Do Ratings Follow Stats or Vice Versa?

[Mr. Marlin]

When I play OOTP for real I usually play stats only. So I never know what a player rating actually is. I rely on performance and scouting report text when making personnel decisions. Please note I am not making ANY claims this play style is better, harder, etc. Lets not open THAT box again.


Anyway the point of this post is to see what everyone else's experience has been. Maybe even Matt will comment. Does a players performance lead his ratings or do ratings lead performance?


I did a little experiment yesterday. Before heading to work a made a fictional test league with the modern MLB set up. I picked the 30 best hitters and the 20 best pitchers and simmed forward 10 years. Then tracked their stats and ratings over the 10 seasons. It seems to me statistical improvement or decline LEADS rating changes. Pretty consistently from this little experiment. So the ratings displayed on the player's page are based on what they have already done, not what they are expected to do.


This is just a point of curiosity for me but I'd like to see if anyone else has noticed. If this were a fact (not stating it is from one data point) would that change how you weight ratings when you play?

[Bobfather]

I see the opposite. I play out my games. This past season I had a 70/70 pitcher who had sn excellent year. This off season he is now a 60/60 pitcher. Just one of many examples per season for me. This is just info.

[Brad K]

Originally Posted by Mr. Marlin

Anyway the point of this post is to see what everyone else's experience has been. Maybe even Matt will comment. Does a players performance lead his ratings or do ratings lead performance?

Ratings are based on real life stats or what the development engine does, not on stats produced by the game. If that were so it would be like lifting yourself up by your own bootstraps or digging yourself into a deeper hole.

However the effects of TCR could easily create the impression that you have about stats leading ratings.

[Bobfather]

Originally Posted by Brad K

Ratings are based on real life stats or what the development engine does, not on stats produced by the game. If that were so it would be like lifting yourself up by your own bootstraps or digging yourself into a deeper hole.

However the effects of TCR could easily create the impression that you have about stats leading ratings.

Please disregard my previous post. I should have known this, but in hurry to help I messed up. Sorry.

[Rain King]

Originally Posted by Mr. Marlin

When I play OOTP for real I usually play stats only. So I never know what a player rating actually is. I rely on performance and scouting report text when making personnel decisions. Please note I am not making ANY claims this play style is better, harder, etc. Lets not open THAT box again.


Anyway the point of this post is to see what everyone else's experience has been. Maybe even Matt will comment. Does a players performance lead his ratings or do ratings lead performance?


I did a little experiment yesterday. Before heading to work a made a fictional test league with the modern MLB set up. I picked the 30 best hitters and the 20 best pitchers and simmed forward 10 years. Then tracked their stats and ratings over the 10 seasons. It seems to me statistical improvement or decline LEADS rating changes. Pretty consistently from this little experiment. So the ratings displayed on the player's page are based on what they have already done, not what they are expected to do.


This is just a point of curiosity for me but I'd like to see if anyone else has noticed. If this were a fact (not stating it is from one data point) would that change how you weight ratings when you play?

This is a good and interesting question. Neither of your proposed premises are quite correct, however.

The game uses an underlying rating scale of 0-550 for most player qualities. Those ratings are what are used to determine performance.

However, you do not see those ratings. Instead, the ratings that are displayed to you take those underlying ratings and apply a number of adjustments.

- Rating Scale: The underlying ratings are modified into the scale you choose (i.e. 1-100, 20-80, 1-10, etc.)

- Relative to League: The displayed ratings are also adjusted based on the average rating within the league (i.e. ratings are going to translate higher if comparing the player relative to AAA vs. MLB averages).

- Scouting Accuracy: The displayed ratings are modified further to represent your scout or OSA. Scouts have their own qualities and accuracy levels that come into play. Scouting reports also only update occasionally (the frequency is adjustable), so you may potentially be seeing a report that was generated before an underlying rating change occurred.

- Evaluation AI Settings: You can choose how much weight both ratings and stats (across the most recent 3 seasons) have towards the evaluation. In this way, yes, previous stats may potentially be affecting the ratings you are seeing (but not the underlying ratings that actually determine performance).

[crashsports1]

To that end, do you find it optimal to keep the Eval A.I. ratings as preset, or modify them?

Originally Posted by Rain King

This is a good and interesting question. Neither of your proposed premises are quite correct, however.

The game uses an underlying rating scale of 0-550 for most player qualities. Those ratings are what are used to determine performance.

However, you do not see those ratings. Instead, the ratings that are displayed to you take those underlying ratings and apply a number of adjustments.

- Rating Scale: The underlying ratings are modified into the scale you choose (i.e. 1-100, 20-80, 1-10, etc.)

- Relative to League: The displayed ratings are also adjusted based on the average rating within the league (i.e. ratings are going to translate higher if comparing the player relative to AAA vs. MLB averages).

- Scouting Accuracy: The displayed ratings are modified further to represent your scout or OSA. Scouts have their own qualities and accuracy levels that come into play. Scouting reports also only update occasionally (the frequency is adjustable), so you may potentially be seeing a report that was generated before an underlying rating change occurred.

- Evaluation AI Settings: You can choose how much weight both ratings and stats (across the most recent 3 seasons) have towards the evaluation. In this way, yes, previous stats may potentially be affecting the ratings you are seeing (but not the underlying ratings that actually determine performance).

[Rain King]

Originally Posted by crashsports1

To that end, do you find it optimal to keep the Eval A.I. ratings as preset, or modify them?

It depends what you are looking for from your simulation. It also depends on what the defaults are.

For example, the Standard game defaults to much higher Ratings weight than a Fictional game.

This is because the dev/roster team is trying to assert a little more control on how the AI initially handles the opening day rosters. Users tend to complain if the AI decides to arrange things a lot differently than the real life organizations from the outset.

Some like high Ratings evaluation because it makes the AI a bit "smarter". They will tend to see underlying rating changes a quicker (if their scouting is accurate) than if they are relying more on recent stats.

However, IMO, that also leads to some odd moves. The AI might release a veteran player when they still look like they are performing well or call up a player who doesn't necessarily have good minor league numbers, etc.

I'm not going for maximum AI efficiency in my saves and play a lot in "commissioner/spectator mode" so I like something more along the lines of 35/35/20/10. I think I am probably in the minority there though.

[Mr. Marlin]

Originally Posted by Rain King

This is a good and interesting question. Neither of your proposed premises are quite correct, however.

The game uses an underlying rating scale of 0-550 for most player qualities. Those ratings are what are used to determine performance.

However, you do not see those ratings. Instead, the ratings that are displayed to you take those underlying ratings and apply a number of adjustments.

- Rating Scale: The underlying ratings are modified into the scale you choose (i.e. 1-100, 20-80, 1-10, etc.)

- Relative to League: The displayed ratings are also adjusted based on the average rating within the league (i.e. ratings are going to translate higher if comparing the player relative to AAA vs. MLB averages).

- Scouting Accuracy: The displayed ratings are modified further to represent your scout or OSA. Scouts have their own qualities and accuracy levels that come into play. Scouting reports also only update occasionally (the frequency is adjustable), so you may potentially be seeing a report that was generated before an underlying rating change occurred.

- Evaluation AI Settings: You can choose how much weight both ratings and stats (across the most recent 3 seasons) have towards the evaluation. In this way, yes, previous stats may potentially be affecting the ratings you are seeing (but not the underlying ratings that actually determine performance).

That would explain why the AI makes some decisions that are baffling based on they way I look at players.

[cheddah]

It seems to me statistical improvement or decline LEADS rating changes.

What methodology did you use? Because this would be correct ONLY if talking about the Overall/Potential ratings. Those meta ratings use statistical performance as criteria in the default settings (you can adjust the weights of stats and ratings to your liking). In other words, you have a down year and your OVR will likely fall. But that is also why OVR/POT aren't a reliable method of forecasting future performance and a more detailed analysis of all ratings is necessary before assembling a team.

The individual component ratings (like Stuff, Power, Control etc.) do not use stats at all. They are the product of the player generation and game engine and are used in the determination of game outcomes, ie. these ratings determine stats. If you look at a significant enough sample, you can see this, with very normal statistical variance giving you the occasional outliers.