The Confusing Star System

[Juggernt]

I trust my own evaluation of players based on their tools, and I understand that the star system is relative to other players at the same position, but it still confuses the hell out of me when a player's abilities and stars move in different directions. Abilities go down, stars go up; abilities go up, stars go down. That the movement is based on an ability going up or down a single point is head-scratching as well.

I'm going to continue to trust myself, but it's easy to get seduced by seeing stars.

[ThePretender]

Just ignore stars, they're not really accurate. Go by your own evaluation.

[NoOne]

it's just a correlation. like you allude to... some 4 star are not as good as a 3 star. if it doesn't adjust to era / statistical environment (settings, not results), it's not even necessarily accurate for your league, either.

as long as accuracy is 100%, no stats included in the evaluation and not relative to position it's definitely consistent in moving up/down and the causes. those are choices a player makes. your own fault if you don't like it (not being snotty, but maybe you didn't realize this and want it different?)

as far as that choice, i'd suggest not making it from your perspective. as a human you can adapt... the AI, not so much. i'd set these tihngs to mold/shape ai behaviour as you see fit. (that's definitely not an argument for 100% accuracy, btw... flaws should exist. ai gm's should make bad choices too. i'd even back up using ai eval including stats, even though ratings are more strongly correlated with results- most often, depends on settings a bit)

you already figured out that you can do better... even if you have ovr/pot calculated in an absolute manner via settings, it's still not as good as you can do.

especially if you use a larger scale for the ratings... the rounding errors alone make it terrible in comparison. (it's a 10pt scale, even though there are 5 stars, any scale larger offers better information).

[Brad K]

Originally Posted by NoOne

it's just a correlation. like you allude to... some 4 star are not as good as a 3 star. if it doesn't adjust to era / statistical environment (settings, not results), it's not even necessarily accurate for your league, either.

as long as accuracy is 100%, no stats included in the evaluation and not relative to position it's definitely consistent in moving up/down and the causes. those are choices a player makes. your own fault if you don't like it (not being snotty, but maybe you didn't realize this and want it different?)

as far as that choice, i'd suggest not making it from your perspective. as a human you can adapt... the AI, not so much. i'd set these tihngs to mold/shape ai behaviour as you see fit. (that's definitely not an argument for 100% accuracy, btw... flaws should exist. ai gm's should make bad choices too. i'd even back up using ai eval including stats, even though ratings are more strongly correlated with results- most often, depends on settings a bit)

you already figured out that you can do better... even if you have ovr/pot calculated in an absolute manner via settings, it's still not as good as you can do.

especially if you use a larger scale for the ratings... the rounding errors alone make it terrible in comparison. (it's a 10pt scale, even though there are 5 stars, any scale larger offers better information).

Great post.

Let me add to this that as far as I know how stars are calculated is not public knowledge. (Ignore the rest of this post if this is incorrect, and please give me a link to the documentation.)

So, if the method isn't known, we don't know if the method reflects reality. Actually we don't know what all of reality is, even as old as baseball is.

For example, avoiding strikeouts. Since players have a rating for this we can assume that somehow its figured into the star ratings. Yet there are present arguments that strikeouts are just another kind of out and the stat doesn't mean much. Are those who say this right?

Well, a player who strikes out a lot probably hits into fewer double plays. OTOH, he may have fewer sacrifice flies. And then there's the team strategy factor. It matters less if his team doesn't do hit and run.

In conclusion, even with star ratings set to accurate, they may not reflect reality, sometimes its unproven what reality is, and may not fit the manager's idea of how to run the game on the field.

[NoOne]

i like your conclusion but not how you got there

overall/potential = star ratings, as i use it in this post. assumes infinite sample when needed too.

the strikeout thing as an example -- from a league-wide perspective and working with LTM/LT that's ~close for sure. everythign has to add up to 1, and a portion of that will be the various types of outs etc.. however, shifting it around will have an impact on individual players in different ways due to their ratings, though.

e.g. a player with high avoid K's in an era with yearly league totals of 10,000k's vs 36,000k's has a completely different meaning. i'd assume it's a bit more important in the league with 36k/year.

I am not sure if overall/potential adjusts for statistical environment, but it easily could and we could easily deduce that too.

we can deduce how it is calculated -- using 100% accuacy, AI eval 100% ratings and not relative to league nor position. change one rating at a time and you'll figure out the equation. ** also using 1-200(+50) editor scale

then, go change league totals and recalc modifers (jsut in case) and see if it changed the ratings.. if not, it doesn't adjust for environment and that's one more reason to never look at overall potential for anything beyond a loose idea of quality.

that alone would be reason enough for discrepancies between results and star ratings and why some may do better than others at the same value, but not the only reason.

the same star rating can have drastically different results is because multiple combinations of ratings equate to that same value... it reads the same, but they are not the same. individual ratings are used for various things in game, not the overall potential... it is visual feedback for your eyes only (and AI logic).

They aren't meant to be read as "equal" if two palyers are 3 stars. Stars are strongly correlated to results (weaker with less accuracy), but don't cause any results on the field.

an individual striking out more is in no way a net-benefit just had to say that too. even a SO in a DP situation reduces chances to score for the team. easy to lose perspective when looking at detail.

[Brad K]

Originally Posted by NoOne

an individual striking out more is in no way a net-benefit just had to say that too. even a SO in a DP situation reduces chances to score for the team. easy to lose perspective when looking at detail.

I don't think I said that. I pointed out there is some thought that a K is merely another out, different from the traditional view that a K is the worst type of failure possible at the plate, identifying some points but not expressing a view.

[Brad K]

Originally Posted by NoOne

i like your conclusion but not how you got there

Thank you! :

(Better to have a disagreement over the method than the results.)