Analysis Results: Aging Modifiers and Career Arcs

[Abu Taha]

I am trying to get players in my league to age slightly more slowly. Problem is I don't know if I can accomplish this by setting the aging modifer to .9 or 1.1. Any ideas? Anybody tested this?

[redmarkYankees]

Slow ageing should be < 1, quicken ageing > 1.

[Lee]

I have run some test leagues with Pitcher aging set at varied levels (1.000, .600, .200) Running a historical sim, with real players from the Arod/Garlon DB. I am still noticing far too many Starting pitchers basically finished at the age of 30....If they don't retire by 31 or 32, they usually languish as relievers, getting used less and less as time goes on.

The fact that this still happens and that the settings on aging don't seem to have too much of an effect makes me question the AI. If this is what is causing it I think a tweak may be needed. There is no way the AI manager should demote a star Pitcher, who still has solid ratings to the Bullpen in favor of a younger Pitcher, just because he is younger.

Guys this has a huge effect on Historical solo sims, please take a look at this if possible. A starter with good ratings, that stays healthy should be able to pitch well into his 30's. I have plenty of examples if needed.

[TC Dale]

Please post examples. How long of a test are you running, etc..? More info could really be helpful here.

[Lee]

Will do TC, I am mostly running samples from 1903 - 1925. I am in the process of running tests with different options and aging features checked. Will have some comparisons put together no later than tomorrow morning for you.

[TheBishop]

There seems to be a minor improvement with pitchers not just completely falling off the table, but it's still pretty prevelant.

While pitchers dropping off at or around 30 isn't an unusual thing (it's actually pretty realistic in some ways; ie Kerry Wood and Eric Gagne this year alone), I think the main problem is that the pitchers go from being good to completely bad, with only a minimal drop in ratings. Guy goes from 30 game winner in 1932 to pitching straight out of the pen the next year, never getting a start, then to the minors (reserve roster for some of us *cough*) for a few more seasons if he doesn't just retire right off. There should be more of a flow from great to good to decent to mediocre to worse to bad.

There is no injury problems with 99% of them (I've been playing on low injuries to test), they just slip in one rating and the AI manager thinks they're done, puts them in the pen and that's that.

The other problem I still see is that no one can win 300 game. Not one pitcher over 40 seasons has won 300 in this test league with patch 2. Four pitchers have lost 300, ironically, but no one goes over 277 wins. Part of that is because the AI manager dumps them so quickly and the game reads their slight rating drop as catastrophic, like they're suddenly missing a limb or something.

btw, all my testing was done with aging on default and pitchers on normal usage.

[TheBishop]

After switching to high endurance from normal before the 1960 season, the league has produced eight 300 game winners, including 1 400 gamer (Whitey Ford went nuts )

Starting pitchers seem to respond better in this setting, though they still die too quickly. At least a few guys finally did it, though.

[rasnell]

In a 30-year fictional sim from 1929-1959, I have 8 pitchers with 300 career wins. But my settings were for 4-man rotations and very high endurance.

[Lee]

Here are the results from my latest test. I have run a historical sim from 1901 to 1920. Pitcher aging set at .500.

So far a total of 43 Pitchers have accumulated 1500 + innings that have either retired or are active, but no longer full-time starters. Pitchers who suffered serious injuries (3) with over 1500 innings were not included. This is all based on AI movement after ratings slipped for the first time.

Only 9 of these 43 remained full-time starters after age 30 (20.9%) Here is the individual breakdown by age showing the last year each pitcher was a fulltime starter.

25-1
26-1
27-5
28-12
29-9
30-6
31-1
32-3
33-2
34-3

• From the list above 25 (58.1%) had winning records their last Season as a starter.
• 12 (27.9%) were 20 game winners their last Season as starter.
• 19 (44.1%) were Starters 5 years or less.
• 24 (55.9%) were Starters 6-10 Seasons.
• Nobody remained a full-time starter for more than 10 Seasons and only 4 (9.3%) achieved 10 Seasons as a fulltime starter.

From these results it appears the AI is a bit too eager to replace Pitchers with someone younger the minute their ratings slip the slightest amount, even with aging set to half the default level.

Just as a random sample for this era I looked at the top 45 Starting Pitchers (Statistically) for the 1910 Season. Of those Pitchers 28 (62.2%) remained fulltime starters after age 30. Here is the year by year breakdown of pitchers ages their last Season as a full time starter IRL for the top 45 of 1910.

23-1
24-1
25-1
26-2
27-5
28-2
29-3
30-2
31-6
32-7
33-1
34-5
35-2
36-1
37-1
38-1
40-1
41-2
43-1

The main years that Starters in this test seemed to hit the wall in OOTP was the 28-30 range in which 37 pitchers (60.4%) spent their last Seasons as fulltime starters.

In the RL comparison sample only 7 (15.5%) hit the wall between 28-30.

The age range in the RL sample where the most pitchers seemed to decline was 31-34. In this grp 19 Pitchers (42.2%) saw their last Season as a fulltime starter in this age range. Compared to 20.9% as indicated above that even made it to age 31 in OOTP.

In the OOTP test nobody remained a fulltime Starter after age 34.

In th RL sample of 1910 Starters, 9 Pitchers (20%) Pitched as full time starters after age 34.

In summary I know this is a small sample, but I think it still shows some tweaks are in order with Pitcher aging to produce more realistic Historical simulations. Especially when the aging for this test was set to .500, which in theory should cause pitchers to decline at a rate of half of the default setting.

Plenty of pitchers are effective past 30. I think the numbers above, showing how many Pitchers won 20 games their last year as a full-time starter, validates to an extent, my thoughts that the AI is a bit too hasty to replace Pitchers that still have a few good years left as starters.

[Seviien]

Results:
1. Career arcs are well modeled in OOTP, but they could be improved
....A: Development is too slow; might be because a lack of variety in development precludes really young players from being in the bigs
....B: Onset point of regression should be more varried
2. Aging and development modifiers have little or no effect

I originally posted this in the main forum. I'm not sure if that is the proper place, so I figured I'd repost here. Apologies if I'm wrong.


Methodology:
1. 24 team league fictional league w 162 game schedule
2. 3-tier minors (rookie level dropped)
3. tests conducted by varrying aging and development modifiers
4. each test was loaded from the same quicksave
5. each test simmed 25 seasons
6. results presented are an average of seasons 16-20 (i.e. "steady state")
7. injuries set to "normal"
8. results examined by distribution of the following by player age:
....A: AB for batters
....B: IP for pitchers
....C: GS for pitchers


Supporting documentation:
http://rapidshare.de/files/25464670/...mmary.xls.html

[Lee]

Dola,

Did some projections on this with the pitchers that were 28 their last fulltime Season. As a group they averaged 12.7 wins through their career.

If were aligned more closely aligned to the sample from 1910, it would have given these Pitchers an additional 2-4 years as starters translating into 25.4 to 50.8 more wins as Starters on average, which IMO would equate to more accurate numbers in Historical sims. OK, now I am off my Soapbox

I would be interested in feedback from others doing Historical sims on their findings as mine always produce results similar to the breakdown I did above.

[RonCo]

Originally Posted by Lee

Will do TC, I am mostly running samples from 1903 - 1925. I am in the process of running tests with different options and aging features checked. Will have some comparisons put together no later than tomorrow morning for you.

I'm a little out of my element in early historical simming, so treat this input with a grain of salt, but here's a pretty extensive Win Share-based study that says that deadballers (1903-1919) were pretty much done by 31-32 years old on the whole.

[Seviien]

I took a pretty comprehensive look at this, and I tend to agree -- though I don't think the problem is quite as acute. I used a different metric, though (IP and GS), so that might have something to do with it. The thread is entitled something along the lines of "Analysis Results: Career Arcs and Aging Modifiers."

I posted it this morning and am still waiting on official comment.

[battists]

Seviien, I'm having trouble getting at this information. Can you host it anywhere other than rapidshare, or just e-mail it to me at steve.battisti@ootpdevelopments.com?

Meanwhile, I logged a ticket, TT # 2195, for Markus to take a look at.

Thanks,

Steve

[Seviien]

email sent with the title: "OOTP 2006 Aging Modifier Analysis"

Thanks Steve

[battists]

Sev,

I was talking with Markus about this, and he indicated that the aging modifiers need to be set more extremely to have a significant impact. Would you like to take another whack at it?

If this turns out to be the case, perhaps we can put together some better guidelines I can add to the Game Guide...

Thanks!

Steve

[Lee]

Originally Posted by RonCo

I'm a little out of my element in early historical simming, so treat this input with a grain of salt, but here's a pretty extensive Win Share-based study that says that deadballers (1903-1919) were pretty much done by 31-32 years old on the whole.

And my research somes something similar, but the game seems to draw that line at 28-29. It is the sudden dropoff for so many 20 game winners, to go from staff ace to the bullpen in one Season is just not realistic when it happens so much.

The other thing here is, nearly every single Pitcher becomes a MR at the end of their careers. This just doesn't happen either.....I would rather see the AI leave them as starters for as long as they hold up. Again I don't think the development curve is the big deal here, I think it is the AI that needs to be tweaked so it doesn't just throw starters in the Pen the minute their ratings drop.

[Seviien]

I can certainly take a look at it again. The thing is, there is a natural limit on the aging speed modifier; I did go down to 50% in my tests, and 0% is the lowest possible value.

As far as the development modifier, did Markus give any indication on how extreme the factor should go to cause a change? 200%? 300%?

[Seviien]

I reran another test this afternoon, and I kinda see what Markus means. With aging modifiers set to 0.2 and development modifiers set to 2.0, I saw a very slight shift.

My concern is that there isn't sufficient leway to make careers longer, since the lowest aging can go is 0.001, and even at 0.200, I'm not seeing much of a difference. The career arcs are really appear too sharp with a steep decline at around 30. I'm not sure if that's an aging issue, or an AI judgement... either way it just feels a little harsh.

I can poke around and runs more tests, but given that the file I forwarded has actual MLB data in it, is there any way to recalibrate the career arcs more effeciently on the back end?

[Seviien]

I updated the summary file after running two sims with more exterme modifiers. In a nutshell, the development can be brought into line using more extreme values, but the aging modifer is broken.

I sent you the file...