Asunto: www.sktables.org - Enhance Your Sokker Experience!

I'm not talking about potential randoms - I tend to agree with you - but about certifying talent.
It does seem a little “presumptuous” to grant such certification when the talent can just as easily be very good as bad.
Iit might maybe make more sense to grant the certification only in cases where the talent is truly determined with certainty by sktable.
This would certainly “sacrifice” the juniors staying too few weeks in the center, but it would also perhaps give more credit to this certification.
At the moment, I notice a little too much variance.
(editado)

I ve already answered you. This junior will certainly randoms up and training up in 3 weeks max, it s most predictable.

You still feint to not understand, or at best you are really strange to use a tool that you don’t trust.
(editado)

I was tracking a certified divine youth pull, which then dropped.

I've already answered previously (before garbages)

In short, if a junior loose certification during time :
1) talent should not shift much and certification will be back again.
2) talent change, and previous certification is lost. Seems to be fair.

"Certified" showcase talent with higher prediction, but it doesn't mean certified is 100% true (errors always remains) or "uncertified" are 100 % false.

If we take this example : Talent: 3~3.3 ✓
It's start at level 4 end 7 after 9 weeks, and talent is certified. Accuracy is 86% (not 100%) and errors 33% (not 0%). So why it is certified? if the player up or down +2 -2 in the next week, it will not change talent, even if certification is lost.
(editado)

I'm not saying this junior isn't talented, but his entry level could have been 6 if the coach had given the wrong estimate.
Which could greatly diminish his talent if he went out at 7 something....
(editado)

But it wasn't 6. That's how statistic tools work, 4 is far more predictable.

Not only you don't care how stats tools work, you just don't care.
So please, Don't use a tool you can trust, and get out of here.
(editado)

Can't we have a normal discussion?

It is nice to try to impress others with statistic knowledge but I am not sure you correctly translated these statistics into the sokker ecosystem.
We are talking about a junior who stayed 8 weeks in the YS, entering with a level 4 that could very well have been 6 with coach wrong estimate, and who exited with a level 7.
Without talking about the very weird +4 pop from 5 to 9....
And still you can ensure that his talent is between 3-3.3
(editado)

4 5 6 5 9 7 6 7 7 give a talent around 3 to 3.2 with a correlation of 0.86
6 5 6 5 9 7 6 7 7 give a talent talent 4.3 and more with a correlation of 0.67

We can add one more random +2 -2 up down levels whatever...

4 5 6 5 9 7 6 7 7 9 give a talent 3 to 3.3 with a correlation of 0.86
6 5 6 5 9 7 6 7 7 5 give a talent 4.3 and more with a correlation of 0.67
(editado)

What is that "correlation"? How does this (theoretical :P) chart also have 0.86 "correlation"?

I see a lot of self-assurance in what you are saying but if you took a step back from your books you'd realize that you are missing a point here.

The 4 ups from 5 to 9 means that on week 3 the coach probably made an estimation error of -2 and on week 4 an estimation error of +2.
(The owner has an unearthly junior coach so the estimation variance would be at worst -2 / +2)

So the real level of this junior was probably around 7 on weeks 3 and 4.
I'm sorry, it's a reason to think that we can't say with certainty as you do that his talent is between 3 and 3.3.
(editado)

Is it a real junior progression ? WTF ?

Certified stand with errors, in that case he is of course not at 79% ^^

No it's not real, I just added bunch of 5 at the end of that junior chart.

I just ask what the correlation is, how you calculate it.

The most disturbing is to see that with a progression of 4 5 6 5 9 7 6 7 7 the chart shows an accuracy of 86% with a certified talent of 3-3.3:

[url=https://postimg.cc/FfFjvSW3][/url]

And with a progression of 5 5 6 5 9 7 6 7 7 (one more level on week 0) the chart shows an accuracy of 19% with a talent from 3 to 5+...

[url=https://postimg.cc/KKcWRJvY][/url]
(editado)

The algorithm is based on “brute force” (all possible combinations of talents/starting levels) and certain constraints. Correlation is the number of talents that match/total talents studied. The Errors are the number of weeks/total number of weeks where the level and the predicted level do not match.

It's broken when talent reach 5 to infinite, since 1) i dont care about talent above 5 2) i have to save brute force time machine, 3) those "dead juniors" are not usual.

Except some limits and talent 5 to infinite, the result is usually close to a linear regression. You can check.

And thanks for your help, you just give another bone to a master troll :-p

Its exactly the same question as before, the answer is the same : '4' is the starting level according to sokker and a better combinaison is found with this series.

You can doubt that 4 is true (coach errors), but nothing will help you to know if any other starting levels are more probable or not.
(editado)

You talk about combination, series, statistics, but have you ever tried to confront your theories and algorithms ?

Seriously, how can we trust your tool when adding an extra level to this junior on week 0 gives us so much discrepancies in % accuracy and talent estimation (86% & 3-3.3 with certified talend vs 19% & 3-5+) ? We are not talking here about marginal but very high variation.

And don't you see that 4 ups from 5 to 9 means that the real level of this junior was probably around 7 on week 4 ?
Which means that the certified talent of 3-3.3 is highly questionable. And if that's so questionnable then the so-called certification is irrelevant.
(editado)