Subpage under development, new version coming soon!
Asunto: »»Youth reform: Changes in youth academies / junior traini
The error is from -2 to +2.... thats why it can have a difference of 4... so when you suggest 1 or 2, I think you misjudge why they can change 4 levels ... a 4 level jump/drop should though actually help in evalute, because you can exclude things with it
Yes, actually I like this, If I understood well your point then I agree. It helps to reduce uncertainty and get a better talent calculation, altough it could increasse uncertainty for the real level calculation if not many many weeks of data
(editado)
Yes, actually I like this, If I understood well your point then I agree. It helps to reduce uncertainty and get a better talent calculation, altough it could increasse uncertainty for the real level calculation if not many many weeks of data
(editado)
That would mean that it's possible to encounter a difference of 8 between past and current level of junior. I'm not saying it's impossible (assuming coach 15+), but I don't recall seeing such big difference either. I saw 4 for sure, probably also 5, but 6-7-8? Doubtfully. Can You show some image/images of such big difference? I am now curious.
1. My point is that you are wrong about your 3-6 talent range. And by bad talents in your database I mean those that show terrible talents when calculated with linear regression, like these ones.
2. No, 16 wouldn't be median, 5.45 would be. If you read my initial comment with understanding then you would know why.
I don't have this number set in stone, it can be bit higher or lower... for example if max youth school talent is 3.1 instead of 3.0 then it would automatically make this median higher. Or if minimal training is different than 10% then median would also be different.
3. It doesn't matter what is the exact talent of a one single junior. Surely Sokker Asistente is heavily wrong about many of my junior talents, especially those with few weeks in youth school, but it doesn't really matter. For every junior that is underestimated (like talent 7 instead of 5) there should be on average one other that is overestimated (like talent 5 instead of 7). In the end overestimations and underestimations will mostly cancel out and the median should stay similar to the real one. To be clear, I know that uncertainty in youth level changes distribution of talents calculated with linear regression, in this case it makes extreme values appear more likely (or not possible at all like talents under 3 or negative talents), I just claim it doesn't change median by much.
Ok it's magic, but in the end, you calculate and compare in senior team 2 different players with different talents, right ? So what if junior's talents are not what you think they are ? How can you expect things until you are not sure at 100% about talent.
You are, again, missing my point. It's not "I can 100% accurately tell what senior talent will be based on youth school progress", because of course there will always be inaccuracy in youth school talent estimation, meaning inaccuracy in senior talent estimation. Sometimes bigger, sometimes smaller, and very often prediction is impossible if youth school talent is impossible to narrow down.
The point is: users shouldn't directly compare youth school talent with talents that are shown in Sokker Asistente, because it may lead to wrong conclusions, like giving up on a player because his youth school talent was too weak when it wouldn't be too bad in senior team.
Real example without magic: when someone has junior with 34 weeks in youth school and pretty stable chart indicating talent 4.30 from linear regression (like this one) then he can make assumption that he is very likely in 4.1-4.5 youth school talent range (with 4.2 - 4.4 most likely) and then based on the table I provided that he will be very likely in 3.65 - 3.85 senior talent range calculated from trainings in Sokker Asistente or my future tool (and most likely in 3.70 - 3.80 range). I checked him with my tool with his full senior training history and there he has 3.80 talent (would be similar in Sokker Asistente, probably a bit lower).
2. No, 16 wouldn't be median, 5.45 would be. If you read my initial comment with understanding then you would know why.
I don't have this number set in stone, it can be bit higher or lower... for example if max youth school talent is 3.1 instead of 3.0 then it would automatically make this median higher. Or if minimal training is different than 10% then median would also be different.
3. It doesn't matter what is the exact talent of a one single junior. Surely Sokker Asistente is heavily wrong about many of my junior talents, especially those with few weeks in youth school, but it doesn't really matter. For every junior that is underestimated (like talent 7 instead of 5) there should be on average one other that is overestimated (like talent 5 instead of 7). In the end overestimations and underestimations will mostly cancel out and the median should stay similar to the real one. To be clear, I know that uncertainty in youth level changes distribution of talents calculated with linear regression, in this case it makes extreme values appear more likely (or not possible at all like talents under 3 or negative talents), I just claim it doesn't change median by much.
Ok it's magic, but in the end, you calculate and compare in senior team 2 different players with different talents, right ? So what if junior's talents are not what you think they are ? How can you expect things until you are not sure at 100% about talent.
You are, again, missing my point. It's not "I can 100% accurately tell what senior talent will be based on youth school progress", because of course there will always be inaccuracy in youth school talent estimation, meaning inaccuracy in senior talent estimation. Sometimes bigger, sometimes smaller, and very often prediction is impossible if youth school talent is impossible to narrow down.
The point is: users shouldn't directly compare youth school talent with talents that are shown in Sokker Asistente, because it may lead to wrong conclusions, like giving up on a player because his youth school talent was too weak when it wouldn't be too bad in senior team.
Real example without magic: when someone has junior with 34 weeks in youth school and pretty stable chart indicating talent 4.30 from linear regression (like this one) then he can make assumption that he is very likely in 4.1-4.5 youth school talent range (with 4.2 - 4.4 most likely) and then based on the table I provided that he will be very likely in 3.65 - 3.85 senior talent range calculated from trainings in Sokker Asistente or my future tool (and most likely in 3.70 - 3.80 range). I checked him with my tool with his full senior training history and there he has 3.80 talent (would be similar in Sokker Asistente, probably a bit lower).
Sokker Asistente estimates the juniors talent based on a linear regression of their values in the school. IS Up to the user to make his own interpretation. The more weeks of data, the less uncertainty, but there's not a fixed number of weeks where you can say that that value is reliable
As I always say, an estimation IS different form a calculation because when you estimate, you are making some Guess, whereas when you calculate, you have full control on the accuracy
On the other hand, when you have a well defined concept like "talent" applied to Juniors, It makes sense to try to keep than concept in the senior players, since It seems It IS the same value for every player
Before SA, people here sometimes manipulated the concept by saying things like "talent 3 to excellent pace" without specifying the age when their player got that numbers. It was ugly to me. For myself I estimated the talent of my players with that Juniors concept of talent and that's what I programmed in SA
Of course It IS not perfect since we don't know the exact values of anything. In fact, I only used one of my players to make the correspondence between the juniors talent and the senior talent. I didn't care too much about it at that moment because there was so much uncertainty in everything that I was Happy with that approximation
I used the factors that best fit to my players and the predictions are great for me, but since It could depend on factors I can't control, like sublevel of coaches or any other unkown factors, I introduced an option so that everyone could configure their own factors
I never kept a Big database of players to make statistics. I mainly used formulas that I read on the forums that serm to work somehow. Maybe it's time to rethink about them with some reliable data?
(editado)
As I always say, an estimation IS different form a calculation because when you estimate, you are making some Guess, whereas when you calculate, you have full control on the accuracy
On the other hand, when you have a well defined concept like "talent" applied to Juniors, It makes sense to try to keep than concept in the senior players, since It seems It IS the same value for every player
Before SA, people here sometimes manipulated the concept by saying things like "talent 3 to excellent pace" without specifying the age when their player got that numbers. It was ugly to me. For myself I estimated the talent of my players with that Juniors concept of talent and that's what I programmed in SA
Of course It IS not perfect since we don't know the exact values of anything. In fact, I only used one of my players to make the correspondence between the juniors talent and the senior talent. I didn't care too much about it at that moment because there was so much uncertainty in everything that I was Happy with that approximation
I used the factors that best fit to my players and the predictions are great for me, but since It could depend on factors I can't control, like sublevel of coaches or any other unkown factors, I introduced an option so that everyone could configure their own factors
I never kept a Big database of players to make statistics. I mainly used formulas that I read on the forums that serm to work somehow. Maybe it's time to rethink about them with some reliable data?
(editado)
1. like I said, I didn’t calc above talent 5, nothing surprising. I didn’t expect any range from skt db (never said 3-6 comes from here) neither from Terrion db, neither from linear regressions without high correlation.
2. You expect 5.45 because it’s the median for all talents in terrion db. It makes sense and it should be almost ok if most of talents was only correct. But it’s not. Coefficient of correlation should be close to 1 for a regular progression, and most of the times it’s very far from 1. You can t conclude something base on mostly unreliable data.
If you retain only (let’s say 0.8 correlation, not 100%) talents in terrion db, then median should be much less biased.
3. 4. If you expect that underestimated things should equilibrate overestimated things it’s very lazy "math", but ok.
I don't know how you (or Terrion) calculate senior talents so I can’t say much about, but I just read « a senior talent is better than its junior talent ». But so what ? training system in junior or senior are 2 different thing, nothing surprising.
But If im not wrong it was not your initial point ? Your claim was "if a player have twice a junior talent compare to another one, senior training speed is not twice as different, but closer". Right ? And my question was : are you sure about bothjunior talents at its base ? Because once again u can't compare things if both talents are not reliable at its base. Just my point, nothing else.
(editado)
2. You expect 5.45 because it’s the median for all talents in terrion db. It makes sense and it should be almost ok if most of talents was only correct. But it’s not. Coefficient of correlation should be close to 1 for a regular progression, and most of the times it’s very far from 1. You can t conclude something base on mostly unreliable data.
If you retain only (let’s say 0.8 correlation, not 100%) talents in terrion db, then median should be much less biased.
3. 4. If you expect that underestimated things should equilibrate overestimated things it’s very lazy "math", but ok.
I don't know how you (or Terrion) calculate senior talents so I can’t say much about, but I just read « a senior talent is better than its junior talent ». But so what ? training system in junior or senior are 2 different thing, nothing surprising.
But If im not wrong it was not your initial point ? Your claim was "if a player have twice a junior talent compare to another one, senior training speed is not twice as different, but closer". Right ? And my question was : are you sure about bothjunior talents at its base ? Because once again u can't compare things if both talents are not reliable at its base. Just my point, nothing else.
(editado)
Lol no
Once again you prove you got no math sense
If -4 to +4 was the range you would see 6, 7, 8 or Even 9 skill levels difference between two weeks
(editado)
Once again you prove you got no math sense
If -4 to +4 was the range you would see 6, 7, 8 or Even 9 skill levels difference between two weeks
(editado)
I remember reading that someone did the math on this and suggested it was +2.2 to -2.2. I can't remember how exactly they came to that number but it was something to do with the opening level of new juniors.
I'm curious how this applies to juniors actually, if the max level they can join at is 7.2, surely we should be seeing some enter at Very Good.
I'm curious how this applies to juniors actually, if the max level they can join at is 7.2, surely we should be seeing some enter at Very Good.
Hahahaha if dtox were right we could see formidable entries xD xD xD
I've seen numerous youth players drop and rise 4 skills in one week, even with magical youth coaches!
The fact you don't see 7, 8 or 9 difference in skill is because,
A) Starting skills are maxed on level 7 and skills can drop or rise 4 skills in one week, but that
B) means it's a correction from the previous level. Those drops and rises only occur once in a youth progress.
I don't get why it should be looked at in 2 weeks. If it can occur in one week with a difference of 4 skillpoints, then the range is -4 +4, right?
I never said then range is from -4 all the way back to +4 in one week. I said that a skill can rise or drop 4 skillpoints in one week.
The weirdest one I've seen in my youth is a player that came in as average, stayed average for a couple weeks, then dropped to unsatisfactory, then rise back up to average and then rose further up to solid, so that's from 2 to 8 in 2 weeks. The week after he dropped to average again. And all that with a magical youth coach. In the end he came out as a formidable youth player. I don't recall the amount of weeks he was in my team. It's been several years now..
The fact you don't see 7, 8 or 9 difference in skill is because,
A) Starting skills are maxed on level 7 and skills can drop or rise 4 skills in one week, but that
B) means it's a correction from the previous level. Those drops and rises only occur once in a youth progress.
I don't get why it should be looked at in 2 weeks. If it can occur in one week with a difference of 4 skillpoints, then the range is -4 +4, right?
I never said then range is from -4 all the way back to +4 in one week. I said that a skill can rise or drop 4 skillpoints in one week.
The weirdest one I've seen in my youth is a player that came in as average, stayed average for a couple weeks, then dropped to unsatisfactory, then rise back up to average and then rose further up to solid, so that's from 2 to 8 in 2 weeks. The week after he dropped to average again. And all that with a magical youth coach. In the end he came out as a formidable youth player. I don't recall the amount of weeks he was in my team. It's been several years now..
I don't get why it should be looked at in 2 weeks. If it can occur in one week with a difference of 4 skillpoints, then the range is -4 +4, right?
You are really bad at math/logical thinking :P
4 levels difference is -2/+2
Enters as [4], shows him as [6], then overcorrects to [2]
You are really bad at math/logical thinking :P
4 levels difference is -2/+2
Enters as [4], shows him as [6], then overcorrects to [2]
chill, I think you are not talking the same as we are.
2 weeks means before training/after training. so.. just 1 update
4 skills in one week, even with magical youth coaches!
yes, because he was for example: week1: level 10, --> week2: level:6. It meanshe had a drop of 4. But the real level is 8, then he jumped inside the +/- 2 of 'uncertainty'
The most extreme case is for example he is shown 6 and then he pop up to 11. it would mean he was 8 before update and then 9 after update (this is a perfect scenario to know the real actual level)
(editado)
2 weeks means before training/after training. so.. just 1 update
4 skills in one week, even with magical youth coaches!
yes, because he was for example: week1: level 10, --> week2: level:6. It meanshe had a drop of 4. But the real level is 8, then he jumped inside the +/- 2 of 'uncertainty'
The most extreme case is for example he is shown 6 and then he pop up to 11. it would mean he was 8 before update and then 9 after update (this is a perfect scenario to know the real actual level)
(editado)
Yes, I'm bad at math.
I see things rationally, not mathematically. That's why I was looking towards the range as being +4 -4, because he can pop and drop 4 levels at once.
But if you're saying the 4 is always a range from the centre out, ok, then that's another way of looking at it.
In the end we're saying the same, but in a different matter.
Fact is that it's lame and stupid to work with probabilities and uncertainties, to work towards certainties.
I thought this was a game, not a mathematical show off. My bad.
I see things rationally, not mathematically. That's why I was looking towards the range as being +4 -4, because he can pop and drop 4 levels at once.
But if you're saying the 4 is always a range from the centre out, ok, then that's another way of looking at it.
In the end we're saying the same, but in a different matter.
Fact is that it's lame and stupid to work with probabilities and uncertainties, to work towards certainties.
I thought this was a game, not a mathematical show off. My bad.
but there's not a fixed number of weeks where you can say that that value is reliable
But there is a tool. You may use coefficient of correlation beside talent in SA then it could help app (or users) to filter unreliable talents.
https://www.geeksforgeeks.org/program-find-correlation-coefficient/
But there is a tool. You may use coefficient of correlation beside talent in SA then it could help app (or users) to filter unreliable talents.
https://www.geeksforgeeks.org/program-find-correlation-coefficient/