Interaction Of Settings With Auto-Calc

[Garlon]

This is a sample of the 1939 season from a current game I am playing (started in 1871 using 5yr recalc) using auto-calc and the results vs the real values. There are many things that auto-calc models, some directly and others indirectly.

1939

Stat: Real vs OOTP

BA: .275 vs .274

OBP: .344 vs .340

SLG: .397 vs .393

OPS: .740 vs .733

BABIP: .289 vs .289

2B per Non-HR Hit: .188 vs .189

3B per Non-HR Hit: .0417 vs .0401

HR/AB: .01702 vs .01657

BB/AB: .09988 vs .09903

K/AB: .10035 vs .10139

R/27 outs: 4.89 vs 4.88

SB Attempts per Reached 1B: .0626 vs .0615

SB Success: .594 vs .591

SF per IPout - K: .0119 vs .0115

Sac per IPout - K: .0247 vs .0241

HBP/AB: .0041 vs .0042



IPouts per GS: 20.2 vs 19.8

GO%: 49.2% vs 48.0%

WP/IPouts: .0067 vs .0074

Balks/IPouts: .0007 vs .0009

ERA: 4.27 vs 4.22

WHIP: 1.46 vs 1.47

SHO%: 10.9% vs 11.5%



League FLD: .971 vs .972

PB/IPouts: .0030 vs .0026

DP per Reached 1B: .09314 vs .08724

GIDP/DP: .812 vs .856

OF PO per IPout - K: .298 vs .292

OF AO percentage: 59.5% vs 60.5%

C FLD: .981 vs .993

1B FLD: .989 vs .989

2B FLD: .966 vs .968

3B FLD: .942 vs .935

SS FLD: .951 vs .951

LF FLD: .970 vs .970

CF FLD: .976 vs .974

RF FLD: .967 vs .974

OF FLD: .971 vs .973

[pstrickert]

Those are pretty fabulous results! I doubt any other baseball sim can match them at this level of detail.

[Brad K]

There is no dispute that auto-calc produces historic output.

[WhiskyTango]

Originally Posted by Brad K

It would have to figure the plusses and minuses from historic for every player and then adjust the development for a zero sum. They get the same result from using auto-calc without much effort.

This seems to be what I'm finding. I run auto-calc on opening day, and produce green and orange.

It may be that there are limits to how much adjustment auto-calc can make but I don't know if there are. If you get historic results with auto-calc in a league with highly diluted talent and atypical player profiles, then if here's a limit you haven't reached it.

Ditto. I don't know how this factors into the discussion in this thread, but I'd say auto-calc seems able to consistently produce the stats that look era-ish to me. Which is pretty impressive given the non-historical demands.

Concerning the items you have unchecked on the Historical tab, I don't see any that I can imagine prevent auto-calc from controlling the league output to match historic league output.

Good to hear. Thanks.

[Brad K]

Originally Posted by WhiskyTango

This seems to be what I'm finding. I run auto-calc on opening day, and produce green and orange.


Ditto. I don't know how this factors into the discussion in this thread, but I'd say auto-calc seems able to consistently produce the stats that look era-ish to me. Which is pretty impressive given the non-historical demands.

You have the most atypical league I can imagine. Auto-calc is very powerful.

[Brad K]

Originally Posted by WhiskyTango

This seems to be what I'm finding. I run auto-calc on opening day, and produce green and orange.


Ditto. I don't know how this factors into the discussion in this thread, but I'd say auto-calc seems able to consistently produce the stats that look era-ish to me. Which is pretty impressive given the non-historical demands.

It's definitely part of the discussion. I thought about your league some more... a third of the players who spent most of their careers in the minors... and you still didn't find something auto-calc can't overcome. With that kind of power, it's trivial for it to overcome things like a change in strategy or slightly askew data.

[NoOne]

If you think it works wonky it's more likely a misunderstanding of what the actual cause and effect is of each - autocalculate and re-calc. In some ways i do think it works wonky, but not because of an error but just a difference of opinion of how the model implements changes like new historic eras and such... they tooka lazy shortcut, imo. It could have been callibrated not to involve shifitng the meaning of ratings, which is what autocalculate resutls in. it changes what to expect from the same ratings, which boggles my mind.

Throw out how you want the game to work and just focus on how it works. Deductive reasoning, not inductive will lead you to the promised land LOL


Autocalc - used when league totals change, like a new historic era. Used when player talent distribution changes -- more players created, historic player creation modifier changes.. anything that cauases a new distribution curve of individual player talent is a great reason to use the autocalulate function.

Using autocalc all the time is generally a bad idea, but if it's what you want to do, do it.

You are basically changing what ratings mean year to year by doing so. 80power might be ~40hr potenetial one year and 30hr the next (that'd take a pretty drastic shift in player disitrubiton,but the concept is true). That to me seems [ridiculous], but it also is how they chose to deal with shifting historical eras and expected results in those eras. This would be better handled by PCM, imo, but at this point re-working that isn't an option, lol. RAtings should cause outcomes, because that's their function.. not "levers" unrelated to causality.

the only time i can see how it makes sense is for something like equipment changes drastically changing fielding percents, or lowering the mound, or a more tightly wound ball. That would make sense for a blanket autocalculate of modifiers. The players aren't different magically overnight, but it mimics a change to balance of competition that acts like a rising tide (or ebbing). When they lowered the mound the guy benching 300 all of a sudden has greater leverage and hits more hr. His "power rating" didn't change. He didn't change. The context changed.

Histroric play context has a lot more reason to use autocalculate, but try to use it minimally, and if you want to tweak a specific outcome, do it by hand after you see the results from the autocalculate. I gaurantee you'll like the results more this way. You get a healthier amount of volatility while still anchored to some reasonable 'median' over time.

Understand thes modifiers merely shiftportions of the same pie. if you reduce hits, that lowers BA, but it also reduces 2b, 3b and hr in proportion to their league total ratios. Increasing walks will reduce AB, which wil reudce hits, 2b, 3b, hr and so on... adding hr will drag BABIP down a bit. (if i recall the BABIP math.. hr are deducted from the numerator portion, right? becaue it technically is not in play). Lots of interrelationships between the modifiers.. squeeze one and something pops out elsewhere, lol. common sense will help you here.

My biggest take for autocalc is -- either focus on keeping league totals in a particular range (preferably a wider range than you expect) or do you want to focus on individual results - season and career numbers looking like what you want. I think the latter is the better way to go. Also, tends to be slightly elevated offensive environment, and that create greater stratification... there'll be more distinct tiers of expected results... the 65power guy won't as often put up numbers like an 80 power guy in a small sample.

e.g. i get my hits modifier and such honed to make sure some players do reach 3000 hits but it's a similar to what has occured in teh MLB to this point (the target being personal preference of course). I do that for hr, 3b, all of it... but if you do that your league wide stats are very unlikely to be historically accurate, fyi. In my experience they'll be a bit elevated if trying to make a 100+-year league leaderboard look similar to MLB's records. It really comes down to what you care about more as far as how you use autocalculate and the modifiers.

The above still stands true.. most effective use is for dealing with new historic era settings, shifts in PCM, number of players created, shifting from seed players to fictionally created or shifts in how historical players evolve thorugh the years blah blah blah that type of stuff. How you use it to tweak what you see above and beyond that is more about personal preferences.


Re-calc - this will impact overall league numbers that result less than you think. This is more about individual player performance. Therefore, more likely to impact the young and the aging more than those in their prime with consistent ratings, for obvious reasons.

A guy that is heading downhill may show degredation sooner with a 3-year recalc rather than a 1 year re-calc. A young player on upswing will see improvments sooner in a 3-year recalc... but for both over that 3-year period it will mostly average out the same way.

The re-calc setting choices will probably have little to no effect on yearly results as the small sample graphs showed above. mathematically expected.

to look at the impact of this settings, would be best to review individual players old and young. to have any confidence in the numbers you see, you may need several replays of any 1 season.

[Rain King]

I don't appreciate the usage of the "R" word.

[Brad K]

Originally Posted by Garlon

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.

I posted about the problem with high SB count players having success percentages lower than historic in 2020. The post concerned OOTP 21 and is located HERE. Lukas posted saying the problem would be looked at. You posted the solution you've posted above.

Some time after that - I don't know when - the problem was fixed by OOTP. So there is no need for you to apply a league strategy setting solution as there is no longer a problem.

As can be in the screen shots below, OOTP 26 functions correctly with or without your suggested strategy settings. The second screen shot shows your settings didn't have an effect when used with OOTP 21. As I have been saying, auto-calc negates the effects of League Strategy settings.

[Garlon]

I made a suggestion for career leagues not replay leagues. I never said the same strategy settings for stealing should apply to both types of games.

[Brad K]

Originally Posted by Garlon

I made a suggestion for career leagues not replay leagues. I never said the same strategy settings for stealing should apply to both types of games.

I test replay leagues because that is the only valid way to compare a player to his historic output.

Do you have evidence the game engine changes depending on what type of league is selected?

[NoOne]

Originally Posted by Rain King

I don't appreciate the usage of the "R" word.

I'm okay with that. Each to their own.

[Brad K]

HERE is a post that demonstrates the need for explanations on the option screen of the effect of various selections.

[Brad K]

Several people have stated a belief that using three year or five year recalc instead of one year produces player output closer to historic than one year recalc. The advantage of a larger sample size is cited.

Output is closest to historic when the ratings match historic performance. The OOTP engine cares about the ratings. It doesn't care about the size of the sample used to create the ratings.

As shown the chart below, multi year recalc results in ratings that do not match the historic performance. Thus multi year is less accurate not more accurate.

The chart is of Andrew McCutchen's 2009-2018 stats.

[pstrickert]

Your first sentence, Brad -- who all specifically made this claim? I'm curious.

[Rain King]

Nobody has claimed that.

People like 3-Year and 5-Year because it can represent individual players better because it smooths out outlier and small sample size seasons. Nobody is saying it makes things "closer to historic" for a single season.

He is using the "make up a problem" so I can provide a solution that gives me something I want tactic. Intentional misrepresentation.

It is also never mentioned in any of these rants how the user is able to customize the Adjust/Weaken settings and/or the "double current year" setting, which make a difference in how all of this plays out as well.

[NoOne]

Careful, cry babies around here.

[Brad K]

Originally Posted by pstrickert

Your first sentence, Brad -- who all specifically made this claim? I'm curious.

I intentionally didn't quote anyone.

We could have a more fruitful discussion if you stated your view on the matter rather than, I assume, defending someone else's view. Have you considered that since they posted they may have changed their mind?

So, does multi year create ratings better able to produce historic output? My chart indicates it can't.

[Brad K]

Originally Posted by Rain King

Nobody has claimed that.

People like 3-Year and 5-Year because it can represent individual players better because it smooths out outlier and small sample size seasons. Nobody is saying it makes things "closer to historic" for a single season.

He is using the "make up a problem" so I can provide a solution that gives me something I want tactic. Intentional misrepresentation.

It is also never mentioned in any of these rants how the user is able to customize the Adjust/Weaken settings and/or the "double current year" setting, which make a difference in how all of this plays out as well.

You are now, as you did before, accusing me of misrepresentation aka lying. And claiming its a made up problem that I created in pursuit of some unnamed goal, presumably, from the context, a nefarious one. I'm not inclined to discuss the issues with someone who wants to post about me not the data.

However I will ask a question. You say the effect of multi year is smoothing of the output and that better represents the players. How can something that prevents players from achieving historic performance better represent them?

If you want to answer that and omit the negative personal characterizations, then perhaps we have something to discuss.

[pstrickert]

Originally Posted by Brad K

I intentionally didn't quote anyone.

We could have a more fruitful discussion if you stated your view on the matter rather than, I assume, defending someone else's view. Have you considered that since they posted they may have changed their mind?

So, does multi year create ratings better able to produce historic output? My chart indicates it can't.

I'm simply asking for names, since I don't recall anyone making the claim. Please name them; we can always ask them if they've since changed their minds.