Ratings scale

[jmuelly]

I use 20-80 and this year I noticed by default rating > max is preselected.

Are others using this? I guess I never have but it was neat seeing some players who project to have extra special stuff beyond the normal elite levels.

[Paul Reuschel's Mustache]

Originally Posted by jmuelly

I use 20-80 and this year I noticed by default rating > max is preselected.

Are others using this? I guess I never have but it was neat seeing some players who project to have extra special stuff beyond the normal elite levels.

I always selected "Yes" for Show Ratings > Max. I believe in past versions, the default was "No". With 25, the default setting is "Yes".

[jmuelly]

Originally Posted by Paul Reuschel's Mustache

I always selected "Yes" for Show Ratings > Max. I believe in past versions, the default was "No". With 25, the default setting is "Yes".

That’s what I thought. I’ve never seen it display like this before. I know it’s not conventional to the 20-80 scale but where we can’t physically see these players it’s maybe kind of nice to see who has greater than elite stuff. Jacob Misiorowski for example in a non controlled 2024 season sim I ran last night was showing over 80 on multiple potentials. I think I may leave it, as it was kind of nice to see the offset in possibly finding those elite+ traits. It kind of would give guys historically like Trevor Hoffman, Mariano Rivera there own “special” ranking as they were creams of the crop.

[Syd Thrift]

I think nowadays those should be suuuuper rare since with the new standard deviations separating the 20-80 scale, an 80 already represents the top 2-3 players in a 30 team league.

[jcard]

Originally Posted by Syd Thrift

I think nowadays those should be suuuuper rare since with the new standard deviations separating the 20-80 scale, an 80 already represents the top 2-3 players in a 30 team league.

What are the “new standard deviations”? I know that the sub-MLB-caliber portion of scales have been made more granular, but (in addition to the fact that the displayed ratings are league-relative) there are multiple posts by developers clearly stating that the OOTP 20-80 scale does not function in the SD-based method as its real-life scouting counterpart (which, incidentally, brings into question whether the default setting of a 20-80 scale is really the best choice for OOTP users).

[Syd Thrift]

Originally Posted by jcard

What are the “new standard deviations”? I know that the sub-MLB-caliber portion of scales have been made more granular, but (in addition to the fact that the displayed ratings are league-relative) there are multiple posts by developers clearly stating that the OOTP 20-80 scale does not function in the SD-based method as its real-life scouting counterpart (which, incidentally, brings into question whether the default setting of a 20-80 scale is really the best choice for OOTP users).

The devs made a point to make the 20-80 scale this year specifically work with the 10 points = 1 standard deviation rule. You’re right, they specifically stayed away from that in the past, and even now it’s hard to do it exactly because ratings are not projected WAR, but this was a change in 25 compared to earlier versions.

[DaBears]

Originally Posted by Syd Thrift

The devs made a point to make the 20-80 scale this year specifically work with the 10 points = 1 standard deviation rule. You’re right, they specifically stayed away from that in the past, and even now it’s hard to do it exactly because ratings are not projected WAR, but this was a change in 25 compared to earlier versions.

Sorry for the dumb question, but how exactly does that work? One SD is 65%, two is 95% and three is 99%. If I assume 50 = 0, 60 = 1, 70 = 3 and 80 = 4.....

That would, as an example, I have 150 SP's (30 x 5), zero to one SP would be rated 80. If I have 240 RP (30 x 8), one to two would be 80. Finally, if I assume nine batters as starters per team, I've got 270 (30 x 9), so again, only maybe two to three (?) are 80's in the ratings. But there are more 80's in the game than that (maybe 5-6 using math above).

Just curious on your thoughts about that math. Interesting discussion item.

[Syd Thrift]

I think the game considers there being more than 5 starters per team and will also include top guys on the DL, the 40 man(?), etc. It’s also been stated that this was the attempt but that it’s a little flatter than real standard deviations, and also I think (pure speculation on my part) they’re just measuring the space from the median to the first standard deviation and then using that same number to indicate further standard deviations instead of being very fiddly with the numbers. I don’t know that anyone on the dev team has like a degree in stats (which, nothing against them; neither do I).

[jcard]

Originally Posted by Syd Thrift

The devs made a point to make the 20-80 scale this year specifically work with the 10 points = 1 standard deviation rule. You’re right, they specifically stayed away from that in the past, and even now it’s hard to do it exactly because ratings are not projected WAR, but this was a change in 25 compared to earlier versions.

Thank you for bringing me up to speed. Does that apply to attribute ratings as well? If so, then a 60 rating and a 40 rating in should imply performance in the associated metric that is equally better or worse, respectively, than league average, with the scale being (uni)linear rather than the split linear scales previously used. A standard deviation of a given population or sample trait is, by definition, “standard”. This does not seem possible for the absolute scale with its long left tail, but perhaps it would be a feature of league-relative attribute ratings. It will be interesting to learn what the expected stats in the current league function reveals about this. Anyway, thanks again for the information quoted above.

[jcard]

Originally Posted by DaBears

Sorry for the dumb question, but how exactly does that work? One SD is 65%, two is 95% and three is 99%. If I assume 50 = 0, 60 = 1, 70 = 3 and 80 = 4.....

That would, as an example, I have 150 SP's (30 x 5), zero to one SP would be rated 80. If I have 240 RP (30 x 8), one to two would be 80. Finally, if I assume nine batters as starters per team, I've got 270 (30 x 9), so again, only maybe two to three (?) are 80's in the ratings. But there are more 80's in the game than that (maybe 5-6 using math above).

Just curious on your thoughts about that math. Interesting discussion item.

The statement in bold is incorrect. A standard deviation is a measurement of dispersion of the trait in a population, agnostic to the shape of that distribution. The percentages you cite are NOT a definition of SD; rather, those percentages are among the defining parameters of a normal distribution: it is the fact that the shape of the trait among the population or sample that is one of the features that makes a given distribution “normal”. In any event, even if baseball talent among the human population is presumed to be normally distributed, the scouting scale is concerned with only a sample of that population located extending from some point on the right tail to the right terminus—professional baseball players. This regular (i.e., diminishing) right tail cannot have a normal distribution. It is asymmetrical, namely declining in occurrence as the aptitude level increases.

[Syd Thrift]

Originally Posted by jcard

Thank you for bringing me up to speed. Does that apply to attribute ratings as well? If so, then a 60 rating and a 40 rating in should imply performance in the associated metric that is equally better or worse, respectively, than league average, with the scale being (uni)linear rather than the split linear scales previously used. A standard deviation of a given population or sample trait is, by definition, “standard”. This does not seem possible for the absolute scale with its long left tail, but perhaps it would be a feature of league-relative attribute ratings. It will be interesting to learn what the expected stats in the current league function reveals about this. Anyway, thanks again for the information quoted above.

I'm pretty positive it's both and as noted it's not going to be perfect because OOTP ratings don't lie in a regular old normal distribution.

Also I SWEAR I made that clarification in this thread but I guess I didn't... that comment I made above was me being a dumb English degree person and not grokking/explaining standard deviations very well. I'm sure the devs are calculating SD properly. I'm also sure you're not going to get the results you'd expect from a normal distribution because OOTP ratings do not lie in a normal distribution and the only way you'd really be able to goose things so that only 3/150 starters were 80s would be if you made the overall SD size larger and therefore had fewer 55s and 60s as well.

[Pelican]

I try to stay far away from discussions on standard deviation, due to memories of math teachers past. I play with ratings from 1-100, because that makes sense to me. What does not make sense is the possibility of a rating above 100. For one thing, you can't (presumably) have ratings under 1, to balance that out. Second, I understood the whole point of a scale, whether 1-100 or the archaic 20-80, to be that all data points would fall somewhere on that scale, distributed (grrrr) according to some formula. So, of course I turn off ratings over 100; but then it bugs me that they are still out there, and some of the guys rated 100 are actually better than others, and wouldn't I want to know that? Math guys, explain please!

[Daniel_09]

Originally Posted by Pelican

I try to stay far away from discussions on standard deviation, due to memories of math teachers past. I play with ratings from 1-100, because that makes sense to me. What does not make sense is the possibility of a rating above 100. For one thing, you can't (presumably) have ratings under 1, to balance that out. Second, I understood the whole point of a scale, whether 1-100 or the archaic 20-80, to be that all data points would fall somewhere on that scale, distributed (grrrr) according to some formula. So, of course I turn off ratings over 100; but then it bugs me that they are still out there, and some of the guys rated 100 are actually better than others, and wouldn't I want to know that? Math guys, explain please!

I understand the reasoning, I think the detail here is that in all previous versions of the game the default was the max rating cut off, I don't know why in this version of the game the default is that you can see the maximum rating.

I don't know why in this version of the game the default is that you can see the maximum rating, but this is easy to change in the global settings.

[Syd Thrift]

Originally Posted by Pelican

I try to stay far away from discussions on standard deviation, due to memories of math teachers past. I play with ratings from 1-100, because that makes sense to me. What does not make sense is the possibility of a rating above 100. For one thing, you can't (presumably) have ratings under 1, to balance that out. Second, I understood the whole point of a scale, whether 1-100 or the archaic 20-80, to be that all data points would fall somewhere on that scale, distributed (grrrr) according to some formula. So, of course I turn off ratings over 100; but then it bugs me that they are still out there, and some of the guys rated 100 are actually better than others, and wouldn't I want to know that? Math guys, explain please!

The 1-100 ratings are not I don't believe divvied up by standard deviations; that is reserved for 20-80. The game just "knows" what a 50 equivalent rating is, or else it figures it out by looking at the league (I don't know the answer here) and then it "knows" how many points on the 1-550 scale you have to go up to go up a point on 1-100. When the scale was 1-250 ratings over 200 were considered every rare and "over 100" if you will but were necessary to account for generational players like Barry Bonds and 1-550 has the same super-high level that few if any players hit.

Also of course if you're playing in an international or minor league, the average player rating is going to be lower and that means that that a player might not even need to inhabit one of those once-in-a-generation ratings levels to be over 100.

If the game just gave you the actual rating divided by 5.5 you'd see:
- basically nobody with low or even below-50 ratings in the major leagues or even AAA
- Nobody with a 100, ever
- Very few people with ratings over 80 or so

[LansdowneSt]

Originally Posted by Pelican

I try to stay far away from discussions on standard deviation, due to memories of math teachers past. I play with ratings from 1-100, because that makes sense to me. What does not make sense is the possibility of a rating above 100. For one thing, you can't (presumably) have ratings under 1, to balance that out. Second, I understood the whole point of a scale, whether 1-100 or the archaic 20-80, to be that all data points would fall somewhere on that scale, distributed (grrrr) according to some formula. So, of course I turn off ratings over 100; but then it bugs me that they are still out there, and some of the guys rated 100 are actually better than others, and wouldn't I want to know that? Math guys, explain please!

1921.

"This one goes to 11..."