Author Topic: Stats. What are they good for?  (Read 15615 times)

0 Members and 1 Guest are viewing this topic.

Re: Stats. What are they good for?
« Reply #30 on: March 05, 2013, 12:18:48 AM »

Offline guava_wrench

  • Satch Sanders
  • *********
  • Posts: 9931
  • Tommy Points: 777
Absolutely nothing.

Show me a basketball stat and I show you itīs fallacy.

The idea that we can use abstract placeholders to describe dynamic processes is theoretically feasible, but practically useless.

Games arenīt played in a vacuum.
How do you quantify experience, chemistry, KGīs will to win or the effect on your motivation a series of wins or losses have?

The only reason stats have entered sports discussions is to have a "killer argument" if you want to justify why Player A is better or worse than Player B.

So, basically, itīs just media talk.
One result of the general ignorance of most people when it comes to stats is the all or nothing attitude. It is hard to convince the stats illiterate of the value of stats when they don't understand how to think mathematically.

For example, it would be complete nonsense to say that if we embrace stats, then we cannot consider chemistry important. That is not how stats are used. No statistician claims that any combination of stats can account for 100% of variation in performance. When they can claim though is that stats can help us find value that is often missed when using our clearly flawed and biased perceptions. They help us make the best choice amidst the always present uncertainty. If a GM can use stats to make 10% better decisions, that will add up over time to serious gain.

So the fallacy is not in the stat. It is in the idea that stats are all or nothing. Understanding the implications of stats help you make better decisions because our intuitions are incredibly subjective and often far from reality. This is obvious is we read this forum or watch sports tv and see intelligent fans arguing for opposite positions. The reason for the disagreement is typically not that either fan is a fool. It is typically due to the reality that sports judgments are highly subjective and even what we are talking about is often very imprecise.

Re: Stats. What are they good for?
« Reply #31 on: March 05, 2013, 12:19:07 AM »

Offline indeedproceed

  • In The Rafters
  • James Naismith
  • *********************************
  • Posts: 42583
  • Tommy Points: 2756
  • You ain't the boss of the freakin' bedclothes.
Best stats I know of: Ts%, total Rebound rate
TS% is an excellent measure of how efficient a player is when scoring. It factors in FT's, added value of 3pters, and horoscope sign.

Total Rebound rate (and defensive rebound rate, offensive rebound rate) is the best measure of how effective a rebounder a player is. Rebounds per 36 minutes is also a good measure, but rebound rate is more comprehensive.

Opponent Production per 48 (from 82 games): only really useful when a large sample size is present, and isn't by any means an end all description of how good a defender one is. It does however provide some light. It can show how well a combo guard defends 2's vs 1's, for example.

Positional offensive production per 48: shows how well a player scores when they're often playing multiple positions in a season. Great example Thad Young. Highlights his inefficiency as a 3, and his mismatch as a 4.

Points/rebounds per 36 minutes: a good number to use when looking at a contextual contrast of a bench player vs a starter on equal footing. Not a great stat because almost always per minute production drops the more minutes per game a guy plays, but useful nonetheless.

"You've gotta respect a 15-percent 3-point shooter. A guy
like that is always lethal." - Evan 'The God' Turner

Re: Stats. What are they good for?
« Reply #32 on: March 05, 2013, 12:22:40 AM »

Offline guava_wrench

  • Satch Sanders
  • *********
  • Posts: 9931
  • Tommy Points: 777
they are interesting as hell. especially the way they're done today. some of the "less analytical" folk may disagree

Thatīs funny.

I thought stats are for the "less analytical" folk.
The reason why the stat haters are typically the less analytical people (we should never generalize to all) is that they are typically the people who would rather rely on their perceptions. They are unwilling to challenge their own perceptions with objective data such at metrics.

Still, someone can be very analytical but still statistically illiterate. There are many ways to be analytical, though some may be very unreliable.

Re: Stats. What are they good for?
« Reply #33 on: March 05, 2013, 12:23:30 AM »

Offline GreenEnvy

  • Antoine Walker
  • ****
  • Posts: 4551
  • Tommy Points: 1031
Anybody that doesn't appreciate the value of stats is [going to avoid making any personal attacks but it's such an ignorant viewpoint I feel like I need to]

Just looking at a basic box score, you can get a great feel for who the best players in the game are. In fact, I'd be willing to wager that a basic box score is far more accurate than anybody using the eye test.

Now ideally, you want to use objective and subjective analysis. But anybody ignoring stats out of fear or ignorance is missing out on a lot of knowledge about the game.

I disagree completely.

But I think you could climb the corporate ladder in Bristol with that mindset. I think ESPN frowns upon their "analysts" actually watching games.
CELTICS 2024

Re: Stats. What are they good for?
« Reply #34 on: March 05, 2013, 12:29:05 AM »

Offline fairweatherfan

  • Johnny Most
  • ********************
  • Posts: 20738
  • Tommy Points: 2365
  • Be the posts you wish to see in the world.
Absolutely nothing.

Show me a basketball stat and I show you itīs fallacy.

The idea that we can use abstract placeholders to describe dynamic processes is theoretically feasible, but practically useless.

Games arenīt played in a vacuum.
How do you quantify experience, chemistry, KGīs will to win or the effect on your motivation a series of wins or losses have?

The only reason stats have entered basketball discussions is to have a "killer argument" if you want to justify why Player A is better or worse than Player B.

So, basically, itīs just media talk.

Literally every front office in the NBA disagrees with you. 

We don't consider "the eye test" or "common sense" or whatever people call the alternatives to stats useless if they can't explain 100% of what happens in a game, so why are only stats held to that standard? 

Statistical analysis isn't useful because it can perfectly quantify every aspect of a game, it's useful because it usually provides better explanations than subjective methods.  Which is why every front office uses it, and many are spending a lot of money and effort trying to push the analytic envelope to gain an edge.

Re: Stats. What are they good for?
« Reply #35 on: March 05, 2013, 12:35:25 AM »

Offline guava_wrench

  • Satch Sanders
  • *********
  • Posts: 9931
  • Tommy Points: 777
Anybody that doesn't appreciate the value of stats is [going to avoid making any personal attacks but it's such an ignorant viewpoint I feel like I need to]

Just looking at a basic box score, you can get a great feel for who the best players in the game are. In fact, I'd be willing to wager that a basic box score is far more accurate than anybody using the eye test.

Now ideally, you want to use objective and subjective analysis. But anybody ignoring stats out of fear or ignorance is missing out on a lot of knowledge about the game.

I disagree completely.

But I think you could climb the corporate ladder in Bristol with that mindset. I think ESPN frowns upon their "analysts" actually watching games.
This is a great example of why people need stats and not just subjective opinions. Your ESPN comments are demonstrably way off-base. ESPN has tons of analysts that don't use much in the way of stats.

We needs stats to control for the fact that people see what they expect or want to see. We are all prone to confirmation bias. Beyond that, we also all are very limited in what we can perceive and very flawed in how we perceive it. Instead of boldly making statements like the one you made about ESPN analysts, the more thoughtful commenter would generate a list of analysts and count the stat-geeks among them.

What gets ratings on TV is loud, opinionated, over-confident analysts, not stat-geeks. So the type of people you seem to prefer are quite well represented.

Re: Stats. What are they good for?
« Reply #36 on: March 05, 2013, 06:26:22 AM »

Offline Casperian

  • Ray Allen
  • ***
  • Posts: 3501
  • Tommy Points: 545
I like how you guys put words into my mouth. Iīm not ignorant of stats, Iīm over them.

Quote
No statistician claims that any combination of stats can account for 100% of variation in performance.

So, the placeholders donīt represent constant values. 1 isnīt actually 1, at least not always. You understand that this little sign (=) becomes a lie, right? All laws derivating from math involving them arenīt really laws, theyīre more like general rules of thumb, yes?

Quote
When they can claim though is that stats can help us find value that is often missed when using our clearly flawed and biased perceptions. They help us make the best choice amidst the always present uncertainty.

Youīre making stuff up now, arenīt you? Which "clearly flawed perception" are you talking about? Stats are just a bunch of numbers without proper interpretation, just like the action on the court has to be interpreted. Considering how I read deeply flawed application of stats on this board alone on a daily basis, I have a hard time accepting their, at best, "general rule of thumb" as the incorruptible judge in any debate.

So, first of all, I question the neutrality of numbers interpreted by humans, and thus, all laws derivating from them. But you understand that, basically, as you wrote

Quote
So the fallacy is not in the stat. It is in the idea that stats are all or nothing. Understanding the implications of stats help you make better decisions because our intuitions are incredibly subjective and often far from reality. This is obvious is we read this forum or watch sports tv and see intelligent fans arguing for opposite positions.

I challenge this view, too. The fallacy is also in the stat alone. The matter youīre trying to explain with not 100% exact placeholders in a closed system is a dynamic one in an open system, in other words, their applicability is limited per se.

Only if the framing conditions stay exactly the same (which they never do), and the stat is trying to explain a simple cause and effect (which every 10 year old could do with words), and 1 is actually always 1, then they tell you something definitive about a limited aspect of the game, which then has to be interpreted and put in itīs proper context by humans with their, as you called it, "clearly flawed perception".

An assist isnīt simply an assist, there are a myriad of angles on the court from which the ball could go from point A to B, and the distance between the assists man, the receiver and the opponent(s) is never the same, but paramount to contextualize the assist. Just this one, lousy assist. To cover this aspect alone, youīd need 3D graphics instead of simple numbers.

Theoretically feasible, pratically useless.

Considering that every NBA team employs professional statisticians, and there are still countless braindead decisions made in the NBA every year, Iīd say the advantage of relying on stats over common sense still has to be proven.

Quote
We don't consider "the eye test" or "common sense" or whatever people call the alternatives to stats useless if they can't explain 100% of what happens in a game, so why are only stats held to that standard?
 

Because if theyīre not even saying what they are supposed to say, then theyīre simply misleading, and have the same meaningfulness as your daily horoscope. These NBA teams might just as well employ a bunch of priests and rabbis to give them advice.

Consequently, and maybe even more importantly, considering their arbitrary nature, their only purpose is to justify whatever anyone wants to justify, and are only distracting from the actual debate, often to the point that itīs impossible to carry on with the discussion without addressing them first in long, drawn-out and boring math exercises.

So, to make an already long post shorter:

I issue my challenge again. Show me a stat, tell me what it is supposed to say, and I will explain itīs fallacy.
In the summer of 2017, I predicted this team would not win a championship for the next 10 years.

3 down, 7 to go.

Re: Stats. What are they good for?
« Reply #37 on: March 05, 2013, 07:21:33 AM »

Offline CelticConcourse

  • Paul Silas
  • ******
  • Posts: 6162
  • Tommy Points: 383
  • Jeff Green
This is why scouts watch players at live games, not GameCasts from home!
Jeff Green - Top 5 SF

[Kevin Garnett]
"I've always said J. Green is going to be one of the best players to ever play this game"

Re: Stats. What are they good for?
« Reply #38 on: March 05, 2013, 07:56:32 AM »

Offline BballTim

  • Dave Cowens
  • ***********************
  • Posts: 23724
  • Tommy Points: 1123
If pierce is averaging 55 points a game but the celtics are losing then it means nothing and its obviously not working.

  I disagree with this. PP averaging 55 a game might be having a strong positive impact on the game, Antoine going 6-25 from behind the arc might be the reason you're losing those games. Trying to figure out whether we need PP to take fewer shots or keep Toine from taking all those threes is valuable, stats can help with that analysis.

  I also disagree with the thought that making s positive impact on a game is only worthwhile if your team ultimately wins the game but that's probably for another thread.

Re: Stats. What are they good for?
« Reply #39 on: March 05, 2013, 08:30:57 AM »

Offline Fafnir

  • Bill Russell
  • ******************************
  • Posts: 30859
  • Tommy Points: 1327
Casperian you dismiss stats because they're "fallacies" because they aren't some sort of higher truth? That's ridiculous, they're just pieces of evidence you use to examine the game. You don't throw away evidence just because its not complete telling of the entire game that happens on the court.

Advanced statistics are useful because human beings are crappy observers of reality. We make bad eye witnesses and are bad at considering the totality of performance, we focus on what we've just seen or what excites us.

Your attitude would have us never investigate physical mechanics or thermodynamics with models because they don't 100% describe the reality of what's happening in the physical world.

Re: Stats. What are they good for?
« Reply #40 on: March 05, 2013, 08:42:27 AM »

Offline Lightskinsmurf

  • NCE
  • Don Chaney
  • *
  • Posts: 1949
  • Tommy Points: 134
If pierce is averaging 55 points a game but the celtics are losing then it means nothing and its obviously not working.

  I disagree with this. PP averaging 55 a game might be having a strong positive impact on the game, Antoine going 6-25 from behind the arc might be the reason you're losing those games. Trying to figure out whether we need PP to take fewer shots or keep Toine from taking all those threes is valuable, stats can help with that analysis.

  I also disagree with the thought that making s positive impact on a game is only worthwhile if your team ultimately wins the game but that's probably for another thread.

I predicted this would be somebody's response the second I posted this.

Re: Stats. What are they good for?
« Reply #41 on: March 05, 2013, 08:46:25 AM »

Offline BballTim

  • Dave Cowens
  • ***********************
  • Posts: 23724
  • Tommy Points: 1123
Absolutely nothing.

Show me a basketball stat and I show you itīs fallacy.

The idea that we can use abstract placeholders to describe dynamic processes is theoretically feasible, but practically useless.

Games arenīt played in a vacuum.
How do you quantify experience, chemistry, KGīs will to win or the effect on your motivation a series of wins or losses have?

The only reason stats have entered basketball discussions is to have a "killer argument" if you want to justify why Player A is better or worse than Player B.

So, basically, itīs just media talk.

  I agree with your statements about how stats can't definitively measure any aspect of a basketball game (although some simpler stats measured over time are fairly accurate) but I don't think that makes them useless. All stats have limitations and only measure what they measure. Whether they indicate other things is a bit more nebulous.

  I personally think that understanding stats can improve your understanding of the game. Frequently I'll see stats that don't seem to jibe with what I see on the court. I'll try and figure out why that is. It generally improves my understanding of what stats do and don't measure and what their limitations are. You shouldn't use stats to form your opinions of players or teams but if my opinion of a player is significantly different from what the stats show I try and figure out why the stats don't reflect what I see. For instance Rondo and PP having the same fg% from a given range would IMO be related to Paul being blanketed by defenders, Rondo having a higher TO% than many point guards would be due to their shooting the ball more often than Rondo and passing less often.

  I will say that the only stats that I generally rely on over my observations are offensive and defensive efficiency for teams. I think that good or bad stretches influences your overall opinion and that I don't spend enough time watching other teams to decide how good they are overall and how they compare to each other. Obviously I can tell top teams from bottom teams but it's hard to see the difference between the 9th team and the 20th if you only see those teams a couple of times a year. Not to mention that a team might just be a good or bad matchup for the Celts. For instance I'm shocked every time I see teh Wizards are among the top defensive teams, not so shocked that I'd claim that they're really as bad as teams that have been giving up 7-8 more ppp than the Wizards for 2/3 of a season.

Re: Stats. What are they good for?
« Reply #42 on: March 05, 2013, 08:51:53 AM »

Offline BballTim

  • Dave Cowens
  • ***********************
  • Posts: 23724
  • Tommy Points: 1123
If pierce is averaging 55 points a game but the celtics are losing then it means nothing and its obviously not working.

  I disagree with this. PP averaging 55 a game might be having a strong positive impact on the game, Antoine going 6-25 from behind the arc might be the reason you're losing those games. Trying to figure out whether we need PP to take fewer shots or keep Toine from taking all those threes is valuable, stats can help with that analysis.

  I also disagree with the thought that making s positive impact on a game is only worthwhile if your team ultimately wins the game but that's probably for another thread.

I predicted this would be somebody's response the second I posted this.

  I can frequently predict the kind of responses my posts will generate as well. I just think there's a certain amount of "throw the baby out with the bath water" element to your statement.

Re: Stats. What are they good for?
« Reply #43 on: March 05, 2013, 08:56:48 AM »

Offline Celtics17

  • Jayson Tatum
  • Posts: 873
  • Tommy Points: 108
Statistics can be very accurate and almost to the point of being unbelievable. They can also be highly misleading. Approxiamately 40 years ago the U.S. and the Soviet Union competed in a track meet with only the two counties in the meet. The U.S. rolled over the Soviets but the headline in the Soviet paper the next day read "Soviet Union takes 2nd place at world track meet, U.S. takes second to last". Well this was somewhat accurate but still very misleading.

One other I heard was this- If you have your head in a freezer and your behind in an oven, on average, you are doing pretty well. Again, very misleading.

Re: Stats. What are they good for?
« Reply #44 on: March 05, 2013, 09:29:42 AM »

Offline BballTim

  • Dave Cowens
  • ***********************
  • Posts: 23724
  • Tommy Points: 1123
Statistics can be very accurate and almost to the point of being unbelievable. They can also be highly misleading.

  I don't like the term misleading, I prefer misused. One places the blame on the stat, the other (rightfully) places the blame on the person using the stats to show what they weren't meant to measure.