Feedback Requested: Tweaked Settings for Aging Curves

[Charlie Hough]

Originally Posted by No Pepper

Charlie, forgive me if I've missed something. In your tests, Injuries were on or off?

Injuries were turned on with normal frequency.

Your data are total number of active players on rosters at each age? If so, do we know what excludes players from being counted? Injury, retirement, or otherwise?

No, these were total numbers of players in the database, which included players on the 25-man rosters and on team reserve lists (since minors were turned off).

It would seem to me that a study making conclusions about retirement should also have data regarding ages at time of CEI and retirement.

I did not include all the individual player research that I did, picking a number of players to track and searching for examples of unusually early retirements, realistic and unrealistic developmental tracks, and so on. Anecdotally, I can state that most players retire for one of three reasons.

1. They suffer a career-ending injury, which appears to be completely random.

2. They decline in quality and thus are no longer useful and retire after spending multiple seasons on reserve.

3. They retire once they reach 35 or older.

Younger players will typically wait more years on reserve before retiring than players who are older. But the difference is often just a year or two.

With these settings, most players that had decent ratings and remained on a 25-man roster retired at the age of 35 or 36, with a few reaching near 40. There seems to be very little regard for their current production when they retire, as a number of players retired while still producing very strong statistics for their teams. The next largest group was comprised of players that simply declined or otherwise found themselves on reserve with seemingly no future, and they retired. Career-ending injuries are relatively rare and don't seem to affect overrall results very significantly. They may well be in line with the frequency of CEIs in real life.

What if modifying development/aging curves shift the injury curve? You'd also want to know if these settings actually shift retirement, not active status alone.

The only way that injuries affect retirement or development is that career-ending injuries will obviously force immediate retirement, and major injuries will tend to cause a significant decrease in ratings, which players sometimes cannot overcome. This sometimes forces players toward an early end to their careers, because they no longer have as much value to major league teams.

So, if you turn injuries off, then certainly no one will ever retire due to a CEI, and the players whose careers suffer from major injuries might last longer. But this would not solve the real problems, which are the age-based decline in ratings for pitchers age 28 to 30, and the unrealistically strong tendency toward retirement at age 35 and over. The aging modifiers seem able to overcome the first problem, but they may create a second problem with younger players. Whatever the case may be, the modifiers certainly cannot solve the problem of premature retirement.

Have you tried 3 separate imports and compare? Also, I don't see a control set in your tests.

I did not run other tests because other forum posters have already done a lot of testing that involved the default modifers, different league settings, and many other options. It is already well established that the default settings produce very unrealistic and unsatisfactory results in historical simulations, with pitchers declining dramatically at the age of 28 to 30, and players in general retiring too early compared to real life tendencies.

The purpose of this test was to take some of the settings recently recommended in this topic, to test them over several simulations and see if they might solve or at least reduce the problems involving player decline and retirements. I also changed one of the factors to a more extreme number, simply to find out if the difference between -0.200 and -1.000 would generate dramatically different results.

[Charlie Hough]

1998 Yankees, I'm glad that you decided to join the forums, and thanks for the encouragement. Based on your original posting, I've created a new topic in the General Discussions forum, and you might be interested in reading it.

[fantom1979]

Originally Posted by Charlie Hough

You can now find further results from my aging and development tests here. I had to upload them all to my webspace, as their is too much information to include on the forums.

http://www-personal.umich.edu/~gandrade/ootp/aging/index.html

I think the results have gone a long way toward identifying the limitations of OOTP when it comes to avoiding player decline and premature retirements in historical play.

In 1991 you had ZERO pitchers over the age of 37 in your league. Has that ever happened in the history of MLB?

[Raidergoo]

Originally Posted by fantom1979

In 1991 you had ZERO pitchers over the age of 37 in your league. Has that ever happened in the history of MLB?

http://www.baseball-reference.com/le...38_pitch.shtml

Wins
Rank	Player	  	                W	  	 Year
1.	Randy Johnson		24		2002
2.	Warren Spahn		21		1959
3.	Spud Chandler		20		1946
 	Roger Clemens		20		2001
 	Jamie Moyer		20		2001
6.	Cy Young		18		1905
7.	Babe Adams		17		1920
 	Bert Blyleven		17		1989
 	Mordecai Brown		17		1915
 	Ed Cushman		17		1890
 	Charlie Hough		17		1986
 	Dutch Leonard		17		1947
 	Jim Perry		17		1974

Innings
Rank	Player	  	                IP	               Year
1.	Phil Niekro		330.3		1977
2.	Cy Young		320.7		1905
3.	Ed Cushman		315.7		1890
4.	Warren Spahn		292.0		1959
5.	Steve Carlton		283.7		1983
6.	Joe Niekro		263.7		1983
7.	Babe Adams		263.0		1920
8.	Walter Johnson		260.7		1926
9.	Randy Johnson		260.0		2002
10.	Spud Chandler		257.3		1946

Wins Including and after age 38
Rank	Player	  	               W	  	    IP
1.	Phil Niekro		156		2641.7
2.	Jack Quinn		122		1926.7
3.	Warren Spahn		117		1722.7
4.	Cy Young		106		1835.3
5.	Charlie Hough		102		1862.0
6.	Nolan Ryan		93		1681.7
7.	Gaylord Perry		83		1490.7
8.	Roger Clemens		81		1037.7
9.	Tommy John		74		1362.7
 	Jamie Moyer		74		1057.3

Most pitchers never make it to 38. The ones that do are the best that ever suited up, and have some sort of gimmick. I see knuckleballers, soft tossers, and glory be to God genetic freaks that have turned their bodies into temples.

If you start to look at 39 year old pitchers, the best all time have stats that are very ordinary ( 10th best all time cumulative ERA post 39 yr+ is 3.50).

This still does not excuse the fact that the game roughly treats young SP, overemphasizes 26-32 yr old SP, and sorely ignores 34 to 40 yr olds.

[Charlie Hough]

Interesting information, but it's not useful in actually determining how many pitchers age 38 or older have been in all the years of baseball.

I had hoped to gather this information using the Lahman database, but unfortunately the data setup in the master file will not allow me to properly subtract final game dates from birthdates, to determine how old players were when they played their final games. Birthdates are separted from birth months and birth days, while the final game dates are in a mm/dd/yyyy format. So it's impossible to do any proper subtractions in Excel.

However, if you do some imports from Lahman to represent various eras, you will find that players have played into their late 30s and early 40s for virtually all of baseball history, and not all of them were starts. Granted, most were very good players in their prime, but there were plenty of mediocre players that made it that far as well.

[arod23]

Originally Posted by Charlie Hough

I had hoped to gather this information using the Lahman database, but unfortunately the data setup in the master file will not allow me to properly subtract final game dates from birthdates, to determine how old players were when they played their final games. Birthdates are separted from birth months and birth days, while the final game dates are in a mm/dd/yyyy format. So it's impossible to do any proper subtractions in Excel.

You can use the master.csv file from our DB (link below). I had to fix all birthdays so they were in the same format. We deleted some players who only had a few appearances so it does not have all the players Lahman does, but it will server your purposes.

[Charlie Hough]

I just checked your database, and the master.csv file shows the birthdays broken out with separate years, months, and days, just like the Lahman database. So, unfortuantely, that isn't going to work. I need the birthdays in a mm/dd/yyyy format, or I need the final game dates broken out with separate years, months, and days.

[Zeyes]

Originally Posted by Charlie Hough

I just checked your database, and the master.csv file shows the birthdays broken out with separate years, months, and days, just like the Lahman database. So, unfortuantely, that isn't going to work. I need the birthdays in a mm/dd/yyyy format, or I need the final game dates broken out with separate years, months, and days.

You could copy the individual parts of the birthdates into a new column, with something like =A1&"/"&B1&"/"&C1 to get them into the proper format.

[Charlie Hough]

Thanks for the help. That solved the problem. But, unfortunately, it looks like Excel will not calculate total years between two dates if one is after 1900 and the other is before 1900. So I was only able to look at players born in 1900 or beyond. But that still provided a sample base of 9212 players from the Lahman database.

Here are the final results based on real life dates of the final games that players played in the major leagues. This does not reflect when these players retired from professional baseball, as it does not account for any further time that players were in the minor leagues after their last MLB game.

Here is the breakdown for each age. I provide the percentage of all 9212 major league players who played their last games at the given age. Again, this includes only major leaguers who are not active players and were born in 1900 or later.

Age PCT
44+ 0.4%
43 0.4%
42 0.5%
41 0.6%
40 1.0%
39 1.5%
38 2.1%
37 2.5%
36 3.5%
35 4.2%
34 4.5%
33 5.6%
32 5.4%
31 6.3%
30 6.8%
29 8.1%
28 8.3%
27 8.8%
26 8.2%
25 6.7%
24 5.6%
23 3.9%
22 2.4%
21 1.4%
20 0.7%
19 0.3%
18 0.2%

[1998 Yankees]

So, gathering up the loose fibers of this thread (based on patch 1.02, build 12262), if I may:

• The problem is more aggravating to historical simmers, but certainly not limited to that group. Anyone looking for baseball realism is, or should be, concerned.
• The problem seems to be more severe than Seviien's data and proposed settings for Player Aging and Development at the top of this thread would indicate. He was on the right track, but negative numbers are now not only considered possible, but actually the way to go for some modifiers.
• The problem is more pronounced with pitchers than batters, as stated by arod23 ("Basically, batters stay forever, pitchers last 10 years or less, even good ones."), Doughnuts?, and TNCubsFan.
• There are a lot of other factors involved that could be affecting results such as league size, amateur draft depth, number of non-MLB leagues, MLE's (all of which can have an effect on the amount and quality of competition for MLB jobs), general manager preference for veterans, injuries, etc.
• However, if you want to try to address this problem head on with the best possible results, it's advisable to use these settings as a starting point:

[quote=Charlie Hough]I used -1.000 aging and 0.100 development for pitchers, and I used -0.100 aging and 0.200 development for batters./quote] . . . but with several provisos:

• Charlie Hough's suggestions appear to produce a more realistic aging curve but his test results are no guarantee of future performance or satisfaction.
• Simmers will still need to test and tweak, given their preferences and the number of other factors involved.
• The problem of premature retirement at 35+ is not addressed as it appears to be hardcoded.
• There is still a need to "reprogram the relevant aspects of the game code" in Charlie's words, as this is "the easiest way to achieve a more realistic distribution of player ages and to avoid premature retirements . . . Retirement in particular exhibits a pressing need for revision, as it appears immune from any modifiers imposed by the user."

Charlie, thanks again for this help. Folks, I would recommend that you at least take a look at Charlie's summary statement at the link that he provided. Very impressive work.

[Doughnuts?]

excellent summary, yankees '98. all the points are valid. someone should forward it to Markus asap.

[1998 Yankees]

Originally Posted by Doughnuts?

someone should forward it to Markus asap.

I have a feeling the developers are watching this particular thread, considering the number of posts and views it's had and how long it's been "alive" on the board. At least I hope so.

[Raidergoo]

Originally Posted by 1998 Yankees

I have a feeling the developers are watching this particular thread, considering the number of posts and views it's had and how long it's been "alive" on the board. At least I hope so.

I've got a secret!

[sporr]

Through a test performed by a fellow beta-tester, it seems that there is no hard-coding of player retirement. That said, age and randomness are factors in whether a player retires or not. But it would seem that age isn't as strong as other influences. In this test, Player Development was turned OFF. The test found an abundance of players ages 35-43 in the league. That leads me to believe that maybe the problem lies in the way the game calculates aging based on the modifiers.

[1998 Yankees]

Ahhhhhh. New legs for this thread. Excellent.

[dougaiton]

POTENTIAL GREAT NEWS BULLETIN

I sometimes get visions, and I have a vision of Markus saying that there is a bug in the game that meant players in their 30's 'magically' lost their skills causing the average age to slip from 28.5 to around 26. This bug will be corrected in the new build.

Or something pretty much exactly like that.

[Raidergoo]

Now all we have to do is run a bunch of tests to cross check it.

[RonCo]

Yes, yes...testing is good.

[1998 Yankees]

I thank any and all people who have, are, and will be working on this problem. They, their ancestors, and their descendants. Forever and ever. Amen.

[Mattymo]

I've read this entire thread, but I'm definitely not as smart as most of you, especially when it comes to numbers.

Right now I'm in 1904, and my fictional league (i'm also running a historical league) is being dominated by guys who are 21-25 years old, especially hitters (the top 6 prospects in the league are all major league STARS). I signed a 33 y/o RF and 2 years later he had nothing left. Should I adopt these aging curves that Charlie Hough found? Or is it an inexact science?

Edit: I was just looking at the financial report for the league, and the teams with the lowest payrolls are winning because they have more younger players with minimum contracts, while the teams with the higher payrolls are finishing in last place because they have veterans who are becoming washed up.