OT: Stats indicate Sergio actually good

[Ed O]

Translation: My eyes do not agree with the statistics, I therefore declare them meaningless...

The team won at a might higher clip when Jones was on the floor than when he was not. Clearly he was not important to their success

Your logic is terrible. I'm not saying that the stat disproves his value... just that it's not relevant.

The team played well overall, and that's fantastic. It wasn't because of Jones, as far as I can tell, and some watered-down stat with a few dozen games' worth of data isn't going to convince me otherwise. Remember that the team had TEN home games during that stretch and played a pretty easy schedule during that time overall. Since Jones played a disproportionate number of his games (he missed more than any regular other than Raef) his win rate looks better. I don't see causation in the stat you mention, and without causation I don't see value.

I can make up a stat, have it "prove" something, and then make fun of you for not believing in it, also. It's an easy thing to do.

Ed O.

 

[andalusian]

I can make up a stat, have it "prove" something, and then make fun of you for not believing in it, also. It's an easy thing to do.

I did not make up the stat. It is recorded by a pretty respected site (82games.com).

I just love it how every time there is such a big bruhaha against a stat that directly measures a team's effectiveness in doing what matters most (win) when a player is on the court, and how 58 games (2 thirds of a season, practically) is irrelevant (especially when talking about that very same season).

We have had these kind of discussions before where you make fun of a "small sample size" only to offer "no sample size" to counter the argument (See Cuningham, a PF or SF in the NBA, earlier this season).

Well, I am delighted to know that you do not consider James as a big reason this team was successful on that specific season despite the fact that the small sample size (around 70% of that season) seems to support it.

I guess I am saying that I find my "terrible logic" better than the "no logic" you offered (It is so because I do not believe it).

 

[LittleAlex]

I did not make up the stat. It is recorded by a pretty respected site (82games.com).

I just love it how every time there is such a big bruhaha against a stat that directly measures a team's effectiveness in doing what matters most (win) when a player is on the court, and how 58 games (2 thirds of a season, practically) is irrelevant (especially when talking about that very same season).

We have had these kind of discussions before where you make fun of a "small sample size" only to offer "no sample size" to counter the argument (See Cuningham, a PF or SF in the NBA, earlier this season).

Well, I am delighted to know that you do not consider James as a big reason this team was successful on that specific season despite the fact that the small sample size (around 70% of that season) seems to support it.

I guess I am saying that I find my "terrible logic" better than the "no logic" you offered (It is so because I do not believe it).

If your logic is based on something that isn't actually a fact then it isn't really logic.

What you are using as proof is a complex formula that tells us that James Jones was probably involved in that Portland win streak at a higher level then most all of the other players on the team. This is far, far different then saying James Jones was absolutely more responsible for all of the wins during that win streak then all but one other player on the team.

Sample size is irrelevent. You could have a sample size of a billion games and it still wouldn't be a undeniable fact. It is entirely possible that James Jones's contribution had little to nothing to do with Portland's success. It isn't as likely a result, of course, but well within the realm of possibility. It is even more possible that his contribution was high, but not as high as this number would indicate.

This is my problem with Wayne Winston and guys like him. He states something that is highly likely as something that is certain.

These numbers are useful evaluation tool but they are nothing more then that. They actually prove nothing.

 

[rocketeer]

Sample size is irrelevent.

sample size is incredibly relevant.

 

[andalusian]

The only reason to use statistics is to measure the things we can not measure directly. If we have a formula that shows us that something happens or does not happen - with no use of statistics - we will use it and know EXACTLY what the outcome is. If we do not have this - we use statistics and try to get a "confidence" level in the result.

Statistics are used to try and give some "mathematical" numbers behind something we can not measure directly - and give you a "confidence measure" of how sure you are about these numbers or not, given some margin of error.

If we take one season and the "number of games all players played" as our population, and look at Jones's number of games played as a sample size for that season - I get (assuming a binomial distribution with 1 = win, 0 = loss) a confidence level of 88-89% that his win% is legit (margin of error - 10%). Would have been nice to have more games from him - but statistically, the numbers are not too bad.

Is there a chance that this stat measures something wrong? Of course, that's what statistics are all about - so - I am pretty confident in saying that the numbers are very likely to have been relevant that his contributions that year were pretty significant. Of course, part of it, in a basketball game is not directly from your own contributions - but from who did not play when you were active - so - maybe it is not so much that got to play James as much as "we had someone better than Webster to play for a lot of minutes". That's a possibility as well - but at the end of the day - it comes down to the same thing - being able to play James, as opposed to the alternatives available to the Blazers that year - are very likely to have contributed to the team's success that year. Does not make James that great of a player individually, but in the context of the Blazers - his contributions were rather relevant.

The facts are (these are not statistics) -

The Blazers won a .58 clip in games James played, and .33 clip in games he did not. His Win% is a fact (not a statistic). What the statistic can tell us is how relevant this win% is given the number of games he played, that his presence that year would have contributed to the same win clip given that all other variables are the same.

His Win%, given the sample size and the population size do not question the result as something with very little confidence or a very large margin of error, so, statistically - I am not really worried about using it as a relevant data point that is worth of discussion. 58 Games is not a lot, and you would have liked to get more to bring either the confidence level up or the margin of error down - but that's what we have.

Anyone that wants to ignore it - is welcome to it - but until I see a better measure out there that proves it irrelevant or shows the opposite... I am going to continue and be a skeptic of people who tell me that the "just do not believe in it" without giving me any other measure.

 

[C_note]

People spend about two weeks of their lives at traffic lights

LOL, my brain went...uhh ok...well...thanks for the useless informa....ooooooh i see what u did there!!

 

[Ed O]

The facts are (these are not statistics) -

The Blazers won a .58 clip in games James played, and .33 clip in games he did not. His Win% is a fact (not a statistic). What the statistic can tell us is how relevant this win% is given the number of games he played, that his presence that year would have contributed to the same win clip given that all other variables are the same.

His Win%, given the sample size and the population size do not question the result as something with very little confidence or a very large margin of error, so, statistically - I am not really worried about using it as a relevant data point that is worth of discussion. 58 Games is not a lot, and you would have liked to get more to bring either the confidence level up or the margin of error down - but that's what we have.

Did you normalize for the strength of schedule? Based on opponents and home-away splits?

I don't believe that the stat does that. Which is one of the many reasons that it's pretty much worthless when we're discussing a guy who played fewer than 1300 minutes.

Ed O.

 

[MARIS61]

Sergio lead his new team to an OT win over the whiz tonight, with another solid stat line.