A Different Perspective On Ratings And League Totals Modifiers

[Brad K]

Summary: A Different Perspective On Ratings And League Totals Modifiers

(Full version with explanatory comments in a later post.)


The interaction of historic ratings and historic League Totals Modifiers does not result in historic league output.

Auto-calc tests the historic modifiers against the league and adjusts the modifiers to produce historic output.

Auto-calc gives the correct output for a replay league. In a random debut, alternate history, or fictional league using auto-calc does not give the correct output. Historic output is not correct for leagues with talent different from historic.

Using pre-calc historic modifiers for such leagues is not a good solution because the errors in the interaction of ratings and historic league totals remain.

The best answer for a league with different talent is to use auto-calc modifiers created by testing against a replay league. The errors will be gone and the talent differences will be present.

OOTP should provide precalculated auto-calc modifiers for each year. The output of each type of league will be correct.

Presently people with non replay leagues can duplicate the results of this proposal by running auto-calc against a replay league and using those modifiers for their non replay league.

[Brad K]

Note how far the two auto-calc tests are from pre-calc yet how close they are to each other. Most of the difference in the sum of lines 14-31 is due to OF assists.

The spreadsheet file is available to those who want it. PM me your e mail to obtain it. It is in Open Document format, readable by Open Office, Libre Office, and MS Office 2010 and newer.

[Brad K]

Full Version: A Different Perspective On Ratings And League Totals Modifiers


I've thought about the issues of player ratings, League Totals Modifiers, pre-calc, and auto-calc for several years. Recently people have posted on these subjects and that has prompted me to pull my thoughts together for this post.

When the editor was changed to show expected stats for modern day, talent, and the current year of the game, I realized my previous attitude that ratings should work without modifiers was wrong. The ratings are talent but the output of that talent is different in 1915 than it is in 2015. So the LTMs take care of that adjustment.

Except (and it's not unusual) historic inputs combined with other historic inputs may not result in historic outputs. I encountered this a decade ago when I was working with a group of people to modify a war game's data files. Some others insisted on historic inputs convinced that since they were historic they must work despite ample evidence to the contrary.

Fortunately OOTP recognized that in this case historic inputs don't give historic outputs and acted upon it. That action is auto-calc. That is the process where the game tests the pre-calc (historic) modifiers against the league and adjusts them so the league will produce output close to historic. In other words, the routine corrects the errors in the interaction of the historic LTMs with historic ratings.

This is a fine solution in replay league where players are on their historic teams and historic lineups are used. However many people play alternative history where players end up on different teams and in different parks, the lineups are different, and the actual player selection is different due to some players having shorter careers due to injury or (if retire according to history is off) players playing past their retirement dates.

Additional factors that make the talent in the league different from historic is having Talent Change Randomness on, Development on, using the Development Lab, or playing with historic minors. Then there's random debut and fictional players which can cause further deviations from historic talent.

When auto-calc is run on such leagues, it not only corrects the errors of the interaction between the ratings and historic LTMs, it obliterates the difference in talent the league has from historic. Regardless of differences in talent, auto-calc aims for the league to duplicate historic output.

In the past the concern about leagues with different talent being hammered into historic output was dismissed with the solution "well just use the pre-calc modifiers". But it's known pre-calc modifiers don't work on replay leagues and so they certainly don't work on any league. That leaves people running leagues with different talent a choice between unrealistic output due to the errors or unrealistic output due to having their league hammered into historic output regardless of talent differences.

But there's a question. How many times has auto-calc been run on replay leagues? Tens of thousands? Hundreds of thousands? Millions? In any case way too many times. Since the replay league for a given year never changes, then close to the same auto-calc LTMs are created each time.

Rather than each human running them for each league, the game could come with precalculated auto-calc results based on a few hundred tests by OOTP. (I need a better term for precalculated auto-calc results!) That's what the replay league people are getting anyway by running their own auto-calc but without the benefit of a large number of samples OOTP could provide.

This has an interesting and beneficial effect on another area. If that same set of LTMs is used with an alternate history league, a random debut league, or a fictional league, it allows that league to function as it would in the environment of the year selected. The precalculated auto-calc LTMs remove the errors in the interaction between historic LTMs and historic ratings while allowing the league's talent differences to be shown in the output.

I suppose there are some people who want their league with different talent to perform as if it had historic talent. But it seems to me that given some thought many would prefer it perform according to its talent within the historic environment. That environment is provided the precalculated auto-calc LTMs described which are the result of testing against a replay league.

So what I suggest is that OOTP provide precalculated auto-calc LTMs. It will work better than the current method with replay leagues. People with alternate history, random debut, or fictional leagues can use it to remove the errors but allow their league to show its talent differences. And if anyone wants a league with different talent to perform as historic, they can run the auto-calc routine on their league.

If this is done there needs to be another change, a notice on the statistical accuracy tab so people have realistic expectations. Something like "It should not be expected that leagues with talent different than historic produce historic results."

Presently, people can create their own LTMs to use in non replay leagues by running auto-calc on a replay league of the same year and transferring the data into their non replay save. Even better, run the auto-calc several times and average the data.

[Larryk007]

Thanks for a deep dive into a great topic.
Do you normally just let the modifiers stay and hope the statistics match by the end of the season?

In some past fictional replays certain statistics would not track, like too many homeruns by July. So in my game I would manually adjust the homeruns modifier down and it seems to work, less homeruns by the end of season matching the expect total year result.

[Brad K]

I'm currently using LTMs from running auto-calc on a 1974 replay league in a alternative history league in the same year. I don't expect the results to match historic because the talent is different. I expect differences to be plausible.

For example, scoring is down a bit. That's plausibly realistic. I'm running historic minors. As of a year ago (game time) 22% of the league's SPs and 41% of the RPs were career minor leaguers. Those percentages probably still apply a year later. That means those career minor leaguers are rated higher than the pitchers who historically had those jobs. So the league has better pitching than historical and lower scoring should be expected.

I happen to think OOTP is rating career minor league pitchers too high but that's not the issue here. The point is they're in my league and affecting its output as a change in talent should.

Here's the thread on pitchers if you're interested. https://forums.ootpdevelopments.com/...d.php?t=360806

[Brad K]

I have discovered that I should have included specific examples of the effects of the proposal compared with running auto-calc on non replay leagues.

Specific effects in non-replay games

Auto-Calc / Proposal

If DH is enabled in a non DH league will scoring increase?
No / Yes

If DH is disabled in a DH league will scoring decrease?
No / Yes

If a big home run hitter has a season ending injury in Spring Training will other players hit his HRs?
Yes / No

If multiple players gain defense ratings through the development lab will league defense improve?
No / Yes

If a league has better hitting than historic will the league hit better than historic?
No / Yes

In a random debut league for 1927 that by chance has Ruth, McCovey, Aaron, Judge, and Sosa, will Ruth's home run total be reduced?
Yes / No

[Garlon]

Remember that auto-calc targets league total ratios.

1. No

2. No

3. If auto-canc happens before spring training, yes. If it happens after, no.

4. BABIP, error rate, and other defensive outcomes are targeted by the modifiers. You will not alter league rates by changing the average defensive ratings across the league. You are just creating a new defensive no dice average rating.

5. No, again you are just changing what the average batting ratings across the league. Average still equals league average.

6. Yes, there is a finite amount of HR to go around based on the HR rate of that season. Not everyone can hit 50 HRs for the season and still maintain the historical rate of the season.


You should really be always be using auto-calc modifiers. There are special cases for the pre-calc modifiers which were originally made so that exhibition games worked properly from the main menu.

[Brad K]

Originally Posted by Garlon

You should really be always be using auto-calc modifiers. There are special cases for the pre-calc modifiers which were originally made so that exhibition games worked properly from the main menu.

That was true when the only other choice was pre-calc which gives terrible results. I have provided a third choice.

Why do you want every league to perform as a league with historical talent? Why do you want the Dev Lab to do nothing? People put a lot of effort into creating that and lots of people like using it.

[David Watts]

Originally Posted by Garlon

Remember that auto-calc targets league total ratios.

1. No

2. No

3. If auto-canc happens before spring training, yes. If it happens after, no.

4. BABIP, error rate, and other defensive outcomes are targeted by the modifiers. You will not alter league rates by changing the average defensive ratings across the league. You are just creating a new defensive no dice average rating.

5. No, again you are just changing what the average batting ratings across the league. Average still equals league average.

6. Yes, there is a finite amount of HR to go around based on the HR rate of that season. Not everyone can hit 50 HRs for the season and still maintain the historical rate of the season.


You should really be always be using auto-calc modifiers. There are special cases for the pre-calc modifiers which were originally made so that exhibition games worked properly from the main menu.

In regards to #6, but let's take random debut out of the picture. If one creates 1921 using historical minors and eliminates the color barrier, teams will immediately begin signing Negro League players. Therefore, there will be a significant change in the overall talent pool. Why should this have any effect at all on the number of home runs Babe Ruth hits? Shouldn't the results lean towards the league hitting more home runs due to the added talent, instead of the same amount spread over more players? Since this is more of a whatif type scenario, I just don't see why Ruth's chances to reach 59 home runs should diminish. This is where I wouldn't mind having a third option that lets the game slide away somewhat from the league totals goal post.

In terms of random debut, I'm not sure there is a solution and the way the game works now, might be the way it needs to work, in order for the historical period to maintain its relevance.

[LansdowneSt]

Presumably the pitching is better too. Dave, you focused on there being more hitting talent coming in because NeL hitter X entered. So did NeL pitcher Y. So, whomever the 4th pitcher was for the Washington Senators, he just lost a job and Bill Foster is on the mound.

[Brad K]

Originally Posted by LansdowneSt

Presumably the pitching is better too. Dave, you focused on there being more hitting talent coming in because NeL hitter X entered. So did NeL pitcher Y. So, whomever the 4th pitcher was for the Washington Senators, he just lost a job and Bill Foster is on the mound.

It's virtually impossible that there would be an equal increase in talent for both hitting and pitching. Talent will be different and so the league should have different output.

Using the proposed method allows the league to have output that shows the talent difference. Running auto-calc on it does not.

An example is random debut where it can be can be easily seen when there is a influx of exceptionally skilled hitters or pitchers. The idea of Ruth having to share his HRs (or Cobb having to share is SBs, etc) isn't logical.

Another place differing talent can be seen is when retire according to history and miss years are turned off. Some people don't want to see Ted Williams spending time in the Marines, Koufax retiring early, or Ken Hubbs, Thurman Munson, Roberto Clemente and Don Wilson dying while active players.

[Brad K]

Originally Posted by Garlon

There are special cases for the pre-calc modifiers which were originally made so that exhibition games worked properly from the main menu.

This is interesting. My assumption was pre-calc (historic) modifiers came first and were found not to work with the ratings and game engine and that led to the development of auto-calc. What was unacceptable about how auto-calc worked with exhibition games?

[Garlon]

The game just does not go through the process of doing an entire auto-calc for exhibition games because it is only loading two teams instead of an entire league.

Markus put together a file that had the average of 100 or 10000 auto-calcs for each historical season with the default strategy settings for each season so that OOTP could just grab that data set and load it for exhibition games. However, that entire file needs to be rebuilt for every edition of the game because there are subtle adjustments to the engine and new modifiers added.

[LansdowneSt]

I feel, Brad K, that this anti-LTM post is just an annual rite of summer for you.

I think I'll just read last year's post about it as everyone is coming on and saying the same things. Enjoy OOTP 26 (or not). if you don't like the engine, no one is going to convince you otherwise. I've seen this movie before.

[Brad K]

Originally Posted by LansdowneSt

I feel, Brad K, that this anti-LTM post is just an annual rite of summer for you.

I think I'll just read last year's post about it as everyone is coming on and saying the same things. Enjoy OOTP 26 (or not). if you don't like the engine, no one is going to convince you otherwise. I've seen this movie before.

I don't know why you think I'm posting anti -LTMs. I'm promoting the use of LTMs. Please see the second paragraph of post #3. Something DID change my previous view.

I'm sorry there's a misunderstanding about what I'm proposing.

[Brad K]

Originally Posted by Garlon

The game just does not go through the process of doing an entire auto-calc for exhibition games because it is only loading two teams instead of an entire league.

Markus put together a file that had the average of 100 or 10000 auto-calcs for each historical season with the default strategy settings for each season so that OOTP could just grab that data set and load it for exhibition games. However, that entire file needs to be rebuilt for every edition of the game because there are subtle adjustments to the engine and new modifiers added.

Hmmm.... are you sure about how the pre-calc was created? Please see the second post for the differences between the LTMs I got and what Markus got using what is described as the same method.

But I'm glad OOTP saw the advantage of having a file the game could just grab and use. That's what I'm promoting.

[Brad K]

Originally Posted by Garlon

The game just does not go through the process of doing an entire auto-calc for exhibition games because it is only loading two teams instead of an entire league.

I expect some people might be puzzled between recommendation to always run auto-calc and not running it in this situation. It's probably wise to explain it.

[Garlon]

The game provides no option to use auto-calc when setting up an exhibition game from the main menu. You also do not get to change the league strategy settings either. Since the league strategy settings are defaults, using a pre-made file based on the defaults for the season averaged across 100 or more trials is fine.

[Brad K]

Originally Posted by Garlon

The game provides no option to use auto-calc when setting up an exhibition game from the main menu. You also do not get to change the league strategy settings either. Since the league strategy settings are defaults, using a pre-made file based on the defaults for the season averaged across 100 or more trials is fine.

Yes, of course, but that "is fine" file is not recommended for use for any other type of play.

I suppose some people have realized that auto-calc won't work with an exhibition because the talent of the teams doesn't match replay league talent. Talent difference has been mentioned here before.

However the file I propose would work great.

[Brad K]

If a light switch in the kitchen is broken, we can conclude it won't work if we install it in the living room.

Since the pre-calc file produces incorrect output in a replay league, we can conclude it is not "fine" for a tournament. Or with any league.

The file I propose works with a replay league and is appropriate for all leagues,