 to 1.08)...and arguably had the better offense (though UNC's was nothing to sneeze at either).

This year, as always, we'll avoid the scourges of pacism and use tempo-free stats when reporting on UVA basketball. Here's what to expect. Please feel free to add to what I said, or ask any questions you may have in the comments (now or any time this year!)

The Four Factors:

Basketball geek Dean Oliver laid out the "four factors" that predict basketball success - here are the 4 stats you need to know, in order of impact on the game. (His full explanation from 2004 is here)

1) Shooting (eFG% = (2-pointers made + (3 pointers made * 1.5))/FGA)  )

(For every shot taken, how many points is the offense scoring?)

2) Turnovers (TO% = Turnovers forced / possessions)

(At what rate is the offense turning the ball over, losing possession often without a shot?)

3) Rebounding (OReb% = Offensive Rebounds / (Offensive Rebounds + Opponent Defensive Rebounds))

(At what rate is the offense rebounding the ball, acquiring a "second chance" for points?)

4) Free Throws (FT Rate = FTs attempted / FGs attempted)

(How often is the team getting to the free-throw line?)

Here's some further explanation on why these stats work, and why we should avoid or limit use of other measures.  I'll focus on scoring and rebounding -


Avoid: Shooting percentages without context.

If Evan Nolte and Mike Tobey both scored 12 points on 50% shooting, who had the better game, scoring-wise? (Hint: They may not have been equal).

To know the answer, you'd need more information: Nolte shot 8 three-pointers and made 4. Tobey attempted 12 lay-ups, and made 6. Remember, players who shoot lots of three-pointers may shoot a lower %, but still be more efficient. Nolte only needed 8 shots to score as many points as Tobey did in 12 shots

Use Instead: Effective FG% (eFG%).

This is essentially the same as FG%...but we are taking into account the fact that 3-pointers are worth 50% more than 2-pointers.  The formula is eFG% = [(2's made) + 1.5* (3's made)] / shots attempted.

It's also useful to just split out the shooting percentages into two-pointers and three-pointers to avoid this issue. For example, "Joe Harris made 49% of his 2s and 40% of his 3s last season." This sentence tells us how Harris fared on these two very different types of shots.

Limit: Points per game (and rebounds per game, and steals per game, etc for that matter)

As noted above. That's not to say we'll eliminate this type of language entirety. Saying that "Malcolm Brogdon scored 20 points and grabbed 5 rebounds" is a completely reasonable way to show that he had a good night. It's accurate, easy to understand, and tells us exactly what happened. What we should avoid is comparing players from team to team like this using averages. At the very least, we should understand that Marcus Paige has an inherent advantage in his per-game scoring average because he has more possessions in which to score every night.

Use instead: Points(/rebounds/steals) per possession, or Offensive Rating

All we need to do is change the denominator and look at per-possession stats.  Looking at teams as a whole, the NCAA-average points-per-possession (PPP) number is generally around 1 (1.04 last season). Teams who score over 1.1 PPP are doing well, while those who allow under .9 have the best defenses.  Often, these numbers, also called "offensive or defensive efficiency" are reported in "adjusted" terms, where strength of opponent is taken into account as well. You'll also see them out of 100 possessions (so an offensive efficiency of 104 would be about average).

Last season, Virginia scored 1.14 PPP (21st in the NCAA), and allowed 90.1 PPP (5th), after adjusting for opponent's strength.

Offensive rating is a more complex way at looking at a player's performance.  KenPom notes, "The formula is very complicated, but accurate." That's basically what we'll stick with. Over 1.1 is good, and over 1.2 is great...but it's also dependent on how often the player is used by his team.


Avoid: Rebound margin

Rebound margin ("The Hoos outrebounded the Hokies 30 to 25") is VERY commonly reported, but it just doesn't tell us much about which team rebounded better.  We all know that defensive rebounds are easier than offensive rebounds, so those are what's driving the "total rebounds" number. Thus, teams that force lots of missed shots will have many of these "easier" chances for rebounds.  Teams that make a lot of shots will not give their teammates this opportunity. Additionally, teams that foul often or teams that force many turnovers limit rebounding opportunities, and will look bad in the rebound margin department.

Use instead: Rebounding percentage

1) Since the number of rebounds is influenced by the number of opportunities, it makes sense to look at this as a rate.

2) Since defensive rebounds are easier than offensive rebounds, it makes sense to split these off and look at them separately so we are comparing apples to apples.

A team's OReb (offensive rebound) % = Offensive rebounds / (Offensive rebounds + Opponent's defensive rebounds)

The denominator there is total opportunities for offensive rebounds - those they failed to get were pulled in by the defense.

A team's DReb % = Defensive rebounds / (Defensive rebounds + Opponent's offensive rebounds).

See why this is better than rebound margin? We got rid of any external factors, and have zeroed in on what we want to know - who is rebounding better?  Note that OReb% of team A = 1-DReb% of team B (since every rebound has to be pulled in by someone). So, noting that "UVA's DReb% was 75%, while UNC's was 70%" means that the Hoos beat UNC on the offensive boards 30% to 25% as well.

Last season, Virginia had an OReb% of 34% (90th) and a DReb% of 74% (5th). The league average was an OReb% of 31.4% (or a 68.6 DReb %)

Where to track this stuff?

Okay, so you generally won't see eFG% reported in your morning sports page. There are two pieces of good news.  First, everything above could be calculated with a bit of mental math from a simple box score.  If that's not your cup of tea, there are lots of cool people who keep track of tempo-free statistics and publish them. - The God of tempo-free stats. He's somewhat of a pioneer in this area, and his work is cited often here (and basically everywhere). You can see his rankings and stats on the front page for free, but the rest is behind a $20 pay wall. It's worth it. - The best place to find live, in-game tempo-free stats. - Miscellaneous information, split by team, that goes more in-depth, for example splitting out whether 2-point attempts are dunks, layups, or jumpers.

John Gasaway - Our favorite ESPN talking head. They put him on Insider because he is so smart, but he tweets fun stuff, such as his weekly "Tuesday Truths," where teams are ranked with their conference-mates solely based on efficiency numbers. - Of course.