Subpage under development, new version coming soon!
Asunto: »»Youth reform: Changes in youth academies / junior traini
depends on how long you train them. I'm discovering training past 26 is horrible.
I have a kid Dakota Hodge who has 3.85 talent. Pulled mid-16, about as early as you reasonably can. Unearthly training, and magical x2/brilliant x1 assistant.
He will end up (estimate) maybe 3x16 if I stop at end of 27. His subskills won't be high either, except pace.
Not sure where this divine is coming from, unless you rob something else.
I have a kid Dakota Hodge who has 3.85 talent. Pulled mid-16, about as early as you reasonably can. Unearthly training, and magical x2/brilliant x1 assistant.
He will end up (estimate) maybe 3x16 if I stop at end of 27. His subskills won't be high either, except pace.
Not sure where this divine is coming from, unless you rob something else.
If you join at 16 with 34 weeks, you exit at 18.6. Anything past that point is sub-optimal, but not useless.
There is an important detail missing in this discussion about talents. I will write it in detail, I could probably make it shorter but I don't want to discuss too much or argue about it later. If someone doesn't understand it then it's not my problem :p
The real "talent" is a value in sokker database, probably in 0-100 scale, where the higher value the better. It's randomly assigned when junior comes to youth school, likely with normal distribution or something similar, likely with a median at 50 (meaning extremely high and low talents are rare). This "talent" value doesn't change between youth school and regular training. It's used as a multiplier to training progress in youth school training and in normal training, BUT it's not used directly, and I'm certain it's used differently in both situations.
Users measure talent in inverted scale - in youth school as a number of weeks needed on average for a pop (or as basic tools measure it, inverse of the slope value for linear regression function, sometimes with some error correction). Sokker Asistente popularised calculating talent from normal training using the same scale - max talent at ~3.0, and worse talents linearly scaling training points needed for pop (for example if player needs 300 points with talent 3.0 to pop in skill X at age Y then with talent 4.0 he would need 400 points), but as I will explain below talent from YS doesn't translate this easily into talent in normal training.
Of course I don't know exactly how talent works in training, but based on real data I can make few points:
- talents in youth school are not that often under 4.0, I would say the average is in 5.0-6.0 range, and sometimes they look extremely bad like 15 or worse, not possible to be explained just by coach errors, so really bad YS talents must be possible
- on the other hand talents in normal training (calculated in Sokker Asistente style) are very often under 4 and I have never seen talent worse than 6, there must be some limit to how bad the effective talent can be in normal training (although probably it's possible to get talents a bit worse than 6, that's just the worst I encountered)
- generally YS talents (that aren't top talents) translate into better normal training talents in cases I could verify, for example ~4.3 YS talent (with 34 weeks of YS data) later had talent ~3.75 in normal training. The closer to max talent the lower that difference will be.
To be clear - I say "calculated in Sokker Asistente style", but I'm not talking about SA calculations, as it's not that good after training changes. I have my own tool that is able to calculate talent in a similar way, but thanks to better training model used it is very accurate (with enough data) and doesn't produce contradictory results like SA (min talent much higher than max talent). I plan to make a simplified public version of talent calculator in few weeks ;)
Back to topic:
If I were writing the code for sokker training then the easiest way to use the 0-100 talent value in training calculation and at the same time have some minimal training guaranteed would be to scale that value to X-100 range for some X. For example if we want the worst possible talent to give 40% training of that with the best possible talent we could scale it by doing this: (40 + (100-40) * talent) and using this as a multiplier in a training formula.
If this is the way greg/damian programmed it then the data suggests that this "minimal training %" is low for youth school training (as terrible YS talents are possible) and it's substantially higher for normal training (as there seems to be a hard limit on how bad that training can be).
10% minimal training for YS training and 40% in normal training produce values that roughly match my data. With this assumption here is a table that would translate "sokker database talent" into user-friendly YS talent value and normal training talent value (SA style). So for example 4.01 (~4.0) YS -> 3.61 (~3.6) normal training. But again, I don't say it works exactly like that, it's a rough estimation based on my data.
Btw. I assume top coach in youth school, unearthly / high magical, if someone has worse coach then obviously it will negatively affect YS talent making the difference even bigger.
The real "talent" is a value in sokker database, probably in 0-100 scale, where the higher value the better. It's randomly assigned when junior comes to youth school, likely with normal distribution or something similar, likely with a median at 50 (meaning extremely high and low talents are rare). This "talent" value doesn't change between youth school and regular training. It's used as a multiplier to training progress in youth school training and in normal training, BUT it's not used directly, and I'm certain it's used differently in both situations.
Users measure talent in inverted scale - in youth school as a number of weeks needed on average for a pop (or as basic tools measure it, inverse of the slope value for linear regression function, sometimes with some error correction). Sokker Asistente popularised calculating talent from normal training using the same scale - max talent at ~3.0, and worse talents linearly scaling training points needed for pop (for example if player needs 300 points with talent 3.0 to pop in skill X at age Y then with talent 4.0 he would need 400 points), but as I will explain below talent from YS doesn't translate this easily into talent in normal training.
Of course I don't know exactly how talent works in training, but based on real data I can make few points:
- talents in youth school are not that often under 4.0, I would say the average is in 5.0-6.0 range, and sometimes they look extremely bad like 15 or worse, not possible to be explained just by coach errors, so really bad YS talents must be possible
- on the other hand talents in normal training (calculated in Sokker Asistente style) are very often under 4 and I have never seen talent worse than 6, there must be some limit to how bad the effective talent can be in normal training (although probably it's possible to get talents a bit worse than 6, that's just the worst I encountered)
- generally YS talents (that aren't top talents) translate into better normal training talents in cases I could verify, for example ~4.3 YS talent (with 34 weeks of YS data) later had talent ~3.75 in normal training. The closer to max talent the lower that difference will be.
To be clear - I say "calculated in Sokker Asistente style", but I'm not talking about SA calculations, as it's not that good after training changes. I have my own tool that is able to calculate talent in a similar way, but thanks to better training model used it is very accurate (with enough data) and doesn't produce contradictory results like SA (min talent much higher than max talent). I plan to make a simplified public version of talent calculator in few weeks ;)
Back to topic:
If I were writing the code for sokker training then the easiest way to use the 0-100 talent value in training calculation and at the same time have some minimal training guaranteed would be to scale that value to X-100 range for some X. For example if we want the worst possible talent to give 40% training of that with the best possible talent we could scale it by doing this: (40 + (100-40) * talent) and using this as a multiplier in a training formula.
If this is the way greg/damian programmed it then the data suggests that this "minimal training %" is low for youth school training (as terrible YS talents are possible) and it's substantially higher for normal training (as there seems to be a hard limit on how bad that training can be).
10% minimal training for YS training and 40% in normal training produce values that roughly match my data. With this assumption here is a table that would translate "sokker database talent" into user-friendly YS talent value and normal training talent value (SA style). So for example 4.01 (~4.0) YS -> 3.61 (~3.6) normal training. But again, I don't say it works exactly like that, it's a rough estimation based on my data.
Btw. I assume top coach in youth school, unearthly / high magical, if someone has worse coach then obviously it will negatively affect YS talent making the difference even bigger.
So basically following youth talent is pointless...cause it is no exact science to knowing the real talent. Obviously, the better the youth talent, the better the real talent, but still, following youth talent as being exact knowledge is actually pointless. You can only say for sure which talent a player has by following his training routine in the A-squad for weeks on. Especially since your youth coach is also only estimating the talent and the skill of the youth players...
(editado)
(editado)
The talent formula will never be worked out, especially under this new youth estimation system. There are too many variables to establish a baseline.
However if you wanted to try, you would need
A perfect head coach
A perfect youth coach
Multiple "perfect" talent youths, ideally at 34 weeks
And you'd need to train them consistently in the same skill over at least 20 weeks to account for the age update.
The data you need is "Where is the baseline?" That is, at what age and level is that perfect 3 talent actually at 3, because we all know such players can pop at 1 & 2 weeks.
Someone data mining this site with their own tools could maybe get this level of information over the space of 10 seasons, with the cooperation of managers.
However if you wanted to try, you would need
A perfect head coach
A perfect youth coach
Multiple "perfect" talent youths, ideally at 34 weeks
And you'd need to train them consistently in the same skill over at least 20 weeks to account for the age update.
The data you need is "Where is the baseline?" That is, at what age and level is that perfect 3 talent actually at 3, because we all know such players can pop at 1 & 2 weeks.
Someone data mining this site with their own tools could maybe get this level of information over the space of 10 seasons, with the cooperation of managers.
Did you miss the paragraph where I wrote about my very accurate tool? :p
You don't need to tell me what I need to do, that job is already done ;) My training model calculations match real player progress (over many seasons of training, for some players as far as training reform goes) with ±1% accuracy. I have it calibrated so the best talent I checked (and I checked many top players) is at 3.05. We could argue if this value should be moved 1% up or down, but it doesn't really matter, because at the same time we will never know what is the max talent possible in youth school, could be 3.00 or 3.05 or 3.10...
You don't need to tell me what I need to do, that job is already done ;) My training model calculations match real player progress (over many seasons of training, for some players as far as training reform goes) with ±1% accuracy. I have it calibrated so the best talent I checked (and I checked many top players) is at 3.05. We could argue if this value should be moved 1% up or down, but it doesn't really matter, because at the same time we will never know what is the max talent possible in youth school, could be 3.00 or 3.05 or 3.10...
I have significant doubts that max talent is anything above 3.14 but it's impossible to prove over the space of 34 weeks.
In the old times when youth level was exactly known there were cases of juniors with 10 pops in a row after 3, meaning their talent had to be under 3.10.
https://sokker.org/forum_topic/ID_forum/1/ID_topic/1455065/pg/50 - here in 2009 Gaudin shows such junior, but screenshots obviously don't work anymore :p Although he also described the progress in text: joined YS as [6], popped in first training to [7], and then pops every 3 weeks to up to [17]
https://sokker.org/forum_topic/ID_forum/1/ID_topic/1455065/pg/50 - here in 2009 Gaudin shows such junior, but screenshots obviously don't work anymore :p Although he also described the progress in text: joined YS as [6], popped in first training to [7], and then pops every 3 weeks to up to [17]
I think 10 in a row still wouldn't guarantee 3.10 due to decimals potentially screwing up the starting point.
I did some math a couple of months back that suggested 3.14 but I posted it in General and can't find it.
I did some math a couple of months back that suggested 3.14 but I posted it in General and can't find it.
Do you really need me to do the math on 10 pops in a row after 3 case? :p
The worst case scenario for talent is starting at 6.99999999 and going in 31 weeks to exactly 17.00, so just over 10.00 levels gains in 31 weeks = talent just under 3.10
The worst case scenario for talent is starting at 6.99999999 and going in 31 weeks to exactly 17.00, so just over 10.00 levels gains in 31 weeks = talent just under 3.10
Can you explain, simplified, how you can be sure of talents and levels of youth players when all data is based on estimations? Like... a player can come into the youth school at level 6 (adequate) for example. And there's an unearthly youth coach at hand. How do you know which is his exact skill at that time? and not only that, but decimals taken into consideration also?
You can’t know the subskills under the current system I’m pretty sure. That’s because you have intentional errors thrown in every week.
In the old system where there were no errors you’d still not know the exact subskills but you could work out best and worst scenarios which is what the above is measuring. Happy to be corrected but that’s my understanding.
(editado)
In the old system where there were no errors you’d still not know the exact subskills but you could work out best and worst scenarios which is what the above is measuring. Happy to be corrected but that’s my understanding.
(editado)
junior school reform should have the highest priority, but at the end of every season we are given updates that have minimal impact on the gameplay rather than what we need
this is all I got today. I am tired of this bullshit that I cannoy influence in any way
Richard Spálenka 18 tragic [0] GK 18
(editado)
this is all I got today. I am tired of this bullshit that I cannoy influence in any way
Richard Spálenka 18 tragic [0] GK 18
(editado)