The Honorable Cock Jowles: That’s a good analogy, but our differences on whether box scores contain ENOUGH information are predicated on the assumption that basketball players either are or are not responsible (primarily) for their own production. Maybe individual defense is hard to quantify (although conjecture about “bad gambling” on aggressive steal or blocking technique drives me made on this site) but offense is something that the box score does well enough. What could be improved? Percentage of assisted shots, points per scoring attempt, a breakdown of shots by distance to the goal, perhaps a rating system of “points against expected points” (based on the exact spot in which the ball is shot against the player’s career averages)? Right now, the box score gives us 99% of what we need to know on a team level. So you want to throw it out because of the possibility of all of these questions producing substantial noise, but we can still use it on the team level.

As far as “gambling” the problem is that you are not taking into account failed block and steal attempts. If Ibaka jumps on every pump fake to boost his block totals, he will get a ton of blocks but he will also get burned a lot jumping in the air, especially by smart players who realize Ibaka has this weakness. Steals and Blocks are always valuable, that we can all agree with, but CHASING steals and blocks is not always a good strategy.

It’s essentially the same thing as counting all the shots a player makes but ignoring the misses. Every time Monta Ellis makes a bucket it provides value to the team, it doesn’t mean he should be constantly trying to shoot, even though a miss could also be put back to mitigate the damage. Going after high blocks per minute isn’t much better than going after a high points per minute, the goal is efficient defense above all

The Honorable Cock Jowles: That’s a good analogy, but our differences on whether box scores contain ENOUGH information are predicated on the assumption that basketball players either are or are not responsible (primarily) for their own production. Maybe individual defense is hard to quantify (although conjecture about “bad gambling” on aggressive steal or blocking technique drives me made on this site) but offense is something that the box score does well enough. What could be improved? Percentage of assisted shots, points per scoring attempt, a breakdown of shots by distance to the goal, perhaps a rating system of “points against expected points” (based on the exact spot in which the ball is shot against the player’s career averages)? Right now, the box score gives us 99% of what we need to know on a team level. So you want to throw it out because of the possibility of all of these questions producing substantial noise, but we can still use it on the team level.

Actually I will fully agree that these stats work on a team level. Team rebounds, offensive efficiency, defensive efficiency, etc. are well defined, and their relation to wins is pretty clear, although I think it could be tweaked some for the playoffs.

Either way, my problem is that there is no valid proof that the value of team stats can simply be transferred down to individuals. Scoring efficiently is always the goal, I just think that a “finishers” like Novak or Chandler are getting too much credit for their excellent shooting and no penalty for their limitations. For instance, there is no penalty for wasting shot clock for someone like Novak who can’t do anything but reset the offense if his shot is covered. Someone like Kidd or Felton would be able to take advantage of the rotating defense, Novak has to kick it back out and make the possession more difficult

Juany8: Here’s the best analogy I’ve heard to explain this little issue: Imagine you lost your keys in a large, messy room that is only lit near the center of the room, leaving large portions of the room dark and difficult to see in. It is obviously easier to search in the lit area since it provides the most sensory information and thus would be the smart place to look. If it’s not in the lit area though, you are eventually going to have to search in the dark area, in which you’re basically fumbling around trying to feel for your keys.
used when they can be, if I said that Kobe is a better 3 point shooter than Novak, it is very easy to objectively prove me wrong. [start off looking in lit area] But if none of the available data is giving me the answer I need, it’s time to start looking in the dark

That’s a good analogy, but our differences on whether box scores contain ENOUGH information are predicated on the assumption that basketball players either are or are not responsible (primarily) for their own production. Maybe individual defense is hard to quantify (although conjecture about “bad gambling” on aggressive steal or blocking technique drives me made on this site) but offense is something that the box score does well enough. What could be improved? Percentage of assisted shots, points per scoring attempt, a breakdown of shots by distance to the goal, perhaps a rating system of “points against expected points” (based on the exact spot in which the ball is shot against the player’s career averages)? Right now, the box score gives us 99% of what we need to know on a team level. So you want to throw it out because of the possibility of all of these questions producing substantial noise, but we can still use it on the team level.

Frank:

And I have to disagree with your last statement – using stats that might be wrong probably is much worse than not using stats at all, if only because it gives you a false sense of security that prevents you from changing your mind (I will refrain from using your devotion to Berri as an example. Oops, I just did it! =)).Bad stats can be confounded/massaged/misread etc. in any number of ways, and certainly in the medical field, have led to countless adverse outcomes

Here’s the best analogy I’ve heard to explain this little issue: Imagine you lost your keys in a large, messy room that is only lit near the center of the room, leaving large portions of the room dark and difficult to see in. It is obviously easier to search in the lit area since it provides the most sensory information and thus would be the smart place to look. If it’s not in the lit area though, you are eventually going to have to search in the dark area, in which you’re basically fumbling around trying to feel for your keys.

The lit area represents events that can be easily be described in an objective way by numbers, while the dark area is everything else, stuff that can’t be easily observed or arbitrarily defined very easily, but still exists. Saying that using statistics is always better than not using statistics would be like refusing to look anywhere but the lit area since it’s better to look by light. At some point the truth won’t always be found in the numbers, dogmatically tying yourself to statistics means you’re locking yourself out of anything that can’t be measured. Statistics should be used when they can be, if I said that Kobe is a better 3 point shooter than Novak, it is very easy to objectively prove me wrong. [start off looking in lit area] But if none of the available data is giving me the answer I need, it’s time to start looking in the dark

ghill: Interesting.That makes a lot of sense. Taking it a step further…. Are more stats always better than less stats?If so, is there a way to confirm that a stat we do use is legitimate?I’m sure this is Introductory Statistics 101 but I thought I’d ask here.

It would be cool to try and measure some of the items you mention above and see if they are legit.I’m write code for a living so I’d be willing to help people crunch some non-structured data if it helps.

You posed the problem perfectly in your comment, which stats should be used and which shouldn’t be? Personally I think the box score is fundamentally flawed to the point that I find it worthless to look at for anything but very general comparisons. The idea behind statistical analysis is that you can analyze the observed events in a basketball game as data points that have a set value.

Unfortunately, the box score does a poor job of giving credit to more than one person on any given offensive possession, defensive possession, and rebounding opportunity. All of those activities are fully influenced by all 10 players on the court at once, as well as the coach, and disentangling the contributions into discrete, directly comparable numbers for all 10 players is insanely difficult.

I’m pretty good at coding too and do predictive statistical analysis as a big part of my job, however the real problem is collecting meaningful data upon which you can construct a model. For instance, the little personal factors that THCJ gave as an example do have a very real and very noticeable effect on performance. There is just no way to possibly collect that kind of data. I’m not comfortable ignoring real data points just because I have no way of measuring them, to the point where I will subjectively try to adjust for those factors rather than blindly assuming they are negligible.

The Honorable Cock Jowles:
I mean, to what do we attribute missed free throws? Should we assume that a player who plays 40 mpg will have a more difficult time making his free throws than a player who plays 5 mpg — due to fatigue? Or should we consider fatigue-based factors like flight time, timezone changes, hotel accommodations, diet, pre-game warmup routine, scheduling variances, etc.? Or should we try to account for the fact that some players may be more “used to” traveling than others, and thus perform better after long trips? Or should we account for the mental health of each player w/r/t spousal/parental/filial/familial relationships, or their financial situations, or other stress-related performance factors?

My argument has simply been that using stats, however flawed, is preferable to not using stats.

I certainly don’t know this for sure, but my guess is that the people who really do this stuff for a living (Vegas) really do get that granular. They know travel times, how certain teams and players play on this many days of rest, who’s wife is mad at them etc. etc. For people who have no time for that stuff (like people with other jobs), it’s too much detail, but that doesn’t mean that the details aren’t important.

And I have to disagree with your last statement – using stats that might be wrong probably is much worse than not using stats at all, if only because it gives you a false sense of security that prevents you from changing your mind (I will refrain from using your devotion to Berri as an example. Oops, I just did it! =)). Bad stats can be confounded/massaged/misread etc. in any number of ways, and certainly in the medical field, have led to countless adverse outcomes

The Honorable Cock Jowles: This is a valid point, of course, and a flaw with “averages,” as Juany8 keeps saying, but if we start making causal arguments about every game (or possession), it’s going to be very easy to slip into a lot of confirmation bias and subjective analyses that attempt to assume causality for whatever factors we privilege.

I mean, to what do we attribute missed free throws? Should we assume that a player who plays 40 mpg will have a more difficult time making his free throws than a player who plays 5 mpg — due to fatigue? Or should we consider fatigue-based factors like flight time, timezone changes, hotel accommodations, diet, pre-game warmup routine, scheduling variances, etc.? Or should we try to account for the fact that some players may be more “used to” traveling than others, and thus perform better after long trips? Or should we account for the mental health of each player w/r/t spousal/parental/filial/familial relationships, or their financial situations, or other stress-related performance factors?

My argument has simply been that using stats, however flawed, is preferable to not using stats.

Interesting. That makes a lot of sense. Taking it a step further…. Are more stats always better than less stats? If so, is there a way to confirm that a stat we do use is legitimate? I’m sure this is Introductory Statistics 101 but I thought I’d ask here.

It would be cool to try and measure some of the items you mention above and see if they are legit. I’m write code for a living so I’d be willing to help people crunch some non-structured data if it helps.

ghill:
If we consider a player’s stats to stay relatively consistent across a large enough sample (understood there’s not 100% agreement on this or how big that sample needs to be) AND it’s fact that those same stats fluctuate between games THEN shouldn’t we focus on why they fluctuate?It seems like the key to predicting wins is not only comparing the players averages between the two teams but also adjusting for the individual players fluctuation between games.Aren’t things like the opponent’s defensive skills and other things not captured by stats (like Jalen Rose’s night before the game nightclub influence) a driver for the fluctuation in easily measurable offensive stats?

Am I just stating the obvious?Is this too generic a question?

This is a valid point, of course, and a flaw with “averages,” as Juany8 keeps saying, but if we start making causal arguments about every game (or possession), it’s going to be very easy to slip into a lot of confirmation bias and subjective analyses that attempt to assume causality for whatever factors we privilege.

I mean, to what do we attribute missed free throws? Should we assume that a player who plays 40 mpg will have a more difficult time making his free throws than a player who plays 5 mpg — due to fatigue? Or should we consider fatigue-based factors like flight time, timezone changes, hotel accommodations, diet, pre-game warmup routine, scheduling variances, etc.? Or should we try to account for the fact that some players may be more “used to” traveling than others, and thus perform better after long trips? Or should we account for the mental health of each player w/r/t spousal/parental/filial/familial relationships, or their financial situations, or other stress-related performance factors?

My argument has simply been that using stats, however flawed, is preferable to not using stats.

Frank:
@103-
I tend to agree with you re: the regular season, but I think lots of people thought SA was the best team out west last year AND they have more rings than anyone in the last 10 years (or close to it).I don’t remember the Vegas odds but my guess is that SA was a favorite against OKC? And i might have taken SA in a 7 game series against Miami because they have Duncan who would have destroyed Joel Anthony or Bosh in the middle, and they have better ball movement than anyone.

It was fun to watch OKC-MIA, and after the 1st 2 games it looked like it could be a great series, but my guess is that SA-MIA would ultimately have been a more competitive series.

the spurs were so dominant those 1st 10 playoff games. i was really surprised how it ended up. a weird mix of bad breaks for them. some decent adjustments by brooks (who knew?). spurs roll players going cold. durant going off sometimes. harden doing his thing and being clutch. and to be honest some weird lineup choices by pop in the later part of series.

i was pretty certain they would win going into the series.