How consistent are our stars: a statistical analysis

Even the table got screwed...


I understand a lot can be attributed to the way the intervals are defined (I thought that 5 point buckets would be significant enough and summarizes the data well). I didn't spent the time to define the best intervals mathematically, too time constrained and lazy...


but I believe this best describes what the common perception sees.

 

Probably the graph shows a better trend. From your table, it
seems that Yao's really not that inconsistent. 53 games out of
71 Yao scores in the range of 15-25, while Tmac's scores range
from 20 - 35. Tmac will always get his 20+ because of the sheer
number of shots he takes. Shooting-wise, he's very inconsistent
compared to Yao. It's not hard to list nights when he shoots
below 33%.

By the way, I don't think the two peaks in Yao's scoring mean
inconsistency. We would be ecstatistic about the 25 pts peak.
The 15 pts peak are just below his average, which is as expected.
If he peaks at 10 pts and 25 pts, then that would be inconsistent.

 

Common perception could be wrong when the observers are
biased. You could see a lot of rants about Yao are not justified,
so were those about Francis last year.

Yao's shooting is very consistent and rarely has a FG% below
45%. If we could get him 20 shots per game, then I bet most
games he would score > 20. Tmac has some horrible shooting
nights but still puts up 20+ pts on 25 shots.

 

There is chance they score under that in all those 1000 games.

 

You are arguing for the sake of argument. His study was about PAST stats, which gives you some indication for the future, MAYBE. But for the past your argument is simly FALSE, because they didn't score less than that for the past N games, therefore, they won't do that in their total games they play in careers.

 

The one-game probability would be raised to the power of 1000
in that case, which makes it 10000000000000 times rarer than
Tmac's 13 pts in 33 secs.

 

Of course I am talking about the future , that is the whole point of probability right? I was simply pointing out the "AT MOST " part is wrong.

 

Who knows, we dont have enough sample space to calc the P of that TMAC thingy.

 

Please read his post before you reply. He said if they only play 1000 games in their careers, then ..... Their careers include PAST games and FUTURE games. If they had only one game where they scored more than that in the past, which they did. It is NOT possible that they score less than that in those 1000 games in their WHOLE careers. Got it? You were making an argument with him, but he's talking about past stats, and you are "OF COURSE TALKING ABOUT THE FUTURE", how's that a valid argument? Just count how many logical flaws and errors in your posts.

 

[delete]

 

Sorry for not being exact , it was a casual one liner. Let me put it this way. If TMac and Yao were to play 1000 together in their career , there is chance that the total point scored by them is lower that certain number in N games , where N = 1000- the number of games they already played.

 

So I see a lot of you guys trying to draw some clusion (or show off) on the CI of the toal score of both based on individual stats. And then arguing about if it’s co-re or not.

Why don’t you add the scores from the two player in each invidual game and analyze, then you can forget about co-re and gets what you need.

 

haven't finished reading the whole thread yet, but here are some of my comments:

1. please note a 95% confidence of 10-35 pts for tmac and 10-20 pts for yao doesn't bring a 95% confidence of combined 20-55 pts. instead, the confidence level would be a LOT higher, something like 99% (I don't have excel to do the calculations now. but if you can, MFW please use the original data, calculate the combined points for each game, and calculate the 95% confidence level, you should see what I mean)

2. the definition of "inconsistency": I feel we have to define it in a bit different way than normal in order for it to make sense. My proposed definition:

when the player performs lower than the MINIMUN of our expected production, that's when he hurts the team and that's when we call him inconsistent. -- for example, personally I'll call Yao inconsistent if he scores lower than 14 pts or less than 7 rebounds. if he plays 14/7 in a game, I won't be happy with it, but I can still take it. Then we can calculate the % of games in which Yao doesn't meet the expectation. Similarly, I would consider Tmac to be hurting the team if he gets less than 20 pts, or 4 rbs, or 3 asts. Then we can calculate the % of games in which He doesn't meet the expectation.

-- of course, the problem here is that expectation is very subjective and everyone would have his own standards. But to talk about "inconsistency" in the normal everyday meaning just doesn't make basketball sense, IMO.


all that being said, good work man... I love your approach to the numbers. please keep up with the good discussion everyone

 

I think FG%, #FTs, TOs, etc. should also be factored in.
If Tmac gets 25 pts on 10-30 shooting, it is a disappointing
performance.

 

awesome calculations. that's what I was trying to say but can't really do the calculations properly to show it.

The exact factors we need to consider can be up to debate. I just hope a large percentage of us can agree that's a sensible definition.

-- and actually, would anyone be interested to start a poll about what's the minimun contribution of Tmac & Yao should produce from them to be considered inconsistent? I would be really interested to see what's the board's perception to mine.

 

The point of the thread was to have a more objective discussion about this. It got lost in a bit of the statistical "noise" . Glad to see we are back on topic.

 

If you are going to talk about inconsistency vs. consistency - let's at least do things right.

And to me, the definition of consistency isn't scoring 50 one night and then 20 the next. That's not inconsistant in the basketball sense.

Inconsistent is the number of crappy performances amongst the decent ones and the great ones.

So a player that averages 18 & 8...we'd first have to agree what that would be. It's not some standard deviation or 2 or whatever. C'mon, we know what a crappy performance is. You look back at Yao's game and say - boy that was crappy, ok, that was at least decent.

This makes your analysis much cleaner - since it's now binary and you are looking at the totallity of their game that night and not one statistical category.

 

can I say we are talking about the same thing while my sucky english prevents me from explain as well as you do, NYC?

 

I read your post and yeah, what you say is pretty much along the same lines.

At the end of the day, we need a definition off a bad performance and an acceptable one. And some of that goes beyond numbers since there are variables - games where a player got hurt, or was pulled because the game turned into garbage time. Also, we need to see the pattern. Do we get 0-1-0-1-0, or does it go 00010100101111011111110011111111 since that demonstrates more "consistency". This is actually a really complex question to quantify.

I still think without a doubt T-mac is more consistent then Yao, which you can figure out by just watching games. Yao varies between tough and soft...and that's clearly inconsistent.