The Pitfalls of Statistical Analysis: Statistics Critique

In a 2-part look at basketball statistics, Vicious D takes a look at how basketball statistics have failed.  Today, he takes a look at how to improve basketball statistics, and proposes that the system is not that difficult to change.

In last week's article, I talked about how basketball is a tricky animal and therefore a tricky sport to analyze.  Basketball, unlike say baseball, often relies on a team play to create an offensive or defensive result rather than a group of individual plays.  It's therefore not surprising that team statistics are often the best indicator of how a team ranks in the league.  If the Raptors have the best free throw percentage in the NBA, then it means that the players that are taking the most free throws are the best in the league in doing so.  For my own tastes, team statistics have offered the best and most accurate picture of a team and can be mostly trusted. 

Individual statistics though, still leave much to be desired.

As most casual fans know, the statistics available to the public and the statistics available to the average general manager or coach can be as different as night and day.  In basketball, we've heard coaches talk about calculated deflections (a steal's 'assist' at times) and other statistics that are not readily available.  Individual player tendencies are also looked through and analyzed to give better defensive schemes.  For example, we've heard that many teams examine such metrics as the effectiveness of a player going to the left or right; again, not readily available statistics to most onlookers. 

The only statistics provided by NBA.com are the basic stats that are used to calculate all other alternative perspective breakdowns, such as those available via Wages of Wins, Hollinger's ESPN Statistical Analysis, and 82games.com.  Out of the core 15 statistics available, statistics such as games, minutes, field goals, and shot percentage are pretty undeniable.  They are hard evaluations of players on the court.  However, based on the 15 core categories provided to us from NBA.com, I'm going to break down where some of the other NBA stats have failed:

Assists:  They're defined as a pass that directly leads to a basket, but they're often left to the scorekeeper's subjective discretion.  Unlike assists in hockey, assists in basketball are hard to cover because they happen quickly and with around 100 combined shots taken in an NBA game, there is little time to go back and correct errors .  Add to the fact that there have been reports of inflated stats and it all equals to some questionable numbers.  Too often though, it's a series of passes that leads to a basket while only the final pass that's scored.  To reward those who attempt a team play, I believe Secondary Assists are a statistic that should be calculated.  In a game that has as much scoring as basketball has though, this calculation could be a monumental task .  However, there are no rules that say that stats have to be calculated directly by the scorekeeper after a play.  Secondary assists could be calculated at a secondary location and would allow fans to have a better glimpse at who is initiating the play.  It would also reward a player for their selflessness in making the pass that leads to the pass. Another metric, Assists to Foul Shots, on the other hand, could be easily calculated.  Currently, assists are not calculated for plays where a player gets fouled and doesn't make his shot.  Players that end up scoring on a play are often given a chance to score by the passer.  By setting up a play where the opposing team has to foul, I'd argue that the assisting player should get recognition for instigating the play.

Rebounds: We talk often about how rebounding is a team effort.  However, rebounding is only ever calculated on an individual basis.  What about players blocking out their man?  How about the players who try to do their part in the entire rebounding mechanism?  Unfortunately, basketball doesn't reward them for their work.  I'd like to include a percentage for how many times a player blocks out a man, but it'd be a subjective measurement of which player was playing out of position.  Instead, I believe that Taps (Rebounding Assist) would be a great new calculation.  How many times has a player tapped the ball out to his team?  How many times has a player been unable to keep control of the basketball because he was blocking out his man, but still manages to tap the ball to a friendly face?  Taps may even encourage a different style of rebounding and possibly a faster paced game.  This was something Rasho Nesterovic excelled at in his first tour of duty with the Dinos; he didn't always corral the final rebound, but many a time he made sure the other team didn't either by tapping it to teammates.

Steals: Steals are actually a fairly objective statistic.  Either you're able to take the ball away from your man and retain possession or you aren't.  However, I think that the addition of Deflections as a stat would add to the entire steals metric.  Steals are often the product of another person poking the ball loose or deflecting the ball to a teammate.  Undervaluing that skill set or not computing it would be doing a disservice to those with active hands.

Blocks: Another area that we as Raptors fans have looked at over the years has been in the blocks category.  Part of the reason is because it's been said that Chris Bosh is one of the best blockers in the league.  It's not necessarily because he has the highest block percentage, but the fact that his blocks keep the ball in play.  Block and Retained Possession would be a good statistic to remedy this and quantify blocks in a different way.  If a center or power forward simply spikes the ball out of bounds (ahem - Dwight Howard), they would no longer be seen as a great shot blocker.

+ / - : I find no either stat as infuriating and useless as the +/- statistic.   While as mentioned, I do feel strongly about the effectiveness of most team stats, this one is too reliant on overall team play and is yet used to measure an individual's effectiveness on the court.  This is especially true on the defensive end.  As we've discussed at length with Jose Calderon last season, one player's injury and ineffectiveness can bring down an entire unit's ability to defend.  Why should players such as Anthony Parker or Chris Bosh suffer due to Calderon's ability to move?  With Calderon getting torched at the point of attack, a solid defender like Parker was sometimes racking up -15 type numbers evne though individually, his defence on the night was fairly good.  A much more interesting stat may be Opponent vs Scoring %.  Basically, the stat would be calculated by figuring out the shooting percentage of the opponent a person is directly guarding.  It would not be reliant on who was assigned to which man, but who ultimately was defending the shot.  So for example, if a team were to run the pick and roll on the Raptors and Bargnani successfully altered the shot of his man, his Opponent vs Scoring % would decrease.  Another easily implemented defensive stat would be Charges.  Frankly, I'm not sure why this stat isn't in the books already except to dissuade flopping.  Nevertheless, charges are such an essential part of a defensive stop that they should be calculated and available to the public.

Finally, Fouls Incurred would be another good measure of a player's prowess to enter the paint and draw fouls.  Yes, foul shots already do some of the work, but to have an accurate breakdown that measures how many fouls a player has drawn from their opponent would give a more accurate picture.  For example, if one looks at free throws attempted, it doesn't tell which resulted in two foul shots and which resulted in mere penalties.  Foul shots also do not measure fouls drawn in a non-shooting situation.  The inclusion of this statistic may encourage more players to attack and be more aggressive.

Now that we have a few more suggested statistics, the question becomes how to track these stats.  It's here where I believe a company such as EA or 2KSports would become beneficial.  As video game companies continue to try to bring the basketball court into our living rooms, they require more statistics in order to derive an accurate portrayal of a player's abilities.  The NBA would not even need to fund the calculations, but would just need to extend their partnership with both companies to create better representations of their athletes.  As a result, we in the public would have access to new stats that would be previously unavailable.

Finally, I'd also like to talk about secondary statistics calculation such as those employed by Wages of Wins, 82games.com, or Hollinger's ESPN stats.  While each site provides their own unique perspective on statistical analysis, each also has an inherent weakness.  The first is that most of them are trying to equalize the players by calculating them in certain situations.  Whether it be for crunch time or for the full 48 minutes, these sites try to equalize the minutes across the league to generate more information.  However, there are always reasons why certain players do not play in certain quarters nor are they able to play more minutes.  A player's efficiency will drop off the more they play and the flaws in their game can make them unusable at times.  In addition, most of these statistics do not take into account that the sample pool that is available may be very small for certain players and to simply equate their numbers with those of more established players can lead to inaccurate conclusions.  Finally, these stats are based on a system that already doesn't provide us enough information.  A player such as Andrea Bargnani is not suddenly going to look like a good defensive player if you calculate his statistics on a 48 minute schedule.  It's because the original statistics are already missing vital information, so any extrapolation on the limited data would only provide limited answers.

At the beginning of this two part analysis, I wanted to set out and challenge our way of thinking about how statistics are portrayed in basketball.  How is it that in basketball that there seems to always be more than a handful of players who seem to be much loved and defended by their local fans for doing the "intangibles"?  Why do others often see these players as "overrated"?  It's because statistics in the NBA still have a far way to go.  As a result, it's been my personal policy to be very careful about the statistics I've quoted in the HQ, preferring to stick with team statistics and to look at objective statistics such as shots taken and shooting percentages.  However, there are solutions available out there so that basketball can be more accurately portrayed and players can be more accurately accounted for. 

It will truly be a great day when we can say that the "intangibles" players are becoming fewer and fewer thanks to a more complete statistical breakdown.