Modifier explination on a historical league

[David Watts]

Originally Posted by Rain King

This is going to seem staged, but I just did this with the following results.

Ratings for Single Season Replay.
As played lineups and transactions.


Season 1 - 49
Season 2 - 54
Season 3 - 54
Season 4 - 56 (Mantle hit 64)
Season 5 - 55
Season 6 - 44
Season 7 - 58
Season 8 - 54
Season 9 - 55
Season 10 - 61

https://www.screencast.com/t/Hgwmk8Fh

Got another question in regards to this test(thanks for running it by the way). When using real as played lineups, one of the things OOTP recommends during setup is for the user to set the (I call them) the make bad settings to half of whatever the default is for hitters. The instructions for this are part of the setup wizard. It's in the paragraph right below the batter splits option. Did you do the half thing? I wonder what difference it makes in the results.

[Brad K]

Regarding the questions, testing was done with one year recalc and optimized for single season play. Injuries set to OFF but not historic lineups (frankly I forgot about the option).

[Rain King]

Originally Posted by David Watts

Got another question in regards to this test(thanks for running it by the way). When using real as played lineups, one of the things OOTP recommends during setup is for the user to set the (I call them) the make bad settings to half of whatever the default is for hitters. The instructions for this are part of the setup wizard. It's in the paragraph right below the batter splits option. Did you do the half thing? I wonder what difference it makes in the results.

I did not. I basically left everything at default with the only changes being.

- Unchecked "Import complete history" (for loading speed)
- Switched Optimize Player Ratings for... to "Single-Season Replay"
- Checked Real Transactions and Real Lineups
- Checked Play in Commissioner Mode

FYI, I did just run a new simulation and made those suggested edits to the Batter settings. Maris sill has 46 HR in the editor. In the simulation he hit 57.

[David Watts]

Originally Posted by Brad K

Regarding the questions, testing was done with one year recalc and optimized for single season play. Injuries set to OFF but not historic lineups (frankly I forgot about the option).

Thanks, I was just curious.

[David Watts]

Originally Posted by Rain King

I did not. I basically left everything at default with the only changes being.

- Unchecked "Import complete history" (for loading speed)
- Switched Optimize Player Ratings for... to "Single-Season Replay"
- Checked Real Transactions and Real Lineups
- Checked Play in Commissioner Mode

FYI, I did just run a new simulation and made those suggested edits to the Batter settings. Maris sill has 46 HR in the editor. In the simulation he hit 57.

Thanks.

[Brad K]

This seems interesting although I don't have a conclusion as to what it means. First is actual BA for 1961 for the top five in the AL and second is their rated BA on the single season setting.

Player Actual Rating Difference

Cash .361 .360 -.001
Kaline .324 .335 +.011
Piersal .322 .336 +.014
Mantle .317 .333 +.016
Gentile .302 .316 +.014

[thehef]

Originally Posted by Brad K

This seems interesting although I don't have a conclusion as to what it means. First is actual BA for 1961 for the top five in the AL and second is their rated BA on the single season setting.

Player Actual Rating Difference

Cash .361 .360 -.001
Kaline .324 .335 +.011
Piersal .322 .336 +.014
Mantle .317 .333 +.016
Gentile .302 .316 +.014

What "adjust" settings did you use? Default, or lower?

[thehef]

Originally Posted by David Watts

Got another question in regards to this test(thanks for running it by the way). When using real as played lineups, one of the things OOTP recommends during setup is for the user to set the (I call them) the make bad settings to half of whatever the default is for hitters. The instructions for this are part of the setup wizard. It's in the paragraph right below the batter splits option. Did you do the half thing? I wonder what difference it makes in the results.

A few things here:

1) The manual doesn't reference this - that the gamer should manually change the settings to half of the default when using real txns & lineups. It should. (I will mention this in the Manual sub-forum of the beta group.)

2) Seems to me that the recommended settings should change, depending upon the historical txns & real lineups settings. For example, the recommended settings for historical are initially 300/50/40/10. However, per the blurb on the Historical Wizard panel that David references, that recommendation changes to 150/25/40/10 (lower for the hitters only) if using real txns & real lineups. Therefore, those defaults should change along with those selections. The issue, though, is that you don't choose your real txns & real lineups settings on page 1 of the wizard; they are chosen later. So maybe that should be addressed: Making historical settings selections in a more-logical order, both for the gamer, and for the ability of OOTP to automatically set recommended settings.

3) Perhaps both sets of default settings - 300/50/40/10 for semi-historical, and 150/25/40/10 for real txn/lineup historical - should be revisited. I understand that no one size fits all, but several gamers have suggested other settings work best. Do we at this time, as OOTP historical has evolved (mostly for the better, but some issues remain, and some issues have been discovered) know that both of those settings are still the best recommendations?

[David Watts]

Originally Posted by thehef

A few things here:

1) The manual doesn't reference this - that the gamer should manually change the settings to half of the default when using real txns & lineups. It should. (I will mention this in the Manual sub-forum of the beta group.)

2) Seems to me that the recommended settings should change, depending upon the historical txns & real lineups settings. For example, the recommended settings for historical are initially 300/50/40/10. However, per the blurb on the Historical Wizard panel that David references, that recommendation changes to 150/25/40/10 (lower for the hitters only) if using real txns & real lineups. Therefore, those defaults should change along with those selections. The issue, though, is that you don't choose your real txns & real lineups settings on page 1 of the wizard; they are chosen later. So maybe that should be addressed: Making historical settings selections in a more-logical order, both for the gamer, and for the ability of OOTP to automatically set recommended settings.

3) Perhaps both sets of default settings - 300/50/40/10 for semi-historical, and 150/25/40/10 for real txn/lineup historical - should be revisited. I understand that no one size fits all, but several gamers have suggested other settings work best. Do we at this time, as OOTP historical has evolved (mostly for the better, but some issues remain, and some issues have been discovered) know that both of those settings are still the best recommendations?

If I remember right, Pstrickert recommends settings for real lineups that differ from the ones OOTP recommends.

[thehef]

Originally Posted by David Watts

If I remember right, Pstrickert recommends settings for real lineups that differ from the ones OOTP recommends.

Right. And Garlon has a few recommended settings, IIRC. In addition, maybe there should be 3 sets of recommended settings:

1) historical without txns & lineups (currently 300/50/40/10)
2) historical with txns but without lineups (currently...???)
3) historical with txns and with lineups (currently 150/25/40/10)

And not to get carried away here, but whether or not you're using real injuries or not could affect things. Some big unknowns at this point.

[Brad K]

Originally Posted by thehef

What "adjust" settings did you use? Default, or lower?

One year recalc optimize for single season. I took the ratings out of one of my existing test games. I stopped them at the end of the regular season so next season calculations aren't done yet. Dev and TCR are off so they're the same as start of season ratings.

[jcard]

In addition to the playing time thresholds for the “weaken” and “adjust” functions, I would strongly recommend that the baselines used for the regressions be revisited to see if they are consistent with their stated definitions (“replacement-level” and “average”, respectively); no thresholds will yield satisfactory results if the regression baseline isn’t appropriate to its stated purpose.
Actually, I already have recommended this in the “Bug Reports” sub-forum. While it apparently hasn’t been deemed worthy of any reply, I think that the people taking part in this discussion should at least give it a read. It is obvious that you all have already spent a great deal of time trying to get things “right”; whatever your conclusions, I think that you will find investing a little more to read this to be worthwhile:
https://forums.ootpdevelopments.com/...d.php?t=331986

[Brad K]

I read your other thread and posted to it.

Your posts would be much easier to read without the formatting misunderstood by the board.

[jcard]

I apologize for the formatting and had not realized the problem. I am not sure why the apostrophes and quotation marks are doing this. I did not think that posting from an iPhone would cause formatting issues, and the formatting looks fine when I am actually drafting the post. Thank you for calling my attention to the fact that I was inadvertently undermining my efforts to convey certain ideas clearly. For the moment, I have tried to avoid using any of the culprit punctuation in this message.

[Brad K]

Originally Posted by jcard

I apologize for the formatting and had not realized the problem. I am not sure why the apostrophes and quotation marks are doing this. I did not think that posting from an iPhone would cause formatting issues, and the formatting looks fine when I am actually drafting the post. Thank you for calling my attention to the fact that I was inadvertently undermining my efforts to convey certain ideas clearly. For the moment, I have tried to avoid using any of the culprit punctuation in this message.

I assumed you were doing C&P from a word processor.

[Brad K]

Originally Posted by thehef

Right. And Garlon has a few recommended settings, IIRC. In addition, maybe there should be 3 sets of recommended settings:

1) historical without txns & lineups (currently 300/50/40/10)
2) historical with txns but without lineups (currently...???)
3) historical with txns and with lineups (currently 150/25/40/10)

And not to get carried away here, but whether or not you're using real injuries or not could affect things. Some big unknowns at this point.

Both Neutralized and Real are neutralized although apparently against different park factors. Only stats neutralized against the game's park factors are useful. So Neutralized is either wrong or redudant.

So a logical set of choices is actual stats with no adjustments (the game neutralizes against the game's park factors file)

1) One year recalc
2) Three year recalc with a double weight current year option
3) Five year recalc with a double weight current year option

The following is an idea for user notification

One year recalc is the actual stats for the year. No adjustments are made except for players not playing for their historical teams. Those players are adjusted for the park factors of their new home park. Randomness may cause a player to exceed his best ever performance.

Three and five year recalc smooths the data so it is less likely a player will exceed his peak historical performance. Weighting increases the possibility a player player will match his peak performance and also the possibility he will exceed it.

While individual years may differ for historic, all settings result in a high possibility of very close to historic performance on a career basis. Variance from historic performance can be increased by activating the development engine and activating Talent Change Randomness. Non historic lineups and non historic injuries also affect career performance.

[Brad K]

Originally Posted by Lukas Berger

Keep in mind that we don't rig the outcomes of specific player seasons like some games may do, where once a player hits their expected level, they then fall off in talent to keep things from getting out of control.

I don't think there's a valid claim to moral purity by not rigging the system by setting caps when the system is rigged in another way to get the same result.

[thehef]

Originally Posted by Brad K

Both Neutralized and Real are neutralized although apparently against different park factors. Only stats neutralized against the game's park factors are useful. So Neutralized is either wrong or redudant.

So a logical set of choices is actual stats with no adjustments (the game neutralizes against the game's park factors file)

1) One year recalc
2) Three year recalc with a double weight current year option
3) Five year recalc with a double weight current year option

The following is an idea for user notification

One year recalc is the actual stats for the year. No adjustments are made except for players not playing for their historical teams. Those players are adjusted for the park factors of their new home park. Randomness may cause a player to exceed his best ever performance.

Three and five year recalc smooths the data so it is less likely a player will exceed his peak historical performance. Weighting increases the possibility a player player will match his peak performance and also the possibility he will exceed it.

While individual years may differ for historic, all settings result in a high possibility of very close to historic performance on a career basis. Variance from historic performance can be increased by activating the development engine and activating Talent Change Randomness. Non historic lineups and non historic injuries also affect career performance.

I like this. Re the part in bold, I also think something should be included for the flip side of peak years: "... and it is less-likely that a player will experience a markedly off year." It can probably be phrased better, but you get the idea

[Brad K]

Originally Posted by thehef

I like this. Re the part in bold, I also think something should be included for the flip side of peak years: "... and it is less-likely that a player will experience a markedly off year." It can probably be phrased better, but you get the idea

I couldn't figure out good phrasing which is why I left it out. But yeah, something about from bad to worse needs to be in there!

Also, with this method the need for answers to the questions about the effect of real stats, neutralized stats, and differences goes away. It's obvious from the names what they do. The description is moslty to deflect in advance complaints from single season settings about why Roger Maris sometimes hits 75 HRs or Chris Davis sometimes hits .138.

[Garlon]

Originally Posted by Brad K

Both Neutralized and Real are neutralized although apparently against different park factors. Only stats neutralized against the game's park factors are useful. So Neutralized is either wrong or redudant.

In OOTP 21 the neutralized stats (which had not been updated for any season since about 2011) were based on the quadratic method, the same method that baseball reference uses and the park factors were based on the quadratic method using 5yr factors and I posted the quadratic park factors calculator too for everyone.

For the new game I went to the actual discrete Bats L and Bats R data for BABIP/2B/3B/HR and created 3yr park factors for Bats L and Bats R for every stadium since about 1905 since 1905. This had to be done custom for OOTP to get the values in a format that make sense for how OOTP uses them. This is why now in historicals you have different L/R values for stadium factors.

With the new park factors I was then able to make a complete rebuild of the neutralized player statistics using the Bats L or Bats R for each player in history. The neutralized stats were specifically made with the new park factors.

The neutralized stats are very close for nearly all players because they still play half their games on the road. Even if you played in 1950 with 8 teams in your league and your home stadium had a HR factor of 1.25 and you had 40 HR, how many extra HR did the home park provide? Well the home park is 1.25, so the road parks must be (8 - 1.25)/7 = 0.964. So the combined home and road environment for the player was (1.25+0.964)/2 = 1.107. So 40/1.107 = 36. So the player had 4 extra HR that season compared to being a neutral stadium.