Posted by JLKennedy44 on 11/23/2019 5:00:00 PM (view original):
Gill, thank you very much for the great answer. Let me throw one more item at you. When does an ability become a negative factor? For example, a PF with an 80 LP and a 19 ATH could indicate that this player is not really that good of a post player. But how bad of a player? And what is the tipping point where ATH rating takes away from a player's LP quality.
To put it another way:
The ATH Coefficient for LP could vary:
0-20 ATH = coefficient = .7
21-40 ATH - coefficient = .8
41-60 ATH - coefficient = .9
61-80 ATH - coefficient = 1.0
81-100 ATH - coefficient = 1.2
The reason I originally asked this question was that I want to start looking at recruiting differently and be more judicious when targeting players. So could the above table work?
to be honest, i am struggling a bit with how to respond to this. we are not quite on the same page about something but its hard to put my finger on what.
if we were to simplify the scoring for a big, into ath and lp, we might say you can approximate with:
0.3 * ath + 1 * lp
with 20 ath and 80 lp, this would be 6+80=86
with 80 ath and 80 lp, this would be 24+80=104
with 80 ath and 20 lp, this would be 24+20=44
this is the 'simple model' for rating evaluation we tend to use on these forums - a model where each ability is a sum of components, where each component is the value of a rating times a coefficient - that way its easier to talk about. its not a perfect model, but its not atrocious, either. that way we can compare the value of ratings for a certain purpose in simple terms. for a certain set of circumstances, is lp twice as important as ath for scoring, 3x, 4x, 5x? those would correspond to coefficients for (lp, ath) of (1, .5), (1, .33), (1, .25), (1, .2), respectively.
does that make sense? it seems to me that is the model most folks have in mind, in responding to you - i spent a lot of time talking about why that isn't perfect, because the situations vary so much - but in the end, am happy to use that model for talking about the value of ratings in a specific context, because its about the best we've come up with as a community over the past 15 years :)
i think in your model, you are looking at taking LP as the base, and increasing or dropping it based on ath. this is not necessarily a bad way to look at things, but it basically breaks down as soon as you consider a 3rd rating - like per (which is probably at least as valuable as ath for the scoring big types you are after - and many abilities really require you to at least consider 3 ratings, so this is important in general). also, it treats ath as equal across pretty broad ranges. so, if you are going to use a simple model like this, i would definitely use the (just as simple) model with a sum of parts (coefficient1 * ath + coefficient2 * lp type of thing). with that model, you can really easily set up an excel sheet or even just do it by hand with the calculator on your computer or something. in excel you can set the coefficients in a cell and reference them so you can change the formula on the fly and see how it impacts the rankings - which is a pretty valuable exercise and a great way to calibrate your formula to your eye test!
anyway, back to your actual question. its a bit of a trick question, to be honest. where does ath take away from lp - i suppose, any time it is below average for the level of play you are playing at in your division - so whats that, like 55-60? i would definitely think the 19ath in your example would take away heavily from your player, but if their LP is high enough, and particularly if you aren't facing very good defense, the guy might still be useful for you. i guess that doesn't really answer your question - here's what i would do -
pick a set of ratings that seem 'solid' to you, and coefficients that seem reasonable. let's suppose this is 60 ath, 70 lp, and ath being worth 30% as much as lp. then you need .3 * 60 + 1 * 70 = 88. then you can do the same calculation on the projected ratings for recruits you are considering, and basically evaluate if they are better or worse than that. you can do the same thing for scorers you consider like, borderline usable, great, exceptional - and do the same thing.