without looking at those stats, I know Scola and Landry aren't that great creating their own offense. Landry is worse. Scola have trouble against double teams. So those stats don't tell much. Stats are what it is, when break down, they are just quantified play by play.
If there is nothing observably different about the team's play, then why should anyone attach any importance to his leadership? Suppose you're a coach on the team. Mutombo joins them, and you see him and his interaction with teammates everyday. Suppose you don't observe a single thing he does that has a positive influence on how players perform on the court. He may crack jokes in the locker room and put people in a good mood, but if he's not doing something positive that translates (it can be something as simple as giving advice that guy's observably implement on the floor), then should you attack any importance to it? I'm not saying that I, as a fan with limited information, should be able to quantify it. But if it can be observed, even if its only by someone with the team sees everything, then can't it also be measured?
Everybody has the same stat, which is just game info in raw form. It's how you use that stat that matters, i.e. how to aggregate and how to interpret.
You and I are apparently very different. I like to be able to validate by observations with facts. Who shoots three pointers at the highest percentage with Yao on the floor in the last 10 minutes of games when the score is within 10 points? Also, who's hit the most such three pointers in those situations? The play by play tells you very little that you don't already know, so perhaps you already know the answer to that question. I am more simple-minded. That's something I can't honestly say I know the answer to, despite watching the games, but it's something I'd be interested in knowing.
That's not true. Publicly, the best information we have is the play by plays. That does not record all the relevant information, unfortunately. Teams keep much more detailed records. Even if we have all the games recorded on video, there's stuff we can't know. Like how players perform with different play calls.
Nothing wrong with that. People use stats to re-enforce or justify things all the time. but stats alone can't do the decision making. Off the top of my head, I'd say Ron or AB. The real question is, why do you want to know that? Is that consistent, reliable information so as to dictate how teams play? why last 10? why not say last 4? why within 10? Why 3 pter? See, before you use stat, you need to pick the stat.
True...but you can't remember every play in a play-by-play. Therefore, statistics are a composite of all the information in the play-by-play. I vaguely see your point but note that they are both tied together and both are needed. Something else to add...I hate when people associate stats with the boxscore. There is more to statistics than just the boxscore.
Chemistry could be represented as a stat. Don't get me wrong, these are complex models that we are talking about here. Scientifically, we use stats in more complex scenarios such as cancer diagnosis, and cancer treatments among other things. Basically, anything that causes change can be observed which then could quantified with a statistic. No exceptions! I would say that 99% of the people in the world would not know how to evaluate complex statistics besides the typical stats such an average.
Well, stats are great. In fact, I check them from time to time. I was not writing as an objection to this thread. Whatever, stat you use, if you ever want to go beyond box score, you still need a "model" -- here, it's basketball knowledge and theories. I'm bother only by people who don't know stats using stats, as if throwing out some numbers instantly validates what they have to say.
If you fail to see durvasa's point then you don't understand the power of statistics and probability. Basically, this kind of information can be used to improve the chances of any team in order to win a game. It becomes a game of probability which is inferred from the statistical analysis. Why the last 10? Why 3 pter? He was just pointing out a hypothetical. It just one of many scenarios that can be mapped out. See, before you use stat, you need to pick the stat. See if I was the GM...I would be calculating every stat imaginable so that I can know what are our best chances of winning. What is our best combination of players that will win the game against a team with their combination of players. The opportunities are endless!
I agree with this statement. Here is an example: PLAYER A: Scores an average 22 pts per game. PLAYER B: Scores an average of 18 pts per game. Which player would you rather have on your team? You would be a fool if you didn't pick PLAYER A. Now say the following: PLAYER A: Scores an average 22 pts per game (STDEV ±8 pts per game). PLAYER B: Scores an average of 18 pts per game (STDEV ±1 pt per game). Which player would you rather have on your team? Now the decision is not that obvious. You have PLAYER A that scores on average four points per game more than PLAYER B; however, PLAYER A is inconsistent when compared to PLAYER B. The more information you have and the more information that you know how to use the better decisions you can make for the team. Obviously, the previous scenario is incomplete to make a decision but you can see how statistical information could be useful.
If you think you can you stats without a model, you don't know stats. My questions are directed at what underline model there is in asking that question. Actually, when you watch games, your brain automatically calculate some stats, such as Rafer sucks - it really says you observe he misses his shot a lot. Implicitly, you assume shooting is important in your model (which is a sound one). Then there are other aspects of the game, you don't have rough stats because you don't pay attention to that. Whether you have an accurate number rather than rough estimate is less important than what you look into. In fact, you need to know what you want to look into before you can aggregate your "stats".
Basketball stats analysis is in a promising stage right now. It's a situation where people are trying to find good questions to ask ("what stats are important?"). This is as opposed to data gathering or solving mathematical models. The data is now relatively available, and the field is novel enough that one doesn't need to bring out super-complicated models. Creativity and insight are what's needed.
My work revolves around stats and I have 15 years of education to prove it. My work revolves around engineering statistical signal processing, and random processes. I think that when people disregard stats they are foolish. The only time you would use a model is if you are going to forecast something. I think that you are confusing a model with observations. The first thing that you need to would be to figure out which statistics you want to learn about through observation and measurements (from here you would need to know the sampling as well) from games. The statistics that you obtain from your analysis is nothing more than a structured method of observing without relying on theories about how the game works. A GM would need to know about the game of basketball to make good decisions. A statistician could then rely on models and tell us what might happen when you make a decision. NOTE this is the first time model is used. That is because we are trying to predict an outcome using probability. The statistics are what will give you the justification of confidence on your probability model. So where do the models come from? Observations, analyisis and interpretation. Therefore, there are no models in statistics unless you intend to create an outcome based on those statistics. So basically, going back to our discussion, when durvasa asked is question about 3 pts and 10 minutes left. Those stats could be helpful along with others to dump in a computer to obtain some set outcomes based on a model. So I hope this clarified the matter some.. I disagree...that is not a stat just merely an observation. A stat is a quantity and none was provided. Just because we don't have stats available to us doesn't mean that the Rockets do not analyze them.
I agree with your statements about statistics, but I disagree with the example. If A scores 22 ppg, and B scores 18 ppg, I can't tell you the answer and it would be silly to draw a very strong conclusion. In fact, ppg is an okay but sometimes misleading statistic. What if A's shooting percentage was 35% and B's shooting percentage was 55%? It is efficiency that matters, not ppg (except in extreme cases, of course). Not only that, but what if B's presence on the floor leads to higher efficiency for his teammates? ppg is only one statistic (it is the common fan's statistic) and it often isn't a very good one for determining which player contributes more to winning basketball. See http://myespn.go.com/blogs/truehoop/0-38-284/The-TrueHoop-Network-Shootaround.html
I was just trying to do a simplified example! In fact, if I was looking into both players I would look beyond efficiency. I would look at stats about their style of play: spot-up shooter vs. dribble penetrator vs. off the dribble jump shooter, I would also look at the range: will they clog the paint, jump shooter from 15-18 ft., variety... The opportunity are endless which means that you can make an educated decision if you know what information to look at and how to use it.
your education is shaky Statistics is not just about measures. Without model, you just have numbers. Law of large numbers is a model. That introduced the use of means, percentages. Random walk is a model, why do you calculate diffusion rate otherwise? Even if you want to narrow down the definition of a "model". Before you use model to forecast, you use it to interpret - checking how your theory works, i.e. in sample performance. That's what most people do, using stats to rationalize. In basketball terms, trying to construct a team, to win a game, is forecasting. Why on earth you calculate a bunch of numbers otherwise? Hypothetically, I can ask: who scores the most in the 4th position of 3rd quarter. But I fail to see why that question is relevant. Oh, that's a rough stat alright, because "poorly" means 1. it measures aggregate 2. it has quantitative implications. An observation would be: Rafer missed a shot.
Ha Ha... Well try this link. All definition of statistics are about collection of data or measurements and interpretation. There is no mention of the word "model". Furthermore, the law of large numbers is a theorem and not a model. A theorem means that it is a proven concept and nothing more. Sure in statistics you have to select the correct statistical measures and techniques to make good decisions but I would call that a statistical system and would not call that a model. By the way there are two types of statistics: descriptive and inferential. Descriptive statistics in the basketball sense would tell us something useful about what is going on the data of a team or a player. Inferential statistics allows us to know things about thing we can't measure or count easily. Inferential statistics could tell us about what is likely to happened if a particular decision is made. I somewhat agree with this statement as it is a little more complex than this. But that is a start. Well statistical analysis could be used to measure performance. Take this scenario: A GM had a wealth of information on the performance of all NBA players. Trade deadline is here and you have the option to trade on of your players to three different teams. Based on observations you could choose the best package but you could use statistics to supplement your knowledge. In this case, you have "calculated a bunch of numbers" and not for the purpose of forecasting. Mostly any GM could do this... The other extreme would be to use these numbers to forecast and try to improve the team in game time decisions. I bet you that not many GMs do this and I do agree that it is a growing field. Darryl Morey is probably leading the front in this new era of statistical analysis that include both descriptive and inferential statistics that can be used to construct a model that can be used to obtain probabilities. Finally, you would close the loop by comparing the probabilities to the statistics to see if it makes sense. PS. I got to get and get ready for my lecture on "Random Walks"
