OVR Ratings Bell Curve?

[cbbl]

OK, another thread on the never-to-have-consensus topic of ratings and evaluation parameters.

I've been fooling around with various combinations between the player rating options (relative to league; pure ratings v stats) and the evaluation AI settings. Ultimately, I'm trying to manipulate the settings to see a reasonable curve in the number of players in each rating category (blue, green, yellow, orange, and red). I have a small major league (6 teams, 154 players). All analysis is done with scouting accuracy at 100% (I play at normal). All ratings on on the 20-80 scale.

Here's my current settings:
* Relative to major league
* Based on AI evaluation
* AI Evaluation settings: 40 / 40 / 20 / 0

This yields the following:
Blue - 1
Green - 10
Yellow - 69
Orange - 69
Red - 5

This distribution seems low ... as the mean is hovering around 45.

Eliminating the impact of stats, I next looked at the distribution by turning off the Based on AI Evaluation setting. I now get:

Blue - 0
Green - 10
Yellow - 65
Orange - 74
Red - 5

This is driving the mean LOWER -- somewhere around 43 or so.

I think, based on pure ratings, when comparing players to each other in the league (and NOT by position), I would see a bell curve. Something like:

Blue - 15
Green - 31
Yellow - 62
Orange - 31
Red - 15

In short, in seems I currently have too many players rated orange rather than green or blue, but I'm not sure 1) whether the bell curve should be expected or 2) what settings would yield the results.

My third and final attempt to get the bell curve saw me change the AI Evaluation settings to all ratings: 100 / 0 / 0 / 0. I suspect that if this has the same impact as my second experiment (Based on AI evaluation OFF), and sure enough it did.

I welcome your thoughts. What do you think is the appropriate distribution between the five color categories?

[ThePretender]

I think the bigger issue is just that ovr/pot do a terrible job of giving you a quick view of how talented a player is or will be.I have one player with a 57 overall/potential, who over his career has been - 5.3 WAR, 3 WAR (99 games, 4.5 WAR per 150), 7.9 WAR, 4.2 WAR (injured, prorated to 5 WAR per 150 games), and this year is on pace for 6.8. So why does Ovr/Pot rate a guy who over his career has averaged 6.2 WAR per 150 games as a 57? Because he doesn't hit for power.

Despite the fact a slightly above average defensive 2B owns a 131 wRC+ for his career, he's a 57 ovr/pot because he hits for no power. When I think 57 ovr/pot, I don't think of guys who average 6 WAR per 150 games. There's no ratings decline, so that's not it. I have another guy, slightly more power, but solid at everything else - 36 ovr/pot. Yet he's been a 4 WAR player for multiple years in a row. Another one of my guys, same deal, low power, good at specific other skills - he's a 45, and you guessed it, another 4 WAR bat.

The reality is that these players are not the outlier, but rather a significant number of players are significantly over or under rated by ovr/pot because OOTP values specific skills (such as power for hitters) more than others, and the result is that overalls and potentials are simply not accurate.

There might be other ways of showing that the quality of players isn't as high while using specific settings options (ie rating players compared to others), but using ovr/pot isn't the best way, because ovr/pot doesn't accurately reflect the ability of a player.

And for the record I don't use scouting so it's not an issue of a scout not recognizing a player's talent. I know showing one example is not "proof", but I'd rather not give away a competitive advantage by sharing which skills are under and over rated by OOTP game engine.

[NotMuchTime]

Originally Posted by ThePretender

I think the bigger issue is just that ovr/pot do a terrible job of giving you a quick view of how talented a player is or will be.I have one player with a 57 overall/potential, who over his career has been - 5.3 WAR, 3 WAR (99 games, 4.5 WAR per 150), 7.9 WAR, 4.2 WAR (injured, prorated to 5 WAR per 150 games), and this year is on pace for 6.8. So why does Ovr/Pot rate a guy who over his career has averaged 6.2 WAR per 150 games as a 57? Because he doesn't hit for power.

Despite the fact a slightly above average defensive 2B owns a 131 wRC+ for his career, he's a 57 ovr/pot because he hits for no power. When I think 57 ovr/pot, I don't think of guys who average 6 WAR per 150 games. There's no ratings decline, so that's not it. I have another guy, slightly more power, but solid at everything else - 36 ovr/pot. Yet he's been a 4 WAR player for multiple years in a row. Another one of my guys, same deal, low power, good at specific other skills - he's a 45, and you guessed it, another 4 WAR bat.

The reality is that these players are not the outlier, but rather a significant number of players are significantly over or under rated by ovr/pot because OOTP values specific skills (such as power for hitters) more than others, and the result is that overalls and potentials are simply not accurate.

There might be other ways of showing that the quality of players isn't as high while using specific settings options (ie rating players compared to others), but using ovr/pot isn't the best way, because ovr/pot doesn't accurately reflect the ability of a player.

And for the record I don't use scouting so it's not an issue of a scout not recognizing a player's talent. I know showing one example is not "proof", but I'd rather not give away a competitive advantage by sharing which skills are under and over rated by OOTP game engine.

I agree 100%... The whole ratings system needs overhauled. I'm very familiar with the scouting/ratings system and it's actually very simplistic and been around a long time.

Is the player a bench player (40); a platoon player (45); an average everyday player (50); an above average everyday player(55); etc.... Is the pitcher a Closer (rarely above a 60); a number #1 starter, a number #2 starter; etc... And yes, the OVR is designed to be a prediction of WAR. An average player (50) should be a 2/3 WAR player.

This has nothing to do with his individual ratings. A utility infielder may have 'tool' ratings that look exactly like a starters ratings, but the OVR reflects the difference. Maybe a good scout rates the utility player a 40/45, while a bad scout rates him a 50. A team tries to make him an everyday player only to find out he isn't going to cut it.

I could go on and on about this. It's what ruins the game for me. It effects the AI, especially trades. Two 40s isn't better than one 60 in a trade (unless it's a salary dump).

[injury log]

It's because your league is so small that you're seeing a distribution like that. I don't know if you're ranking by position or against all players, but if you imagine a system by position where the top 5% is blue, the next 20% is green, the next 50% is yellow and so on, well if you have 6 players at each position, you don't have anyone in the top 5%.

I'm making up numbers above, but the system does work by ranking players against each other, and any time I've had a very small league, I've noticed the distribution of the OVR ratings is unusual.

[NoOne]

for what you are trying to do, turning on AI Evaluation is making it more difficult to really analyze distribution... same reason you turned on 100% accurate ratings... don't use stats in overall/pot...


if you have real players, don't waste your time... fictional will be the only thing you can manipulate, so don't waste time studing real players.

also, don't focus on overall/potential... it's not a true measure of ability, relative to the game.

individual ratings - power, control, contact... these are all created in a very even distribution curve, then how they have randomly occured together is the results and it may ebb and flow a bit, but it will be similar to a belll curve.

you can shift this curve left and right by adjusting PCM - as you pinch it toward the min/max allowed, it will get flatter, i'd bet. you can adjust resulting stats with LTM.


overall / pot = a 45 could outpefrom a 55 - given an infinite sample. you could argue that he is actually a "Better" rated player even though overall/pot does not say the same thing as what was proven to be true. overall and potential are not absolute, they are derivative. since they are derived from stats in an imperfect way, you should assume overall/pot is an absolute sclae... i.e. 36 might be worse than 25, so don't assume the opposite either.