Follow John on Twitter.

Who is the best overall point guard in the NBA? The answer to this question depends on your criteria. Are you using the eye test or are you more interested in basic stats like points, assists and turnovers. Or do you want to use advanced statistics like Player Efficiency Rating (PER) or plus/minus numbers?

I am a stathead, but numbers only go so far. A point guard’s leadership, presence, poise, vision and defense is just as important as his shooting percentage or his assist to turnover ratio. As a former college player I think I know the game pretty well and I understand that numbers only go so far.

That said, let’s look at some numbers!

I like to use Efficiency Per Minute as a good overall gauge of a player’s production when he’s on the court. For this study, I took the top 30 players in Efficiency Per Game and then added a few players so that every team would be represented at least once.

EPM doesn’t take into account defense (except for steals, blocks and defensive rebounds), so for that side of things, I am putting together a custom statistic that uses both Opponent PER and on/off defensive stats from 82games.com. So, for Tony Parker, I’ll use his 48-minute opponent PER (15.1) and his Net Defense per 100 Possessions (-3.8). Both of these stats are flawed. Opponent PER doesn’t take into account what kind of team defender Parker is while his net defense could be impacted positively (or negatively) if his backup is really poor (or really great) defensively. There may also be bias of strength of schedule (i.e. there are better point guards in the West). But this is what we have to work with.

Using both Opponent PER and Net Defense, I came up with a Defensive Rating that weights each number equally versus others in this study. If the Defensive Rating is positive, he’s a better than average defender. If it’s negative, he’s below average.

Keep in mind that the EPM numbers are for an entire season while the 82games stats are only for a player’s current team, so the defensive rating for players who were traded midseason — Ray Felton, Chauncey Billups, Mike Bibby, Deron Williams, Mo Williams, Baron Davis, Devin Harris and Kirk Hinrich — will only factor in games after they were traded.

Here’s a look at the resulting chart. As always, click on the chart for a bigger version.

A few random thoughts:

— Generally speaking, you’ll find the “best” point guards in the top-right quadrant, though there are some big names (Derrick Rose, Russell Westbrook and Deron Williams) in the bottom-right quadrant. Anyone left of the Y-axis has a below-average EPM with relation to the players in this study. If my Defensive Rating is a good measure, the players in the top-right quadrant can defend well while also filling up the box score.

— Derrick Rose is an interesting case because while his Opponent PER is great (12.7) his team gives up 8.4 additional points (per 100 possessions) while he’s on the court. C.J. Watson is Rose’s primary backup and by the numbers (12.6, -6.9), he’s an excellent defender. But it results in an average Defensive Rating for Rose, so situations like this in mind.

— Steve Nash fared a lot better defensively than I thought he would. His opponent PER is solid (13.8) and the Suns are actually 1.5 points better defensively when he’s on the court. Given these numbers (and his intangibles), one could certainly argue that he’s the best point guard in the NBA.

— Kyle Lowry had the second best defensive rating of any point guard in the study. His Opponent PER (16.0) is about average, but his Net Defense (-7.0) was the second-best of this group. (Ray Felton had the best Net Defense, but he had a limited sample size due to the trade.)

— Perhaps Stephen Curry and Jose Calderon are both getting a bad rap for their defense. Point guards have been just so-so against Curry (15.7 PER) and the Warriors are a little better defensively when he’s on the court. The same goes for Calderon.

— I included Mike Bibby and Derek Fisher, but the two are so bad in terms of EPM that had I plotted their points, the rest of the names would have been virtually unreadable.