clock menu more-arrow no yes mobile

Filed under:

Fun With Stats

Statistical analysis...whether you love it, hate it, or are totally indifferent you can't avoid it.  Every NBA TV broadcast is peppered with statistical references, overt (this team averages 108 points per game) or covert (this is the best rebounding team in the league).  Most of the theories you see advanced at a site like this also contain such references.  In most cases they are quick mentions, drive-by number dropping if you will.  The implication is that the stat cited is the magic bullet that proves the argument without further need for explanation or dissent.  Nothing could be further from the truth.

Statistics are invaluable to analysis, of course.  Like the frame of a painting they provide solidity, boundary, and access to the illustration.  Continuing the analogy they are inadequate when presented in isolation, distracting when they dominate the scene, and downright ruinous when paired with a subject they don't fit.

It would be one thing if only amateurs were guilty of statistical over-reliance and misapplication but I've seen NBA writers, television personalities, and even players misuse statistics horribly.  Johnny Wordsmith can hardly be blamed for pulling statistics out of their context when we see it repeatedly in our sound-bite heavy, analysis light coverage.

So...I figured a primer on statistics would be in order, to at least give us a leg up on the right questions to ask next time somebody says, "The L*kers are the best rebounding team in the league!"

Because the subject is exhaustive, I am limiting this first post to team statistics (as opposed to individual) and I am going to discuss only stats you'd find in a common rundown.  We will eventually get to individual stats and more exotic measurements.  This is to make sure everybody gets in on the ground floor.  I'm also doing it this way because you're never going to hear Mike Rice say, "His effective field goal percentage is spectacular but his PER is a little low."  Most of the mass-marketed claims you hear will use these stats so that's where we'll start.

For each stat we're going to look at what it means, its strengths and weaknesses, and how useful/predictive it tends to be when measuring good team play.

Points For

This stat measures how many points a team scores on average each game.

Can the oft-made claim, "This is the best offensive team in the league" be substantiated using this stat?  Yes and no.  In the most general sense, sure.  Offense equals points.  However the weakness of this stat is it doesn't show how a team scores its points.  For instance they may make a ton of buckets off of offensive rebound put-backs or forced turnovers and run-outs.  This is technically offense but not the kind of shooting, slashing wizardry we associate with the term.  The stat doesn't measure pace either.  Pace tells you how many possessions each team averages.  If one team gets 100 per game while another gets 80 the 100-possession team really should score more each game on average.  But it's quite possible that the 80-possession team has better offensive players and hits a higher percentage of its shots.  In this sense describing offense solely in terms of points falls short.

This is part of why Points For is not a great predictor of team success.  Teams like the Wizards and Warriors (you know that music, old-school Nintendo fans) have been notorious for putting up tons of points but still not advancing in the playoffs, or even getting there some years.  If you go quick enough and allow the opponent enough easy shots you can run your score up into the stratosphere.  Often it means you lose 122-116 instead of 99-88.

Points Against

This is the other shoe:  how many points your opponent scores each game on average.

Because pace is not factored into this stat either it shows even less about defense than Points For shows about offense.  You can at least figure a huge scoring team knows how to put the ball in the bucket somehow.  A team can limit its opponent's point simply by slowing the game down to a crawl even if they are lousy on defense.  In fact one of the preferred strategies for coaches who know their teams are overmatched is the tortoise game.  (See also:  any Mike Fratello-coached team in the last two decades.)  The reasoning is fewer possessions narrows the scoring gap, perhaps allowing your team to stay closer and make a comeback at the end.  It's no accident that the Blazers played slow during their entire rebuilding phase.  Their Points Against standing was relatively good compared to their record.  That doesn't mean they were a good defensive team.  Immediately suspect anyone who uses this stat as a shorthand synonym for good defense.

Like Points For, Points Against is of relatively little value in predicting good teams.

Point Differential

This stat measures the difference between Points For and Points Against.  If you score more points than you allow you have a positive differential.  If you allow more than you score the stat is negative.

So...Points For is kind of a garbage stat.  Points Against is kind of a garbage stat.  This stat which combines the two must be a garbage stat as well, right?  WRONG!  This is actually one of the best predictors of team success...THE best really.

The reason is simple.  Pace can affect offense.  Pace can affect defense.  But whatever pace you play at affects both equally, so a stat that takes both sides into account negates the imbalance.  In other words Point Differential doesn't discriminate between teams that play fast or slow, it measures how much more you score than your opponent does no matter what pace you play at.  Since final score is the measure of winning, scoring more than your opponent on an average night is going to lead to plenty of wins.

One of the arguments sometimes brought up against the importance of Point Differential is the blowout.  A win is a win whether it's by 1 point or 30 but a 30-point win obviously affects Point Differential far more than a 1-point win.  Even more, if a team lost four games by 4 points each and won 1 game by 21 their differential would be +1, which is decent, though their record was a dismal 1-4.  Doesn't this indicate bias? 

In reality that argument only works if you consider a short-term window.  Every stat falls apart when you consider only a few games.  The season is 82 games long.  You'd have to blow a team out by 82 points to raise your point differential one point for the year.  You could also do it with four 20-point blowouts.  So then, to raise it somewhere around 6 points using only the blowout you'd have to get 24 20-point wins.  Guess what?  If you blow out 24 teams by 20 or more you're an elite squad.  There's no way you're losing the other 58 games by 1 point each and demonstrating an aberrant differential.  In fact research has shown (you're going to have to ask  Henry Abbott for the link, as it was a while ago) that blowouts aren't accidents and are in themselves a decent indicator of elite status.

Go back through the stats for the last 25 years or so.  Look at the teams that win championships and go deep in the playoffs.  You're going to find again...and again...and again that these teams also rank highly in Point Differential.  If you're going to pick one stat to measure elite status or predict success, this is it.  Just make sure you let enough games go by to even out the peaks and valleys before you make such a claim.

Field Goal Percentage

This stat measures how many of your shots you hit compared to the total number taken.

This stat provides a somewhat more accurate indicator of offensive prowess than does Points For because it is pace-neutral.  You can't fake hitting or missing shots.  However pace ignorance is also one of its weaknesses when it comes to predicting offensive output.  A team that hits 45% of 100 shots is going to score the same as a team that hits 50% of 90 shots.  Some teams have strong Field Goal Percentage stats but rely on slow-down isolation buckets from a couple stars.  Other teams brick tons of shots but get so many up they score 100 anyway.  In general high percentage shooting teams tend to be among the better teams, both in scoring and winning, but you get aberrations.

The field goal percentage area has been fertile ground for advanced stat computation.  Effective Field Goal Percentage and Adjusted Field Goal Percentage measure field goals by the number of points they actually produce, thus valuing a three-point shot half again as much as a two-pointer.  These attempt to bridge the gap between shooting and scoring and are probably better indicators than just percentage alone.

Field Goal Percentage Against

This stat measures how many of your opponent's shots they hit compared to the total number taken.

This stat has the same advantage as Field Goal Percentage:  pace neutrality.  Whether you play up-tempo or slow-down making your opponent miss is good.  It doesn't suffer from the same drawbacks as its offensive counterpart, however, as no points are tallied for forcing misses.  Making someone miss a two-pointer is exactly as valuable as making them miss a three.  That means the best defensive teams are almost always the teams with low Field Goal Percentage Against numbers.  It's as solid as defensive stats come (which may be damning with faint praise).

Field Goal Percentage Against is not in itself an airtight measure of team success.  Some teams defend quite well but can't score at all.  Many good teams defend as well as they need to in order to maintain an advantage without being excellent defenders.  You'll see teams with relatively weak Percentage Against stats compile good regular season records.  The truth often comes out in head-to-head playoff matchups, however.  Excellent defense really shines when you get a chance to prepare for your opponent and play them repeatedly.

Three-Point Percentage For and Against

These duplicate their Field Goal Percentage counterparts, except they measure only shots from the three-point arc and beyond.

These stats carry the same strengths and weaknesses as their regular counterparts with one added consideration.  If you're going to use these stats to prove or disprove distance shooting prowess, at least in terms of effect on the game, you have to factor in the number of three-point attempts a team takes as well.  The Oklahoma City Thunder, for instance, average around 37% on their three-point shots, good for 11th in the league.  But they take only 10 three-pointers per game, far and away the lowest number among all teams (and it's not even close).  Are they a "better" shooting team than the Knicks, who average around 36% (16th league-wide) but attempt almost 30 per game?  Certainly not in terms of production they're not.

Most successful teams tend to shoot the three-pointer adequately or better.  In isolation, though, Three Point Percentage is not a great predictor of success.  Being in the lower echelon of the league tends to hamper success, however.

Three Point Percentage Against is sometimes more indicative of an active backcourt than great defense.  Some poor teams, in terms of defense and win-loss record, have decent three-point defense.

Free Throw Stats

Free Throws Attempted measures just what it says.  Free Throw Percentage measures free throws made versus free throws taken.

If you want to measure free throw prowess these need to be treated just like three-point percentage.  You have to couple percentage and attempts to get an accurate picture.  It doesn't do a ton of good to have great foul shooters if they never get to the line.  You could say the same about teams which draw fouls but can't hit the shot, but in general it's more valuable to be a team that gets a ton of attempts than it is to be a team that shoots well from the line.  There are two reasons for this.  First drawing fouls puts opponents in foul trouble which is to the good.  Second you can always get hot and starting hitting those extra shots in any given game, whereas a team that shoots well but doesn't draw fouls doesn't get that opportunity for extra points. 

That said, Free Throws Attempted is also subject to the vagaries of pace which is the reason bad teams sometimes get a lot of attempts while great teams don't always.

Turnover Stats

Own Turnovers measures the number of turnovers you commit whereas Opponent Turnovers measures the number your opponent commits.  Turnover Differential measures the difference between the two.

Both Turnovers For and Against reflect pace.  For this reason it's not safe to lean on them solely as measures of taking care of the ball or good/bad defense without considering number of possessions as well.  The "Own" stat is more suspect, as you can control it much more by pace.  If you go slower you will have fewer opportunities to commit turnovers which can make you look better even if you're lousy at taking care of the ball.  At least with the "Against" stat you have to help create the opponent's turnover, which is a small measure of success no matter what the tempo.  However Turnovers Against is not a strong indicator of strong overall defense.  Some teams gamble to create turnovers and are quite happy to do so even though it can mean giving up a higher shooting percentage and more opponent points. 

Unlike with Point Differential, Turnover Differential is not a telling stat.  Some offenses value possession over risk, placing a high premium on the ball and limiting turnovers.  Others reverse that.  Some defenses value pressure over percentage.  Others reverse that.  It's possible for combinations to lie anywhere on either spectrum.  Some combinations yield higher differentials, others lower.  That doesn't mean one combination is better or generates more wins than others.  The best teams are all over the map.

Turnover stats, in general, are not good predictors of success.

Rebounding Stats

Rebounds are split into Offensive (off of your own miss) and Defensive (off of an opponent's).  Combined these yield Total Rebounds.

We've talked about stats that are affected by pace, such as turnovers.  Rebounds are more than affected by pace, they're directly linked to it.  Every missed shot must be accounted for by a rebound in the ledger.  The more shots that are missed the more rebounds there will be.  For this reason teams that create a lot of possessions will always look like better rebounders than teams that play slowly.  You cannot tell whether a team rebounds well or poorly by looking at their rebounding totals, offensive, defensive, or combined.

Rebounding Percentages are a much better indicator of how a team cleans the glass.  Rebounding Percentage divides the total number of rebounds available by the number your team actually got.  Pace doesn't affect this number a bit.  Golden State and Portland are on the same turf.  Measured by Total Rebounds the Warriors rank 9th in the league whereas the Blazers come in at 13th.  One look at the rebounding percentages, however, (.536 to .472 total) shows us that the Blazers blow the Warriors out of the water on the boards.

Over time defensive rebounding will be more indicative of good teams than offensive rebounding.  Mediocre and even bad teams can rank highly on the offensive rebounding scale.  Great defensive rebounding teams tend to be good teams overall.  The disparity is easily explained.  Offensive rebounds can be demoralizing and create extra possessions, often for scores.  However offensive rebounds happen after a mistake...a miss for you.  Defensive rebounds happen after something good...the opponent missing.  Teams that excel in defensive rebounding are also excelling at defense.  Teams that get a lot of offensive rebounds are also missing a lot of shots.  That doesn't mean offensive rebounds are bad.  You should go for every one you can get and pump up that offensive rebounding percentage.  But if you're going to try and predict a team's success based on one kind of rebounding, make it defensive.

Miscellaneous Stats:  Assists, Steals, and Blocks

These stats tend to be catch-alls for things that happen during the course of the game that are noteworthy but not uniquely definitive.

An Assist is simply a score, covered in Field Goal Percentage and Points.  It's a score with a pass contributing directly to it.

Steals are a specialized kind of turnover.

A Blocked Shot doesn't necessarily lead to points or turnovers, it's simply an event.

All three can be affected by pace.  All three are a measure of style as much as effect. 

Teams that run isolation plays for their superstar may rank low in assists even though they are quite proficient at scoring and winning. 

Steals have the same benefits and pitfalls as other turnovers with the exception of being more likely to lead to a direct score.  Note that Points off of Turnovers doesn't measure direct or indirect scoring, though many assume it does.  The fast break off of the steal and the post bucket scored against a set defense on a play that followed the opponent's 3-Second violation both get counted as Points Off of Turnovers.  It should probably be called "Points that Happened to Follow a Turnover which May or May Not Have Been Related to Them."

Blocks are great for intimidation and even better when they lead to turnovers.  They can sometimes be misleading, however.  A classic example came in the days when Theo Ratliff and Joel Przybilla were roaming the paint for the Blazers.  In 2004-05 they averaged a combined 4.6 blocks per game, which is an enormous number.  That also meant that Damon Stoudamire, Sebastian Telfair, and Derek Anderson were letting everybody and their uncle drive right down the lane into the centers' arms.

Being at the bottom of the league in blocked shots is usually not a great sign because it means opponents are free to drive on you.  But there's no correlation between being a great shot-blocking team and being a great team overall.  Assists and Steals are a matter of playing style and there's no easy correlation between their presence or absence and success or failure either.

Hope that helps make things a little more understandable.  I hope it also helps the next time you hear somebody on TV or elsewhere make a claim about a team based on basic stats.  Fine-tuning those (ahem) baloney detectors is a good thing sometimes.

--Dave (


BallHype: hype it up!