Asunto: NEWS Sokker- big changes are coming!

I also expect that in the subgroup of these youths with top talent and 30+ weeks in ys there will also be 80%+ with initial 0.x gk skill.

You can expect, but nothing tell you that this subgroup is not part of 20% starting at least 1 gk... It's a wrong data analysis if you think that there'r 80% of probability for each subgroups.

If your theory was true then it would completely break this correlation, because initial skills would have much bigger impact on final skills.
For example someone starting at level [2] and finishing at level [12] would need to have about ~half of the final sumskill of someone starting at level [4] and finishing at [14]

Well, bring back to this conversation...

Level/sumskills progression is relative to talent (in my expected model), you compare 12/2 14/4 but it's not the same ratio (or talent). Levels progression are not relative to level itself, i didnt catch your point.

If i take your words :

level 2 * 3.2 = 6.4 sumskills
level 12 * 3.2 = 38,4 sumskills

level 4 * 3.2 = 12,8 sumskills
level 14 * 3.2 = 44,8 sum-skills

Even if level 2 is twice of level 4 sumkills, ending levels 12/14 are not "half of...".
Not such a big diff in the end.

My theory is not against linear progression — initial skills/levels/sumksills growing at same factor — it could even match your ~3.2 skills, so it's not typically against your own correlation. For sure starting skills have impact, but not as much as you seems to say in term of absolute skills.
(editado)

I remember getting brilliant junior with about 38 sumskill.

I confirm :-/

But 19 years ago, sokker coaches mistaskes didn't exist and even if the sumskills didn't match, we could see a linear progression without a doubt.

I dont understand why we should introduce "random skills" or even logarithmic progression to explain sumksills diff for an old linear model that certainly didn't change after reform (coach misjudged)

you compare 12/2 14/4 but it's not the same ratio (or talent).
So +10 levels in X weeks from 2 to 12 is different talent than +10 levels in X weeks from 4 to 14? Is this a part of your model?

Even if level 2 is twice of level 4 sumkills, ending levels 12/14 are "not half of".
Not such a big diff in the end.

Of course it's not the half of, because I was talking about how progress is calculated in YOUR model and why it is wrong.



You have factor X, and Y weeks, so effectively you just multiply each starting skill by (1 + X*Y)... meaning you also multiply their starting sumskill by (1+ X*Y), in your example it's 1.4. It means if we have two players with the same talent and the same number of weeks then in your model the proportion between their starting sumskills should be preserved as proportion of their sumskills after finishing the youth school. If someone starts with sumskill 10 and has X*Y = 1.5 then will end as sumskill 25. If someone else starts with 20 then with the same X*Y will end with sumskill 50. If someone starts at 0.00 he will end with sumskill 0, even if he has top talent and 34 weeks in youth school.
Is that correct in your model or not?

Let's say we have two youths with the same top talent and high number of weeks, but one will start as [1] youth, and other will start as [4] youth (so few times higher starting sumskill). Will they gain the same number of overall levels in youth school?

The logarithmic regression should be used to determine the talent and final skill weeks before, its just a method to predict.
This is not related to sumskill, that one is determined by the endskill, so in the example of my outstanding (12) junior you will not apply logarithmic regression on the last day before promoting him. You are not predicting anymore. The endskill of the player is 12.00 to 12.99 therefore sumskill 39 to 43 according to the Gaston's tool.
Logarithmic regression can help to determine the value between 12.00 and 12.99, but not the sumskill, although its a function of endskill.

The reason why i am suggesting logarithmic regression over linear is that the higher skills jump slower than lower skills. This is true for the A team players, but i do assume that the same happens with juniors. If that is correct then applying linear regression is wrong.

The reason why i am suggesting logarithmic regression over linear is that the higher skills jump slower than lower skills. This is true for the A team players, but i do assume that the same happens with juniors.

It's not, in the old system with 100% correctness (no youth coach judgment impact on skill level) the talent was constant, without any changes depending on the level

[same player had same talent at level 5-9 and 10-14, irrespective of age change and current skill level]

Lets agree to disagree. I dont have proof, because the history in sokker viewer i have is only from 2020.
But i remember implementing linear regression into oSokker and knowing it was wrong. I think if you would dig in the forum, you would find my comment admitting that.

Talent is constant = junior level gain is constant. Still junior can pop up for example after 4, 5, 4, 4, 4, 4, 5, 4 weeks.

Sum-skill gain per junior level is not constant. I will show my function graph later. I need to take a closer look to data I've collected.

In the new system you cannot determine that, you can just assume. Lets start a new "challenge":

10 weeks in school should be enough for linear regression to determine the talent, 19 left. According to linear regression he should leave the school at 11.5 skill.

I am just not sure if i will remember this challenge in 19 weeks. The other was shorter 3 weeks only :-), although seem to be finished after 1 week already.

But i remember implementing linear regression into oSokker and knowing it was wrong.

And I remember that age/level had no impact on talent graph before the changes :)
But it's almost impossible to find info from that period [pre-2010], everyone was posting screenshots and those are already unavailable, deleted from servers.

Damn, is oSokker that old? :D

And damn, so am i :D

I will check also how much linear regression is mistaken after 10 weeks vs full training.
(editado)

effectively you just multiply each starting skill by (1 + X*Y)

So if (1 + X*Y) = 1.5 you multiply sumskills by 1.5

If sumskills 10 so 15
If sumskills 20 so 30

But anyway, you are right the model I wrote doesn't work, it seems too much cumulative. May be something like end = init * f*w.

If it's not a model with factor, in that case only random distrib can explain that 0 at start remain at 0 at ends as you explain, but in the end it doesnt explain why a predictive ending level (leg. high coefficient of reg) is sometimes -4 under this prediction and/or sumskills didn't work.
(editado)

And I remember that age/level had no impact on talent graph before the changes :)

so it is, and greg always said that this didn't change with the new school system

but anyway, school is Rauls next reform project and maybe we will get something new soon

I found info from 2006:)

"Seweryn Chudy nsm nsm nsm nsm znak znak znak znak śwt śwt śwt śwt cel cel cel cel cel bdb bdb bdb bdb solid solid solid solid db db db db db"

it's in reverse order [from highest to lowest] so I'm writing it "the normal way"

7 7 7 7 7 [5]
8 8 8 8 [4]
9 9 9 9 [4]
10 10 10 10 10 [5]
11 11 11 11 [4]
12 12 12 12 [4]
13 13 13 13 [4.end]

30 weeks, as you see it kept the same "tempo"
(editado)

I wonder...
Youth coach always makes an estimation.
So the end level of the youth at the time he can enter the A squad is also an estimation.
Isn't it plausible to assume that the effective end line might differ with the estimation of the coach and therefor does not line up correct in all cases, considering youth talent?
So the brilliant player with only 38 skillsum should be considered outstanding and the coach was 2 levels wrong in that case?

but anyway, school is Rauls next reform project and maybe we will get something new soon

I can't say that I'm looking forward to it... they already presented their idea for "new youth school" and it was nothing but a sk-money grab to lower the inflation...

for example they wanted to introduce possibility to "improve youth facilities" only so that you COULD hire a brilliant/magical/unearthly coach... and that was the main idea of these changes, you'd have to pay more to get the same that you have now without paying...

that itself is an amazing idea... there ALREADY is a problem with not enough nice youngsters on the market, so they will make it even harder to have a decent youth school, so people who don't have luck with YS will decide to close them more often... bravo
(editado)