Interaction Of Settings With Auto-Calc

[Brad K]

This is not an anti-auto-calc post. I want people to understand what it does so they can make the choice between it and pre-calc that is best for them.

The data below is from 1978. There are 33 starting pitchers in the study. Starting pitchers who had a relief appearance in either real life or the test are not included. There is significant variance in the tests done with auto-calc. However the data is much more usable than the Complete Game data.

The variability of the output can be seen in the fact that Very High stamina resulted in fewer IPs than Normal stamina. Also, both are higher than historic IPs. The setting that produced output closest to historic was Very Low, missing by only 28 IPs.

Of note is that the biggest variance - almost twice any other - is between auto-calc Very Low and pre-calc Very Low. This is well outside the other variances and is evidence auto-calc is over riding the League Strategy setting. And it must as typical SP stamina modifiers are Very Low 1.119, Normal .917. Very High .850.

The full data by pitcher is included for those who might be interested in more than the summary. If anyone wants the save files to examine them, contact me.

[Brad K]

Previous posts to increase understanding of the effects of auto-calc have shown that auto=calc alters the effect of some strategy settings and completely over rides some others. This post addresses its affect on three year and five year recalc.

It's a reasonable assumption that using three year or five year recalc on player ratings provides a smoothing effect. That is the function of a moving average. But a league doesn't perform better or worse on its own. It performs better or worse based on how the players perform. Auto-calc over rides the effects of the moving average on players in favor of keeping every league performing as a historical replay league.

A good three year period to consider is 1976 - 1978. During those years average HRs per team was 93, 140, and 114 with a three year average of 116. On one year recalc for 1977, players are rated to produce 140 HRs/Team. On three year recalc they're rated to produce 116 HRs/Team. Yet with auto-calc teams will produce about 140 HRs/Team whether the players are rated for 140 or 116.

The graph below shows that there is no smoothing effect with three year recalc. The output of both it and one year recalc closely track historic output. Ratings based on a moving average do not deliver what is expected when auto-calc is used.

Presumably auto-calc was introduced to adjust for errors in the data. It does this well. The problem is differences in ratings intended by the user are treated as errors by auto-calc which over rides them.

[Rain King]

Auto-Calc tunes the overall results to the specific season you are simulating if you choose the League Totals for that season. I do not know why you would expect a "smoothing" result just because you are running Recalc over 3 or 5 years.

If you want the statistics to "smooth" across 3 or 5 year periods then you should adjust your League Totals to be an average of those 3 or 5 years and THEN run auto-calc.

[Brad K]

Originally Posted by Rain King

Auto-Calc tunes the overall results to the specific season you are simulating if you choose the League Totals for that season. I do not know why you would expect a "smoothing" result just because you are running Recalc over 3 or 5 years.

If you want the statistics to "smooth" across 3 or 5 year periods then you should adjust your League Totals to be an average of those 3 or 5 years and THEN run auto-calc.

Puzzled a bit by your response. What then is the purpose of 3 year and 5 year?

[Rain King]

Originally Posted by Brad K

Then what is three year and five year recalc supposed to do?

Recalc rates players compared to league average...not compared to specific statistical results.

So, when you run 3 or 5 year recalc you are rating the player compared to league average across those seasons instead of using just 1 season.

Their stats are then produced based on the League Totals/Modifiers you use.

So, if a player averages a Power rating (i.e. home runs per plate appearance) of 1.15 times the league average across the 3-5 years...then they will be given a rating that should produce 15% more home runs than a league average player.

The number of home runs that a league average player will hit is determined by the League Totals/Modifiers.

[Brad K]

Originally Posted by Rain King

The number of home runs that a league average player will hit is determined by the League Totals/Modifiers.

Yes. I agree totally. If players on average have lower ratings due to use of three year or five year auto-calc over rides that and bumps their performance. As a result I don't see benefit to three year and five year. Perhaps you do? Maybe I've overlooked something.

[Rain King]

Originally Posted by Brad K

If players on average have lower ratings due to use of three year or five year auto-calc over rides that and bumps their performance.

I do not understand what you mean by this sentence. Can you elaborate?

[Garlon]

As I mentioned earlier, the strategy settings are still very important when using recalc and autocalc if LTM.

For SP stamina I use Low from 1920-1961 and Very Low from 1962-present.

I previously mentioned that if your SP stamina is set too high forever that the modifier will bottom out and not be able to properly function and you can see that with the 0.850 modifier which is as low as OOTp will generate.

Another issue with the default OOTP settings driving up SP innings is that the stamina modifier by default puts 1/3 of its weight on targeting CG%. Doing that drives up the innings for all SO in the league because OOTP keeps boosting the stamina a LTM to try to get that CG target which is really not that important. This in turn drives down the distribution of IP between SP and RP. I exit the engine file to put 99% of the weight in IP per GS and 1% of the weight on CG percentage so that there is a proper balance in the distribution of innings.

I also set Stealing to Very Often for all eras for career leagues without real lineups and transactions so that individual player SB success rates are closer to their historical values.

[robc]

It seems like the real problem isn't how Autocalc and the settings work, it's that there isn't good documentation for the game so it is difficult for a player to know what should happen. It seems like there should be some type of message that gets displayed if the strategy settings get changed after Autocalc that warns the player of the consequences. Let them do it if they want, but help them understand the effects on the league totals.

Luckily there are people like Garlon, Rain King and others that have a good understanding of how the game works and they are willing to share it with others. I just wish all of this knowledge existed in a reference so customers could understand how the game works. Many times we don't know what we don't know, so we don't even ask the questions and the game doesn't behave how we think it does.

[Brad K]

Originally Posted by Rain King

I do not understand what you mean by this sentence. Can you elaborate?

Yes I can. Autocalc does not change the ratings. It changes how the engine interacts with the ratings.

Whether one year or three year recalc is used, auto-calc produces the same league output even though the ratings are different.

An equal measurement tool is predicted OPS in a modern neutral environment which is listed in the editor. The graphic below shows that players are rated differently with 1 year and 3 year.





Despite these different ratings the league will produce historic output. However auto-calc affects the league as a whole. It doesn't target individual players. All 3 year ratings will get a bump because on average they're lower than what is needed for historic output in 1977.

Note some players ratings are higher on 3 year even though 1977 was a stronger offensive year for the league. The auto-calc boost is applied to these players too who are thus likely to over perform in 1977.

Auto-calc on three year makes the league produce historic output while increasing the chance of outlier performances by individual players.

I have screen shots of the OPS but it seems a bit excessive to put 14 screen shots in a post. Anyone who wants them please contact me. Also, the save files are available upon request.

[Rain King]

I don't see any problem with any of that.

If you are simulating the highest offensive season out of the 3 years involved in Recalc, then players are going to put up better offensive numbers than average across those 3 years. Subsequently, pitchers will put up worse numbers than their average.

You seem to be missing that Recalc and Autocalc are completely separate processes. They are not intertwined. You can use Recalc without using Autocalc and you can use Autocalc without using Recalc.

Also, you seem to be misunderstanding how Recalc works. Hitters do not get better hitter ratings because a season had higher league offense. The ratings are based around the average hitter and a player gets rated by how much better (or worse) they are compared to the average numbers for each season (or set of seasons).

You can completely change the League Totals set for your Autocalc at any point. You can Recalc across 3 seasons and set Autocalc to a completely different environment and then your players are going to play to that environment which might be very different than their real life seasons. This is all customizable.

As I mentioned previously, you can get the results you seem to be expecting...but you would need to change your League Totals to reflect the average across the 3 or 5 years that Recalc is using. I could see the game offering that as an option instead of only having the single season League Totals available to load. I'd probably use that.

[Brad K]

Originally Posted by Rain King

I don't see any problem with any of that.

If you are simulating the highest offensive season out of the 3 years involved in Recalc, then players are going to put up better offensive numbers than average across those 3 years. Subsequently, pitchers will put up worse numbers than their average.

Correct. Again we agree.

So since the league is going to perform to historic, the lower batter ratings with 3 year don't matter. Auto-calc over rides the ratings. Might as well use 1 year, right?

[Rain King]

Originally Posted by Brad K

So since the league is going to perform to historic, the lower batter ratings with 3 year don't matter.

The rating always matter.

Auto-calc over rides the ratings.

This is not an accurate description of what happens. Auto Calc runs simulations to set the modifiers so that the "average" rated player for your situation (i.e. whatever Recalc produced in your case) will put out average stats for your environment (the League Settings you have).

Might as well use 1 year, right?

You can do whatever you want. If someone would prefer to rate the players across 3 or 5 Years, they should use 3 year or 5 year recalc.

I don't understand the need to say what is the "right" way to do it. Everyone has different objectives.

If you want to use 1-Year recalc, use 1-Year recalc.

[Brad K]

Originally Posted by robc

It seems like the real problem isn't how Autocalc and the settings work, it's that there isn't good documentation for the game so it is difficult for a player to know what should happen.

I'm providing documentation with evidence to back it up. For the most part, it's been rejected. Hopefully the evidence will eventually be accepted and action taken. It took a long time for that to happen with my thread that resulted in a new pre-calc file but in the end it was accepted.

[Brad K]

Originally Posted by Rain King

The rating always matter.



This is not an accurate description of what happens. Auto Calc runs simulations to set the modifiers so that the "average" rated player for your situation (i.e. whatever Recalc produced in your case) will put out average stats for your environment (the League Settings you have).



You can do whatever you want. If someone would prefer to rate the players across 3 or 5 Years, they should use 3 year or 5 year recalc.

I don't understand the need to say what is the "right" way to do it. Everyone has different objectives.

If you want to use 1-Year recalc, use 1-Year recalc.

Well, this time we don't agree. A shame to break the streak!

[Rain King]

Originally Posted by Brad K

Well, this time we don't agree. A shame to break the streak!

What is it specifically you disagree with?

(Queue response that goes on some tangent that has nothing to do with my post).

[Brad K]

Previous posts have included evidence that auto-calc over rides the different ratings that result form using 3 year recalc. There is no significant difference from historic league output regardless of which rating calculation method is used.

Actually it doesn't matter how the ratings change. Expand a league when it didn't expand historically and the average MLB player is less capable. Contract a league contrary to history and the average MLB player is more capable. It doesn't matter.

The screen shot below shows three seasons where output hits historic targets. 1974 was run normally. At the end of 1974 the top 32 batters were retired. And the end of 1975 13 additional top batters were retired. This huge change in talent resulted in no change in output. Everything is green. Auto-calc over rides ratings.

To have ratings affect the league output, it is necessary to use the pre-calc file in build 26.4.59. It will work if coaching is disabled.

[Rain King]

Originally Posted by Brad K

The screen shot below shows three seasons where output hits historic targets. 1974 was run normally. At the end of 1974 the top 32 batters were retired. And the end of 1975 13 additional top batters were retired. This huge change in talent resulted in no change in output. Everything is green.

Yes.

That is the entire purpose of Auto Calc.

What that doesn't mean is that "auto calc overrides ratings", nor does it mean "to have ratings affect the league output, it is necessary to use pre-calc". Your applying it to an extreme case does not invalidate it.

Auto Calc uses the ratings that currently exist in your league, runs simulations against your League Totals and adjust Modifiers to get the overall output of those League Totals. That's it. It isn't some big conspiracy. The ratings obviously have to "matter" for that to occur.

I know that you are purposefully misinterpreting all of this because you don't like it. I really wish you would give it a rest and stop trolling or that the mods would ban you for constantly misrepresenting this stuff.

You prefer not to use Auto Calc. Great, don't use Auto Calc.

Auto Calc has it uses for those that want certain results....and does exactly what it is intended to do.

[robc]

Originally Posted by Brad K

Whether one year or three year recalc is used, auto-calc produces the same league output even though the ratings are different.
...
I'm providing documentation with evidence to back it up. For the most part, it's been rejected. Hopefully the evidence will eventually be accepted and action taken. It took a long time for that to happen with my thread that resulted in a new pre-calc file but in the end it was accepted.
...
So since the league is going to perform to historic, the lower batter ratings with 3 year don't matter. Auto-calc over rides the ratings. Might as well use 1 year, right?

First, the other guys know a lot more about this than I do, so if I misstate something I apologize. It sounds like you have put a lot of work into this, but I think your understanding of how these features work are off. I'll try to word it in a different way than the others to see if it resonates with you.

First there is the season environment you want the results to conform too, say 1975 but it could be anything. Then there is the recalc method (if you choose to use recalc). This only controls what stats are used to produce the players' ratings - the current season, the current season + the two surrounding years, or the current season + the surrounding 4 years). So of course some players will become more productive and some less depending on what recalc period you choose.

The following might not be 100% accurate on how it works, but I think it is conceptually. For power, think of the ratings as a representation of the share of how many of those HRs a particular player gets. If they are playing in an environment where the league hit 3000 HRs, they will get a share of those. It if were 1000 HRs, they would get a share of those so even with the same power rating they would hit less HRs. You probably knew that stuff. This is why you can take players out of the period they played in and create a league for an entirely different period. They are rated compared to other players in their period, but then those ratings are applied to another league environment.

Autocalc is only trying to massage those league totals in OOTP to match those of the chosen year it's trying to emulate. The game still uses the same player ratings, so proportionally the players' performance should be the same.

So based on your comments I quoted above it doesn't seem like you understood the point of recalc and how it works. I'm not sure if my explanation is the best way to think about it, but hopefully it makes sense.

[Brad K]

Originally Posted by robc

First, the other guys know a lot more about this than I do, so if I misstate something I apologize. It sounds like you have put a lot of work into this, but I think your understanding of how these features work are off. I'll try to word it in a different way than the others to see if it resonates with you.

First there is the season environment you want the results to conform too, say 1975 but it could be anything. Then there is the recalc method (if you choose to use recalc). This only controls what stats are used to produce the players' ratings - the current season, the current season + the two surrounding years, or the current season + the surrounding 4 years). So of course some players will become more productive and some less depending on what recalc period you choose.

The following might not be 100% accurate on how it works, but I think it is conceptually. For power, think of the ratings as a representation of the share of how many of those HRs a particular player gets. If they are playing in an environment where the league hit 3000 HRs, they will get a share of those. It if were 1000 HRs, they would get a share of those so even with the same power rating they would hit less HRs. You probably knew that stuff. This is why you can take players out of the period they played in and create a league for an entirely different period. They are rated compared to other players in their period, but then those ratings are applied to another league environment.

Autocalc is only trying to massage those league totals in OOTP to match those of the chosen year it's trying to emulate. The game still uses the same player ratings, so proportionally the players' performance should be the same.

So based on your comments I quoted above it doesn't seem like you understood the point of recalc and how it works. I'm not sure if my explanation is the best way to think about it, but hopefully it makes sense.

Everyone agrees that auto-calc controls the game output. No dispute there. That idea and the idea ratings matter are mutually exclusive.

Then there's the statement made many times on the board by many people: There are only so many home runs to go around. That means auto-calc, not ratings, control output.

I have stated my views and provided proof. Those who say I am wrong have provided no proof.