Now that we know where the Hornets will draft, it's time to start looking at who they should draft, and that means deciding how to judge college basketball players. Today's Staturday introduces HOOPWAR - a statistic created by Benjamin Miraski, a sports journalist with an engineering and finance background. He introduces the idea here and expounds more on the statistical underpinning here.
Rohan and I might very well be using HOOPWAR over the next month or so to help statistically describe draft prospects, and, more specifically, justify our crush on Damian Lillard. Over the next couple of Staturdays, I'll try to determine whether HOOPWAR predicts NBA performance (spoiler alert: I have no idea), as well as identify which players that look like they could be undervalued.
A popular baseball stat is Wins Above Replacement (WAR). It calculates how many wins a player is worth and calculates the difference between that player and a hypothetical AAA call-up that is assigned a particular value. Therefore, the stat tells you how many wins a player is worth above an average player that a team could pick up off the scrap heap.
Miraski wanted to do the same thing, but for college basketball, and his brainchild was HOOPWAR. It's a relatively simple formula:
(Points Saved + Points Earned)*Average Pace/Team Pace - Replacement Level/30
We'll get into points saved and points earned in a minute, but the rest of it is relatively simple. Dividing Average Pace by Team Pace allows us to compare players from different teams that play with varied tempos. Without the adjustment, a player at Kentucky (which plays like they're on roller skates) would have higher stats than an equivalent player at Arizona State (which plays like they're in quicksand).
Subtracting the replacement level is the key to making HOOPWAR an "above replacement level" stat. And dividing by thirty is the way to convert points into wins - Miraski calculated that about thirty points either created or saved is equal to a win in college basketball.
These points are reasonably uncontroversial - the techniques Miraski used are common in sports statistics, and, if changed, would most likely affect every player the same.
The questionable part comes in how he calculates how many points a player is worth.
Points saved = (Defensive Rebounds + Steals + Blocks - Fouls)*Team defensive points per possession
The problems with this formula:
1) All defensive stats are rated equally - does anybody really think a defensive rebound is worth the same as a steal?
2) Multiplying by team defensive points per possession punishes players who play in good defenses. Meraski contends that players in good defenses are getting lots of help. However, a good defense would only help you play tighter defense and reduce the number of points your man scores - a good team defense won't help the individuals boost their rebound, steal, or block rates.
3) As with any basic defensive metric, this doesn't take into account anything that isn't covered with a defensive statistic - a shot altered or a poached passing lane or an excellent hedge on a pick and roll can't get picked up with this system.
The good things about this formula:
1) There is some evidence that equal weighted statistics are better for prediction. You can read about that here (pdf).
2) It uses stats that can be found among all college basketball players - sometimes a challenge, since lower-level basketball games don't always have the most detailed score-keeping.
Now, on to points created:
FG * (2 * (1-TeamAstPct)) + 3PT * (1-(Team3ptShotPCT*TeamAstPct))+ FT + Off Reb + Ast * TeamAstPct * ( 2+Team3ptshotPct) - ( TO * Team offensive Pts per Poss)
Miraski was a tad unclear about what a couple of these stats mean, but I defined TeamAstPct as the percentage of the team's shots that were assisted, and Team3ptShotPct as the percentage of a team's attempted shots that were three pointers.
The bad things about this formula:
1) It doesn't take possessions into account, since the statistic includes field goals and free throws made, not attempted. Based on this formula, a player who takes a hundred shots and makes one is the same as the player who takes one shot and makes one. In short, it doesn't tell you whether a player is efficient or not.
2) Some of the weights seem kind of weird - are assists really more valuable when they're for a three point shot?
The good things about this formula:
1) It takes into account all four offensive factors
2) Once again, it uses available college stats that can go back for years - this value can not be understated.
In general, I think this is a pretty good catchall stat for college. It has some glaring weaknesses, like not accounting for efficiency or strength of schedule, but its simplicity makes it valuable. Over the next few weeks, we'll explore whether HOOPWAR is good at predicting college value, as well as examine which players appear overrated or underrated based on HOOPWAR.
I've calculated HOOPWAR for a handful of players - you can check it out in the table below. Remember - 1 HOOPWAR equals one win above a replacement college player.
|Damian Lillard||Weber State||13.42|
|Harrison Barnes||North Carolina||3.85|
|John Henson||North Carolina||8.04|
|Kendall Marshall||North Carolina||5.07|
|Tyler Zeller||North Carolina||8.58|