Stats: Handy or Useless?

I dislike Hollinger. Using stats is one thing but he ONLY uses stats and trusts them even if he doesn’t think them true. Really irritates me.

Yeah I’d agree with that, I don’t look at stats enough. I trust my eyes and base my judgements from there from suspiscion of stats. Should check them a bit more and see whats out there. Lots of new ones these last couple of years.

According to stats, Scals is useless. That’s not 100% true, but it’s not far off either.

Could someone explain how PER stats work? Is it just points/rebounds/etc per minute? Does it matter when the minutes were played: crunch versus garbage time for example?

I like /- the best. It captures “intangibles” and defense better than any other stat. The one shortcoming is that players’ numbers can be a product of the rotation that they play with. But I think, over the course of the season, and certainly over the course of several seasons, players are on the court with too broad a combination of players to simply attribute their /- numbers to their floormates’ performance.

By the way, Garnett’s /- is off the charts! And check out Rondo’s, too…

My wife is a clinical psychologist, and from what she tells me about the studies done on memory, I don’t think that our “lying eyes” are as reliable as we think. Once we think something is true, our brains have a tendency to disregard or minimize contrary evidence. To me, stats help me to question or confirm my presumptions. Also, a stat is only good for what it is supposed to measure. I like per 48 minutes stats to show me what a part-time player’s potential is. But, I don’t extrapolate that this is what they can do as a starter, because I understand that staying on the floor is in itself an indicator of skill and ability. Same with Hollinger’s PER. A player like Pierce who has been asked to pay outside of his game for the last 4-5 years is going to have more turnover, forced, shots, etc. that make PER go down. As long as I know to take that into account, I still like the PER stat to compare players’ value across positions better than points/rebounds.

According to stats, Scal is not useless; he’s less than useless, detrimental to the Celtics’ cause every minute he’s on the floor.

W-L. The only stat you really need to know. Everything else should be up to the coach and GM. They should know the strengths and weaknesses of all the players. Data and statistics can be shaped to say anything at times. Agents certainly use these kinds of things to skew the truth. Come to think of it, so can a president when he wants to go to war.

PER is pretty dumb. But the worst one is the ‘’per 48 minutes’’ stats people always throw around.

#1, No one plays 48 minutes.

#2, players play the minutes they ear, so if a guy is playing 5 minutes, it’s ridiculous to quote his per-48 stats.

1. (the MOST important point), even if that guy did play 48, his per-minute production would decline precipitously with the extra time. It’s called the Law of Diminishing Returns. It’s an economic theory, but it applies to almost everything in life.

It’s rather simple actually. Just ask Cousy or Russel or even Tommy how much stock they put in statistics. What it really boils down to is a player’s character and his willingness to sacrifice himself for the team and play whatever role is needed of him to get the win. If that means setting solid picks and moving the ball, then so be it. If it means scoring 30 points, then that’s fine too. But if it helps the team win and the win would not have happened with one of these guys missing, then they are both equally valuable. That goes for all the teamates that contribute to the W.

Statistics are good for fantasy leagues and people like Hollinger who have never laced up a pair of shoes. In order to gauge a player’s value, try to envision what role he would need to fill and what are his strengths and weaknesses based on that role. And by strengths and weaknesses I don’t mean his PER stats. Just like there is more to defense than rebounds, blocks, and steals… there is more to offense than points, assists, and percentage.

Um, can someone take Darko out back to the woodshed and straighten this punk out?

Sure, /- is a useless stat on a game by game basis, but over the course of the season, or several seasons, for a player who plays 25 or more minutes a night, you can’t attribute /- stats to the other players on the floor. Any given player is likley to spend significant quantities of time playing with 8-9 other players over the course of the season.

achansen- saying that stats don’t matter is like saying that the C’s were champions last year because they sure tried really hard and were good guys. Life is about results. A player may not fill the whole stat sheet and be considered valuable, but he better be really dominant at a couple of those columns!

migit – you are dead on with that post. That is why we saw our young guys as much more valuable than the rest of the league did. We liked them and therefore ignored their deficient areas.

Stats are information, and the more information you have, the better. You wouldn’t look at your computer and say: Well, the specs tell me I don’t have enough memory to perform this function, but darn it, I’m going to trust my eyes— it looks like a pretty good computer to me!

That’s not to say stats are always accurate or helpful, but people who say that stats don’t matter are essentially disregarding the whole language with which we talk about sports.

That said, I’m a little skeptical about stats like PER that attempt to turn a player’s value into a number. Not that it doesn’t have its place, say, in making comparisons between players, but I prefer to look at a lot of different stats and draw a complete picture that way.

Also, I do agree with those who say basketball isn’t the ideal sport to analyze with stats. In my favorite stat sport, baseball, pretty much everything can be expressed by the numbers, and people who put intangibles ahead of performance are (in my judgment) kidding themselves.

Basketball flows so fast and depends so much more on a direct physical contest between players that I’m willing to admit intangibles play a much bigger role in this sport— Larry Bird and Michael Jordan certainly had ’em. But even in basketball, perceptions are pretty useless without a way to check them. Thus, stats remain at least somewhat helpful.

PER would be better as an offensive-only stat. Steals and blocks don’t necessarily measure a player’s defensive skill anyway, so why include them in the number? Bruce Bowen gets very few steals and blocks, so his PER stinks even though he’s the best defender in the league.

Hollinger should give up all pretense of measuring a player’s total worth and focus just on offense, which is much more quantifiable than defense is (yes, you can’t measure a pick very easily, but I said “more quantifiable”, not “perfectly quantifiable”).

I think I meant “offense-only”, not “offensive-only”. :)

ejplaya – When I refer to a players Character, I don’t mean if he is a good guy or if he molests collies. I mean if he is willing to do whatever it takes to get the win. Now of course trying to get the win is different than actually getting the win. The only stat that counts is the score. But a player’s worth should be measured by how he contributed to that score, providing it was higher than the opposition. Because last years Cs played hard and were willing to sacrafice and still lost reflects the fact that they weren’t good players. (Which doesn’t mean they may become good players) By the way, what is Bruce Bowen’s PER? Does anyone doubt that he is an integral part of San Antonio’s success?

Not sure why the whole world thinks Bowen is such a brilliant defender. He’s solid. Are there any stats out there that show the number of turnovers caused by him, or the fg% drop of the player he is guarding? That might convince me. I’d take Prince over Bowen as far as defensive ability.

CelticsBalla32: But if that’s the case, then Average Basketball Fans aren’t using statistics properly. You can’t take one number— even scoring average— and build an opinion around it. But what about a player who has higher scoring, rebounding, and assists and a better field-goal percentage than another player, while commiting fewer turnovers? Surely the numbers matter then!

Or what if two players are essentially equal otherwise, and that extra 3 points a game is the main difference between them? In that case, it matters. Statistics should never be used to justify an argument, but they should be used to inform an argument. And yes, they do tell you some things that you might not notice just by watching the games.

EJPLAYA: I really wish the NBA would keep a stat called “field-goal percentage against.” I think I saw something like it in something of Hollinger’s once, but having that as a major, accepted stat would be huge for evaluating defense statistically, even though it would probably be a pain for the scorekeepers.

EJPLAYA wrote: Are there any stats out there that show the number of turnovers caused by him, or the fg% drop of the player he is guarding?

Maybe there is, maybe there isn’t. Who cares. San Antonio likes him around and guess what. They know a thing or two about team basketball. The point is that Bruce Bowen is a piece of a championship puzzle regardless of his stats. Stats people would rather have a statistically better player than a player that contributes to wins. Intagibles go a long way towards a team’s success and it’s not just a cliche.

JB – Are you telling me that John Hollinger knows more about talent evaluation than NBA GMs? If that’s the case then why haven’t any teams hired him. I guess being a cheesy sportswriter is more lucrative and exciting than being a GM.

JB -

I really liked that article. He also discussed how stats weren’t the end all be all, for example the fact that a guy was injured and out of the league was something traditional scouting could have predicted, but not his stats.

His method did a great job of maximizing NBA potential AND minimizing busts. I think its a solid way to evaluate college (and Euro league) draftees, especially for an established team, that needs to add contributors, but is unwilling to take long shots (say on a GG.) His method also liked Rondo fwiw.

I have not been able to access the 2007/8 PER projections, every link I click, redirects me to the 2006/7 projections. Very frustrating.

One thing about PER is that it doesn’t take into account games. Maybe a player doesn’t control his minutes, but you do need to take into account games as well. Take a guy like Camby, as good as he is, your backup center is going to see starters minutes for 20 games a year, and your third string guy is going to see backup minutes during that time.

SABR guys in baseball often fill in playing time at replacement level to evaluate durability, i.e. a guy like Bonds who might only play 130 games, gets 32 games at replacement level added in to make up for his inability to play every day, while another guy who may be a lot worse, only has 3-4 games added in. I also agree that their are some guys who look like HoFers in a reserve minutes, that would look like NDADLers as starters.

Finally I haven’t seen a site with it, but I do like the idea of offense only PER (ooPER?) and offense only PER against (ooPERa?). Eventually stats like /- should be able to be refined to the point where we can measure this stuff. It took a long time to get to VORP, OPS , and figure out things like pitchers don’t control what happens on balls in play, unless they are knuckleballers, and then they might – in baseball. But this analysis has been to a large extent driven by fans and alt-media, and has changed the way players and rosters are evaluated.

I expect basketball will eventually close the gap.

I realize this is off topic but Led Zeppelin just announced that they will play for the first time in 19 years at a one-time concert in England this November. That announcement, along with the start of a new, promising season in November, this fall something to look forward to.

The important thing to understand about PER, and the reason why it is so valuable, is that it is the only stat in basketball which tracks the game at it’s basic level. It is revolutionary in that it actually tracks the currency of the game, which is the possession. Once you understand that there is nothing in basketball more important than a possession, you can then understand why a stat like PER is not only so important, but so predictive. Of course, it is going to drastically overrate random solid performances by people without enough of a sample size, but for players with a sample size, it is amazingly accurate and properly values the correct things.

The reason why the stat needs to include stuff like blocks, steals, and rebounds is that each of these events leads to a change of possession, and each possession leads to a scoring opportunity. Given that NBA players score on more than 40% of their shots, on average, the game can break down quite mathmatically if you simply generate enough possessions to win on most nights. Good or bad luck will cause outliers, but effective use of a possession (through FG%) and effective possession modifiers (REBs, STL, etc) have a strong and necessary correlation with victory. Any other single stat does not combine this information, so it follows that they do not accurately represent a player’s value to his team.

Stats are great, but they need to be accompanied with context. Stats without context are simply useless numbers. The context is what is important in any argument, and the stats should support the context, not the other way around.

Stats are a great help to differentiate a players performance, but they are a burden when you try to rank it in relation to others, cause you naturally tend to generalize.

Migit has it exactly right. We tend to only view things that reinforce our pre-existing opinions. Stats help us re-evaluate those opinions, based on different metrics. Do I really think Manu Ginobli had a better season then Steve Nash? No. But that doesn’t change the fact it’s a useful metric.

My only real problem with Hollinger is that he knows that PER is an imperfect measurement. His scouting reports are top-notch, and he’s always willing to concede the flaws in PER. But his arguments, at least for ESPN, sometimes rely too heavily on that little number.

Now that I know PER is averaged out over 40 minutes, I think even less of it.

Look, if you have one guy who plays 38 minutes a game and another who plays 11 minutes a game, you CANNOT average them both out to 40 minutes a game. It’s making a statistically invalid assumption. The guy who only plays 11 minutes a game would likely be much less productive per minute with more playing time due to fatigue and wear and tear.

Cousin It – Sure, if a player isn’t playing many minutes, it’s invalid to assume that their production is going to translate out over 40 minutes. It’s all about sample size. That said, someone playing 25 minutes a game who’s putting up great per minute stats probably deserves more playing time. Perfect example of this right now is a guy like David Lee. In years prior, guys like Okur, Redd, and Randolph were poster boys for PER’s usefulness. Fact is per minute stats usually DO transfer over when a player is given more minutes, in fact usually a player’s production per minute increases.

All in all though, PER is a function of usage, and to me there in lies the biggest flaw. Chris Webber can be a terribly inefficient player, but because he uses the ball so much, his PER is still above average. I prefer looking at oRtg, which is a summary of a player’s offensive efficiency, and combining that with usage rate. If I see a player who has an above average usage rate and an above average oRtg, that a pretty effective offensive player. Then there’s your high oRtg, low usage guys like James Posey who are the perfect role players, and the high usage low oRtg guys like Jerry Stackhouse who are useful only on teams that need another high usage guy. For example, Jerry Stackhouse on the Celtics would be absolutely useless, because he’d only take away shots from more efficient scorers.

repole,
Your post is very intelligently written, and after reading it, I’m convinced you know what you’re talking about, but I have to admit … I still don’t fully understand what you’re saying. I get confused by terms like usage rate. I like to watch the game and look at everything a guy does and how consistently he does it. I am unconvinced that numbers on paper will tell me anything about a player that my eyes can’t see.

I don’t like PER. The big problem is by Hollinger’s own admission it’s not particularly accurate. It’s not measuring much of anything except how well people fill out box scores according to the PER statistic.

With normal statistics you can say what they measure. Rebound rate measures how good a rebounder you are and so on. Statistics like PER seem designed to measure how good a PLAYER you are but Hollinger has a cop out and freely admits it doesn’t work all the time.

It’s a statistics that’s supposedly valid when HE wants it to be valid – otherwise you can down play it. That’s not really a very good statistic, IMHO. For example if Eddie House’s PER is 15 and Rondo’s is 12 then can we say House is better then Rondo? Well of course probably not…

And Hollinger admits this. But that begs the question what the heck does it measure then? Can we say someone with TWICE the PER of another guy is a better player? But sometimes Hollinger DOES say a higher PER makes a guy a better player. He famously championed Michael Redd with this statistic..

Hollinger can’t have it both ways like he tries to. The Win Share people say flat out guys with more win shares are better players then guys with less. They at least ATTEMPT to measure something. PER does nothing cept combine box scores into handy little numbers with built in pre selected biases.

It’s not really worth this entire discussion to be honest. /- is a far more useful statistic simply because it brings something interesting to the table. PER does not..it’s a dead been there done that statistic that doesn’t measure much of anything.. It’s a handy box score compiliation but I am just as okay using other ones like efficency or tendex.