Modifier explination on a historical league

[Brad K]

Originally Posted by Rain King

I checked the editor in a handful of the seasons and in each of them the "Resulting Stats (in a neutral, modern Major League environment)" says 46 HR's.

That's interesting. None of my imports rated 46 HRs. (See post #85,)

[Rain King]

Originally Posted by Brad K

That's interesting. None of my imports rated 46 HRs. (See post #85,)

Settings-wise, my sims match-up with the line that has 47 HR in your post. I just created another league and again got 46.

[Brad K]

Originally Posted by Rain King

Settings-wise, my sims match-up with the line that has 47 HR in your post. I just created another league and again got 46.

Strange.

[thehef]

Originally Posted by Lukas Berger

The problem is, this is not how probability works and if you use real probability, that means there are going to be outliers, both high and low.

Sure. But based upon my experience and what I think you are saying is that the low outliers will generally outnumber the high outliers (for HR's, but not necessarily for batting avg), and that Maris is more-likely to hit fewer - rather than more - HR's in a 1-year recalc. Right? Or am I misunderstanding?

Originally Posted by Lukas Berger

There's no way to get what you seem to want, which, correct me if I'm wrong, Maris hitting something like 59-63 HR's every time you reply the 61 season, without abandoning real probability and manipulating things behind the scenes to force guys to perform within a very narrow band.

Lukas: I am 100% with you/OOTP in that the results should not be rigged. And no, I'm not saying that Maris should hit 59-63 HR's every time in '61. I'm just saying that it should generally be in the ballpark of 61, and should be just as likely to hit a few (or even several) more than 61 as he is to hit less than 61.

Rain King's testing revealed the following:

Originally Posted by Rain King

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

That's an average of 54 HR's. If the above is the expectation, then something is wrong. Without splitting hairs on the exact range, it should average out to something more like in the 57-65 range.

Now, if you're saying that Rain King's 10-season test is too small of a sample size, and that if, say, a 100 seasons were included, the avg might be more in line with (say) that 57-65 range, then ok. But that's not we're seeing so far.

Assuming 1-year recalc and real-life usage, what I (and most of us?) want to see is that a) it's just as extremely unlikely that Maris would hit 41 as it is he would hit 81, b) Maris would hit somewhere in the range of 61, with the "a little higher" and "a little lower" chances being equal, and c) Maris' 1961 season would look in OOTP just as it did in real-life: an outlier as viewed against the rest of his career (rather than his career OOTP stats looking like 3- or 5-year recal were being used).

[Syd Thrift]

I’m of the opinion, based on the actual numbers Maris produced in his career, that even for him 1961 was an incredibly lucky year. It was an expansion season, which probably helped a little, but if normal distribution centers around 54 that, frankly, sounds about right.

Most replays you will not have Maris hitting 61, it’s true. Most replays will have someone - Maris or Mantle or Aaron or Hank Greenberg or Chuck Klein or George Foster or Reggie Jackson - break 60 before McGwire and Sosa.

[Lukas Berger]

Originally Posted by thehef

Sure. But based upon my experience and what I think you are saying is that the low outliers will generally outnumber the high outliers (for HR's, but not necessarily for batting avg), and that Maris is more-likely to hit fewer - rather than more - HR's in a 1-year recalc. Right? Or am I misunderstanding?

I suppose that depends on exactly how you define 'outnumber' but the most extreme outliers are more likely to come on the high side rather than the low, even if they're not actually more frequent. If you target 61 for Maris, maybe he's equally likely to hit over 61 or under in a given sim, but in rare outlier seasons he's going to be much more likely to hit 81 than he ever would be to hit 41.

Originally Posted by thehef

Lukas: I am 100% with you/OOTP in that the results should not be rigged. And no, I'm not saying that Maris should hit 59-63 HR's every time in '61. I'm just saying that it should generally be in the ballpark of 61, and should be just as likely to hit a few (or even several) more than 61 as he is to hit less than 61.

Rain King's testing revealed the following:

That's an average of 54 HR's. If the above is the expectation, then something is wrong. Without splitting hairs on the exact range, it should average out to something more like in the 57-65 range.

Now, if you're saying that Rain King's 10-season test is too small of a sample size, and that if, say, a 100 seasons were included, the avg might be more in line with (say) that 57-65 range, then ok. But that's not we're seeing so far.

Assuming 1-year recalc and real-life usage, what I (and most of us?) want to see is that a) it's just as extremely unlikely that Maris would hit 41 as it is he would hit 81, b) Maris would hit somewhere in the range of 61, with the "a little higher" and "a little lower" chances being equal, and c) Maris' 1961 season would look in OOTP just as it did in real-life: an outlier as viewed against the rest of his career (rather than his career OOTP stats looking like 3- or 5-year recal were being used).

Yes, but we've now come full circle on this. The problem is, if you make 61 the target as the median for Maris, then as long as you're using real probability, you're going to have outlier seasons where he's hitting in the 70's or maybe even over 80 and even if you're ok with that, in my opinion the majority of the folks who are playing with single-season recalc or doing single-season replays would be far more unhappy with that than with what we have now.

[thehef]

Originally Posted by Syd Thrift

I’m of the opinion, based on the actual numbers Maris produced in his career, that even for him 1961 was an incredibly lucky year. It was an expansion season, which probably helped a little, but if normal distribution centers around 54 that, frankly, sounds about right.

Most replays you will not have Maris hitting 61, it’s true. Most replays will have someone - Maris or Mantle or Aaron or Hank Greenberg or Chuck Klein or George Foster or Reggie Jackson - break 60 before McGwire and Sosa.

I'm not sure why - even if '61 was lucky for Maris, and it was an expansion year - that with 1-year recalc as the setting, normal distribution of 54 would sound correct. To me, with 1-year recalc, normal distribution on any statistic for any player should be right around what he actually did IRL.

- With 5-year recalc, normal distribution ought to be around 34 1/2, assuming similar usage.
- With 5-year recalc, double-weight on current year, n/d ought to be around 39
- With 3-year recalc, n/d ought to be around 44
- With 3-year recalc, double-weight on current year, ought to be around 48 1/2
- With 1-year recalc, n/d ought to be around 61

With the last one in particular, I don't see why any expectation other than "around 61" would be anything other than inaccurate. I mean, do we really want OOTP making a determination that a player got lucky in a particular year and shouldn't have hit as many HR's as he did, or hit for as high an average as he did, or struck out as many batters as he did? (Did Davey Lopes get lucky when hit a career high 28 HR's in 1979, and therefore OOTP should - in essence - 3-year or 5-year recalc him, even if the gamer chooses 1-year recalc?)

[thehef]

Originally Posted by Lukas Berger

I suppose that depends on exactly how you define 'outnumber' but the most extreme outliers are more likely to come on the high side rather than the low, even if they're not actually more frequent. If you target 61 for Maris, maybe he's equally likely to hit over 61 or under in a given sim, but in rare outlier seasons he's going to be much more likely to hit 81 than he ever would be to hit 41.

Re the parts in bold, I think that is exactly the problem. He should not be more likely to hit 81 than he is to hit 41. He should be approximately equally (un)likely to hit either of those numbers. And then OOTP wouldn't need to set his ratings lower than they should be in order to avoid the high outliers.

Re the part in dark red, I think Rain King's testing indicates that Maris is not equally likely to hit over 61 as he is to hit under 61. Rather, he's more-likely to fall short.

[Lukas Berger]

Originally Posted by Syd Thrift

I’m of the opinion, based on the actual numbers Maris produced in his career, that even for him 1961 was an incredibly lucky year. It was an expansion season, which probably helped a little, but if normal distribution centers around 54 that, frankly, sounds about right.

Yeah, I agree. And I haven't gone into this so much in the thread so far since it's a different argument than the more mathematical one I've been using so far, but to me from an artistic or gameplay perspective, it also makes more sense to treat Maris' 61 season like this, as well as most other outlier seasons.

If you could somehow replay the real-life 61 season multiple times, I don't think it would be realistic to expect Maris to break the record more often than not or even be equally likely.

The real season was incredibly tense, there was huge amount of drama and question as to whether he could break Ruth's record. When he finally did, there was such a sense of shock and astonishment and wonder and of seeing something remarkable around that.

If you set things up so that he's more likely to break that mark than not in any given sim, then you take away the main part of the feeling and spirit of replaying the '61 season and turn the nail biting suspense into a boring ho-hum fait accompli, where the only drama or suspense or intrigue is around whether he'll hit 71 or not.

In my opinion, that's not what best serves the majority of the folks who do single-season replays. Obviously that's just opinion and there's no right or wrong.

I have a lot of respect for thehef and all he's done helping out with the game over the years, and I'd never even want to imply my take on this is better than his. It's just two different ways of looking at it. Both are valid, but since we do have to choose one method to use by default, in my opinion this is the choice that will make more folks happy, even though there's no choice we can make here that will satisfy everyone.

And for anyone who disagrees, there are plenty of very simple options that can make Maris more likely to hit 61 in a given sim, like using the 'make bad' feature to make a bit more aggressively like I'd mentioned earlier or even just boosting his power rating manually if you want to.

[Lukas Berger]

Originally Posted by thehef

Re the part in dark red, I think Rain King's testing indicates that Maris is not equally likely to hit over 61 as he is to hit under 61. Rather, he's more-likely to fall short.

Right, but this is why I said "If you target 61".

I wasn't saying we are targeting it now, I was saying basically 'if we changed things to do like you want to, this would happen'.

Originally Posted by thehef

Re the parts in bold, I think that is exactly the problem. He should not be more likely to hit 81 than he is to hit 41. He should be approximately equally (un)likely to hit either of those numbers. And then OOTP wouldn't need to set his ratings lower than they should be in order to avoid the high outliers.

The problem here is that this is not really a philosophical discussion where there's a mechanic you can argue against or that we can change if we wanted to (at least not without completely redesigning the game from the ground up or rigging things). It's just how probability in the context of the game's mechanics actually works.

So if someone is rated to hit 61 HR's in a season where only 7 players in the league hit more than 40, he's always going to be more likely to have his extreme outliers be too high than too low, because he's already standing out so much from the crowd on the high side. He's too good to ever realistically get unlucky enough to hit 41 HR's, assuming normal playing time, but he'll sometimes get lucky enough to hit 81 (or 75 or whatever, I'm not sure he'd necessarily hit 81 specifically). The more you make him stand out from the crowd, the 'worse' this gets.

As far as I know, the only way to prevent that would be to start building in brakes and fiddling with the probability behind the scenes, like I've already discussed.

[thehef]

Originally Posted by Lukas Berger

Yeah, I agree. And I haven't gone into this so much in the thread so far since it's a different argument than the more mathematical one I've been using so far, but to me from an artistic or gameplay perspective, it also makes more sense to treat Maris' 61 season like this, as well as most other outlier seasons.

But OOTP doesn't treat outlier seasons for batting avg in the same - lowered expectation - way. Why not? I get distribution part of it, that from an overall statistical perspective the hits required for a .400+ season aren't as big of an outlier at the HR's required for a 55+ homer season. However, to me, that's simply and explanation of why something isn't working quite right. It doesn't take away from the fact that something isn't working quite right. If, in OOTP, Ted Williams is expected to hit around .400 in 1941, but Maris is not expected to hit around 61 dingers in 1961, then something isn't working quite right. It may be working the way it is expected due to the design of the program, but that doesn't mean it's working the way it ought to work. Just my opinion, apparently in the minority of the few who are interested

Originally Posted by Lukas Berger

If you could somehow replay the real-life 61 season multiple times, I don't think it would be realistic to expect Maris to break the record more often than not or even be equally likely.

The real season was incredibly tense, there was huge amount of drama and question as to whether he could break Ruth's record. When he finally did, there was such a sense of shock and astonishment and wonder and of seeing something remarkable around that.

With all due respect, that's all totally irrelevant for a sim that's set up for 1-year recalc. OOTP is supposed to be stats-driven, so it's not like OOTP is programmed to take into account that these intangible factors that Maris faced would make his real-life output unlikely, despite the fact that he actually did it real life. Davey Johnson's 43 HR's in 1973 was - from an outlier perspective - more unlikely than Maris' 61 in '61. But there was no "drama and question as to whether he could" hit such an unusually high number of dingers (compared to his history).

Originally Posted by Lukas Berger

If you set things up so that he's more likely to break that mark than not in any given sim, then you take away the main part of the feeling and spirit of replaying the '61 season and turn the nail biting suspense into a boring ho-hum fait accompli, where the only drama or suspense or intrigue is around whether he'll hit 71 or not.

I don't think anybody's asking for that (certainly not me). Rather, I think Maris should simply be rated to hit in the vicinity of 61 HR's. With the likelihood of a few more or a few less being about the same.

Originally Posted by Lukas Berger

In my opinion, that's not what best serves the majority of the folks who do single-season replays. Obviously that's just opinion and there's no right or wrong.

In my opinion, what best serves those who do single-season replays and/or 1-yr recalc historical, is for players to be rated to approximate their real-life performance - without OOTP interpreting it as a "lucky" (or whatever) outlier and adjusting it downward - and then letting the internal "dice rolls" account for randomness & variations.

Originally Posted by Lukas Berger

I have a lot of respect for thehef and all he's done helping out with the game over the years, and I'd never even want to imply my take on this is better than his. It's just two different ways of looking at it. Both are valid, but since we do have to choose one method to use by default, in my opinion this is the choice that will make more folks happy, even though there's no choice we can make here that will satisfy everyone.

Fair enough (and thanks). To me, the issue of HR hitters generally falling short of their big-masher seasons is on par with the stolen base success % issue of top base-stealers. Both are problems in that the replay results become unrealistic. The difference I guess is that you (Lukas) and the OOTP devs agree on the stolen base issue. Why you don't agree on the HR issue is perplexing, although I get and respect that addressing either one is really big from a coding perspective.

Originally Posted by Lukas Berger

And for anyone who disagrees, there are plenty of very simple options that can make Maris more likely to hit 61 in a given sim, like using the 'make bad' feature to make a bit more aggressively like I'd mentioned earlier or even just boosting his power rating manually if you want to.

I'm thinking the better option would be one that allows the gamer to choose between player ratings (in this case, Maris' HR rating) being a) reflective of the actual season, or b) adjusted downward by OOTP. I thought that 1-year recalc was the a) option, but apparently not.

Originally Posted by Lukas Berger

The problem here is that this is not really a philosophical discussion where there's a mechanic you can argue against or that we can change if we wanted to (at least not without completely redesigning the game from the ground up or rigging things). It's just how probability in the context of the game's mechanics actually works.

So if someone is rated to hit 61 HR's in a season where only 7 players in the league hit more than 40, he's always going to be more likely to have his extreme outliers be too high than too low, because he's already standing out so much from the crowd on the high side. He's too good to ever realistically get unlucky enough to hit 41 HR's, assuming normal playing time, but he'll sometimes get lucky enough to hit 81 (or 75 or whatever, I'm not sure he'd necessarily hit 81 specifically). The more you make him stand out from the crowd, the 'worse' this gets.

Yep, I understand all of that - how it works (at least from a high-level perspective) - but that just sounds like a mechanic trying to justify why I should accept that my car isn't quite running right. It may be a very good explanation of what's actually happening, but the problem still exists.

I think what the overall problem is - and that, due to complexity, OOTP is unlikely to address anytime soon (which I understand) - is the following:

- Ruth (60) and Gehrig (47) together hit 24% of the league-wide HR's in 1927. If you take those two guys out of an 1927 OOTP sim (1-year recalc), other players in the league will combine to make up that significant difference.

- In reality, however, had Ruth & Gehrig been unable to play in 1927, their absence would not have caused other players to hit more HR's (with the exception of the players who replaced them, who would hit a few more HR's due to increased playing time, but - by not being epic mashers like Lou & The Bambino - would not actually have hit all that many more).

At any rate, I think this horse is fairly well beaten beyond death and I need to just file it under it is what it is.

Thanks, Lukas, for engaging with detail. Appreciated

[Syd Thrift]

You need a lot more hits to get to .400 over an entire season than homeruns to get to 61. Thus the normal distribution probability arc is going to be smaller relative to the accomplishment. That said I’d also have real issues with a game where Ted Williams in 1941 was as likely to hit .450 as .350; while unlike Maris he did get close to .400 again (he hit .388 in the 50s) it’s very clear that .406 was a high water mark for him and not some kind of magical median he should be hitting.

[Brad K]

As I stated earlier I don't have a personal interest in single season.

A person who selects multi season ratings base knows what he's getting. He knows he's getting a smoothing of the peaks and valleys. It's clear from the selection.

A person who selects single season doesn't know what he's getting. It doesn't say "base on single season except top performances reduced and part of their production given to lesser players." This is especially misleading when the person has selected "real" stats even if he's read a couple thread here and knows "real" isn't "real".

It appears from the Maris imports I did that perhaps neutralized optimized for single season does not reduce the peak performance. But it's hard to recommend that setting to anyone since it's been discovered that neutralized stats are neutralized against something other than the park factors in the game.

People who play single season are the voices to listen to on how to rate the players for single season, not mine. But at least name the choices so it's clear what people are getting.

[thehef]

Originally Posted by Syd Thrift

You need a lot more hits to get to .400 over an entire season than homeruns to get to 61. Thus the normal distribution probability arc is going to be smaller relative to the accomplishment.

Yep, I tried to explain that I understand that part. However, (to tangle up my metaphor) that is basically the "auto mechanic" telling me that I should accept my engine running less than optimal for single-season, 1-yr recalc, that "historically-high-outlier batting averages will play out pretty well in OOTP, but historically-high-outlier HR's will not"

Originally Posted by Syd Thrift

That said I’d also have real issues with a game where Ted Williams in 1941 was as likely to hit .450 as .350;...

Agreed. Teddy Ballgame hitting .350 or .450 in '41 and Maris hitting 41 or 81 in '61 - all four scenarios - should be approximately/equally extremely unlikely.

Originally Posted by Syd Thrift

...while unlike Maris he did get close to .400 again (he hit .388 in the 50s) it’s very clear that .406 was a high water mark for him and not some kind of magical median he should be hitting.

While true, the facts that TW nearly repeated the feat in a different season, but Maris did not, should have zero influence on a single-season replay and/or 1-yr recalc sim. (And even so, how then do we explain lowering The Bambino's HR ratings? As we know, he had multiple high-HR / outlier (relative to other players) seasons.)... That effect - factoring in other seasons to smooth things out - is what 3- or 5-yr recalc is all about. But 1-yr recalc shouldn't be factoring in what may or may not have happened in previous or future seasons.

[Brad K]

Originally Posted by thehef

That effect - factoring in other seasons to smooth things out - is what 3- or 5-yr recalc is all about. But 1-yr recalc shouldn't be factoring in what may or may not have happened in previous or future seasons.

Bravo.

I have a 1961 test loaded, single season, "real", and because of other seasons interfering with single season Mantle has a higher HR rating than Maris.

[thehef]

Originally Posted by Brad K

A person who selects multi season ratings base knows what he's getting. He knows he's getting a smoothing of the peaks and valleys. It's clear from the selection.

A person who selects single season doesn't know what he's getting. It doesn't say "base on single season except top performances reduced and part of their production given to lesser players." This is especially misleading when the person has selected "real" stats even if he's read a couple thread here and knows "real" isn't "real".

It appears from the Maris imports I did that perhaps neutralized optimized for single season does not reduce the peak performance. But it's hard to recommend that setting to anyone since it's been discovered that neutralized stats are neutralized against something other than the park factors in the game.

People who play single season are the voices to listen to on how to rate the players for single season, not mine. But at least name the choices so it's clear what people are getting.

This ^^^. Very much this ^^^. Especially that 2nd paragraph.

[Brad K]

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

OK, well since I've been able to eliminate the two neutralized stats selections as not functioning correctly in the game I decided to do a test too. I'm still getting 47 for Maris' home run rating unlike the 46 you get. Anyway, single season real stats

Mantle rating 51 Maris rating 47

Mantle/Maris

60 46
55 36
48 34
37 40
46 56
50 45
52 51
51 50
41 43
57 46


Mantle Average 49.7 Actual 54
Maris Average 44.7 Actual 61

[David Watts]

Originally Posted by Brad K

OK, well since I've been able to eliminate the two neutralized stats selections as not functioning correctly in the game I decided to do a test too. I'm still getting 47 for Maris' home run rating unlike the 46 you get. Anyway, single season real stats

Mantle rating 51 Maris rating 47

Mantle/Maris

60 46
55 36
48 34
37 40
46 56
50 45
52 51
51 50
41 43
57 46


Mantle Average 49.7 Actual 54
Maris Average 44.7 Actual 61

Was that with as played lineups?

[thehef]

Originally Posted by Brad K

OK, well since I've been able to eliminate the two neutralized stats selections as not functioning correctly in the game I decided to do a test too. I'm still getting 47 for Maris' home run rating unlike the 46 you get. Anyway, single season real stats

Mantle rating 51 Maris rating 47

Mantle/Maris

60 46
55 36
48 34
37 40
46 56
50 45
52 51
51 50
41 43
57 46

Mantle Average 49.7 Actual 54
Maris Average 44.7 Actual 61

Just to clarify, was this with 1-year recalc? If so, it would make no sense that Mantle would consistently across your tests a) have a higher HR rating, and therefore b) hit more HR's... Maris had more plate appearances, so actually these two guys ended up homering at an almost even rate (Mantle 8.4% of the time, Maris 8.8%). So either OOTP is looking at more than just one year for 1-year recalc - which would make no sense - or something else is amiss.

[Lukas Berger]

Originally Posted by thehef

Just to clarify, was this with 1-year recalc? If so, it would make no sense that Mantle would consistently across your tests a) have a higher HR rating, and therefore b) hit more HR's... Maris had more plate appearances, so actually these two guys ended up homering at an almost even rate (Mantle 8.4% of the time, Maris 8.8%). So either OOTP is looking at more than just one year for 1-year recalc - which would make no sense - or something else is amiss.

Yeah, I'd tend to agree here. Something feels off.