Context just as important as numbers in world of stats

BOSTON -- Basketball analytics is becoming mainstream.

That became quite clear at the fourth annual MIT Sloan Sports Analytics Conference, where Rockets general manager Daryl Morey hosted 1,000 other stat-heads at the Boston Convention and Exhibition Center last Saturday.

The conference was made up of 16 panels covering several different sports and many topics that apply to every activity that involves a ball or puck. Several people who were there already have done a fine job of chronicling the day's events, and you can find those pretty easily elsewhere on the Web. You've probably already read at least one or two accounts of what went down at "Dorkapalooza 2010."

If you've been reading this column at all this season, you've seen the value of advanced statistics. Teams can't be evaluated accurately unless pace is taken into account and offense and defense are measured on a points-per-possession basis. A player's value is more about how he affects his team's numbers than what his points, rebounds and assists might indicate on their own.

Having at least a basic understanding of advanced statistics will greatly help in understanding and analyzing the game. There's no doubt that there is a correlation between teams that use analytics and those that succeed, no matter what some may spout to the contrary.

In his Monday column, David Aldridge quoted a Western Conference executive, who referred to the Sloan conference as "people who haven't won anything, who think they have something to teach us."

First, it's amazing that the person who said that -- someone who may have a say in investing tens of millions of dollars on players every year -- doesn't want to know everything posible about players.

Second, it's unclear why anyone in any field would resist an opportunity to learn more, or at least hear a different perspective.

Third, the NBA analytics community has at least one championship ring in its pocket. Mike Zarren, who was on the basketball analytics panel Saturday, has been with the Celtics for six years and is considered to be one of the leaders in the field.

Celtics general manager Danny Ainge may not have needed advanced stats to tell him that trading Al Jefferson for Kevin Garnett would be a good idea. But you had better believe that (like Morey and his staff in Houston) the Celtics have used analytics to determine the best way to defend Kobe Bryant and the Lakers.

Statistical analysis has its limits, of course, especially in a game in which nine other players can have more influence on a player's numbers than the player himself. The only box score stat that isn't affected by a player's teammates or opponents is free-throw percentage.

That doesn't mean that any statistic should be ignored.

A quick example: In Darko Milicic's first game with the Timberwolves last month, he was a plus-35, which for some people probably reaffirmed their belief that the plus-minus stat is meaningless.

The raw plus-minus stat clearly has its limits. But it's never meaningless, even on a single-game basis. It just needs to be taken in context.

It's not like anyone is claiming Darko to be a great player because he had a great plus-minus one night in February. Even the biggest plus-minus advocates wouldn't look solely at that number and claim that the Timberwolves played well because of Darko.

But it's indisputable that the Wolves played well when he was on the floor. While the gaudy plus-minus number might not say much about exactly what Darko did in his 19 minutes, it does help show a general sense of how the Timberwolves' played. From there, it's possible to dig deeper to get the full story.

Essentially, that's what's important to do with any stat in a box score (or elsewhere). No stat tells everything; it all needs to be taken in context. And sometimes, stats are very inconclusive.

One number that many teams depend on is adjusted plus-minus, which uses regression analysis to remove the influence of a player's teammates and opponents.

Another example: Daniel Gibson is a raw plus-145 for the season, a pretty gaudy number. But once it's accounted that he plays with the rest of the Cavs (one of whom happens to be the best player in the world), Gibson has a negative impact for Cleveland on the scoreboard.

For many in the field, adjusted plus-minus is considered to be the closest thing to the holy grail of basketball numbers. But, as with any stat, there's better information with a large sample size. A full season of data may not be enough to make a fair evaluation of a player using adjusted plus-minus. Just ask Mark Cuban, who gave Evan Eschmeyer a 6-year, $20 million contract based on a year's worth of data.

Jamal Crawford, for another example, may have a strong adjusted plus-minus rating, but that doesn't mean Mike Woodson should play him for a defensive possession down the stretch. When you separate offensive numbers from defensive numbers, you find that Crawford's value is mostly on offense.

A skeptic might say that no one needs statistics to tell that Crawford is a better offensive player than he is a defender. It's obvious to anyone that watches a game. But the greatest thing about statistics is that can confirm what is seen, and sometimes they can force thinking in a new direction.

The Nuggets' Dean Oliver had one of the best remarks of the conference. "Individuals see a game better than the numbers," Oliver said. "But the numbers see all the games."

Dismissing the numbers is ludicrous. Relying on them solely is ludicrous, too. To get a true understanding of the game, both an open mind and an understanding of context is needed to make it all work.

John Schuhmann is a staff writer for NBA.com. You can e-mail him here and follow him on twitter.

The views on this page do not necessarily reflect the views of the NBA, its clubs or Turner Broadcasting.