Pitcher's Dropping Off Despite Doing Well

[knuckler]

100% of his examples included ratings drops, but yes, I'm open to hearing more detail.

He also noted in his first post about players being sent to the bullpen, and as ksyrup mentions, his guys are being sent to the minors. That's why I said we need to hear more or see more examples.

I'm not trying to downplay the existence of an issue. I'm an independent tester with no ties to SI.

I know, and I respect everything you have done and the way in which you have gone about it. I have seen this issue myself(more what ksyrup is experiencing) and posted about it in the beta forum, albeit late in the process. As I said earlier, I share your concern that the main problem maybe the inability of the AI to use stats primarily to make decisions.

[Ksyrup]

Here's a perfect example.

Spring training, 1940, the Yankees release 2-time repeat Cy Young winner Lefty Gomez, coming off a 24-6, 2.46 ERA season. Yes, his ratings dropped from 7/20/10 to 6/15/8, so maybe a demotion/trade/release is justified - maybe - but to be even remotely realistic, the game can't just look at his ratings and dump him before he's even thrown a pitch the next year, short of his pitching arm being severed in a wheat thrasher incident back home on the farm. Completely unrealistic...as is the fact that he's been changed to a MR already. The game needs to model this so that the AI is reacting to what happens on the field, with an assist/confirmation from the ratings, as opposed to proactively taking action that makes no sense in light of recent statistical history.

I hope this is getting some attention, because it's a big issue, IMO.

[jbmagic]

Originally Posted by Ksyrup

Here's a perfect example.

Spring training, 1940, the Yankees release 2-time repeat Cy Young winner Lefty Gomez, coming off a 24-6, 2.46 ERA season. Yes, his ratings dropped from 7/20/10 to 6/15/8, so maybe a demotion/trade/release is justified - maybe - but to be even remotely realistic, the game can't just look at his ratings and dump him before he's even thrown a pitch the next year, short of his pitching arm being severed in a wheat thrasher incident back home on the farm. Completely unrealistic...as is the fact that he's been changed to a MR already. The game needs to model this so that the AI is reacting to what happens on the field, with an assist/confirmation from the ratings, as opposed to proactively taking action that makes no sense in light of recent statistical history.

I hope this is getting some attention, because it's a big issue, IMO.

Does this happen if you put AI weight evaluation at 100% for stats?

[Ksyrup]

I haven't tried that because this isn't a test league and I don't want to throw things out of whack, but according to previous posts above, even changing the ratings weights doesn't help much at all.

[RonCo]

Originally Posted by jbmagic

Does this happen if you put AI weight evaluation at 100% for stats?

Weighting toward stats seems to reduce the issue, but does not remove it completely.

[Ksyrup]

There must be something coded screwy here, because half the problem is these guys getting dropped in the first place, but the other half is that no team is picking them up. Early Wynn was released after a 16-11 3.35 1946 season, and sat in FA for all of 1947. His current ratings are 5/17/11, and actually improved right before he was released (at age 27).

Release him, OK, I still have issues with that as discussed above, but no other team wants him? I'm playing a universe with the Federal League as a "minor major" league, basically taking the scraps from the MLB, and just after a quick survey of pitchers on the FL teams' rosters, Winn's ratings are as good or better than about half the pitchers on those teams' 25-man rosters. And, some of their minor league teams have only 1 player! So it's not like they don't have a place for him. It's like the game is giving him a CEI without the injury.

Oh, and Lefty Gomez from my example above retired after sitting in the FA pool for one year. I expect Winn will do the same.

[RonCo]

Just because I got interested, I decided to to try to quantify this a little. I went to the Lahman DB and pulled data that matched the following criteria:

Pitchers whose last season consisted of:

wins >= 10
ERA <= 5.00
Age <= 34

I then removed players who I knew for fact had left the game due to injury. This left me with 41 pitchers. My most controversail decision was to leave Sandy Koufax in the mix. I did this because he proved he could still pitch, despite pain. You can simply make the number 40 if you disagree with my assessment. Seventeen of these players are 30-34 years old.

Notes:

1) I am 100% certain there are more pitchers who left the game due to injury in this list, but the information I have at my disposal is not perfect, and I'm not going to spend a week chasing every one of these guys down. At the end of the day, if you assume 50% were injured in some fashion, you still get over 20 guys who were under 34 years old and who left the game in the season after pitching relatively "well." Twenty players in ~ 100 years is rougly 2 per decade.

2) A majority of these pitchers had careers between 1900 and 1945 (taking this into account, #(1) above becomes more like 4 per decade). This leads me to think that a majority of these are due to either (a) injury, or (b) the business element of baseball.

3) Three of the 34 year-old pitchers on the list had their last seasons in 1945--before the boys came home from Europe and the Pacific. Make of that what you will. One (Larry French) had his last season in 1942, and I believe left baseball in order to run a business considered crucial to the war effort.

Number 2(b) is one thing that we lose track of sometimes in this world of OOTP. It makes me wonder if some of these OOTP pitchers have priced themselves out of baseball--in otherwords, are they asking for too much money after being released, for example. This would essentially be retiring in order to run the family business or be a sportscaster or whatever. In that sense, it works as a simulation. Remember, in 1925, a guy was likely to be able to make a lot more money at some other business than baseball...and negotiation with the team owner wasn't all that pleasant of a chore.

Regardless, it does seem that OOTP is releasing these guys a bit too often, especially if you're running a more modern environment. Historical environments (1900-1940) are apparently subject to this more frequently in real life, and perhaps it's a law of supply-and-demand in action. During those years there were fewer major-league teams, so fewer spots. A restricted environment like that makes it more likely that decent talent would sit on the shelf in real life...just a theory.

Here is the current list of guys who fit my search criteria, sorted by age of last appearance and with the results of their last season.

Last	First	Year	Age	W	L	G	GS	ERA
Kaiserl	George	1915	22	15	15	41	29	2.24
Lush	Johnny	1910	24	14	13	36	24	3.2
Bedient	Hugh	1915	25	16	18	53	30	3.17
Fiske	Max	1914	25	12	12	38	22	3.14
Marion	Dan	1915	25	12	9	35	25	3.2
Durham	Ed	1933	25	10	6	24	21	4.48
Stratto	Monty	1938	26	15	9	26	22	4.01
Moran	Harry	1915	26	13	9	34	23	2.54
Moren	Lew	1910	26	13	14	34	26	3.55
Nops	Jerry	1901	26	12	10	27	23	4.08
Lange	Erv	1914	26	12	11	36	22	2.23
O'Neill	Mike	1904	26	10	14	25	24	2.09
William	Lefty	1920	27	22	14	39	38	3.91
Eason	Mal	1906	27	10	17	34	26	3.25
Lewis	Ted	1901	28	16	17	39	34	3.53
Hughes	Jay	1902	28	15	11	31	30	2.87
Mercer	Win	1902	28	15	18	35	33	3.04
Carrick	Bill	1902	28	11	17	31	30	4.86
Iburg	Ham	1902	28	11	18	30	30	3.89
Broberg	Pete	1978	28	10	12	35	26	4.62
Johnson	Chief	1915	29	17	17	46	34	2.75
Doheny	Ed	1903	29	16	8	27	25	3.19
Swetoni	Steve	1933	29	12	12	31	21	3.5
Koufax	Sandy	1966	30	27	9	41	41	1.73
StrickleElmer	1907	30	12	14	29	26	2.27
Cronin	Jack	1904	30	12	23	40	34	2.7
Frey	Benny	1936	30	10	8	31	12	4.25
Auker	Elden	1942	31	14	13	35	34	4.08
Malarke	John	1903	31	11	16	32	27	3.09
Suggs	George	1915	32	11	17	35	25	4.14
Douglas	Phil	1922	32	11	4	24	21	2.63
Fisher	Ray	1920	32	10	11	33	22	2.73
Donahue	Red	1906	33	13	14	28	28	2.73
Ruether	Dutch	1927	33	13	6	27	26	3.38
Denny	John	1986	33	11	10	27	27	4.2
Leonard	Dutch	1925	33	11	4	18	18	4.51
French	Larry	1942	34	15	4	38	14	1.83
Butcher	Max	1945	34	10	8	28	20	3.03
Mungo	Van	1945	34	14	7	26	26	3.2
Toney	Fred	1923	34	11	12	29	28	3.84
Seats	Tom	1945	34	10	7	31	18	4.36

[f.montoya]

RonCo, let's call it a bug now and give this to Markus. I just revisited this thread after a bit of time away and I'd say the above and the rest is more than enough to justify the need for a fix.

[f.montoya]

Tt# 3311

[f.montoya]

Now, forget the fact that in real life he retired on top...but what GM in his right mind would release Koufax after this kind of season:


Koufax Sandy 1966 30 27 9 41 41 1.73</pre>

I'd take Koufax with twice as bad of ratings over almost any other pitcher available.

[RonCo]

Originally Posted by f.montoya

Tt# 3311

I had already TTed it as TT#3298.

I agree with you on Koufax. I'm just looking at the entire picture of real-life incidents to get a feel for how good/bad the OOTP model really is. The issue we have with modelling a Koufax situation is that you have no immediate injury preceeding it that you can point to--instead, he left at the top of his game (ratings sky-high, no "injury") due to pain most likely due to an accumulation of abuse. His retirement is 100% justified in real life, but how does OOTP model that? The fact is OOTP doesn't model it directly, yet it happens. I'm betting it happens quite often, though I would have to dig deeper to understand exactly how often, and I'm sure I'm not _that_ interested.

Now, forget the fact that in real life he retired on top

Right....it's hard to forget that. Koufax is the highest profile guy to have this happen to because he was at the top. If it happens to another guy who's merely above average, it makes very little indentation in the public memory, and there's every reason to think it happens to more guys than Koufax.

The kicker here is that we've gone to great pains to get the overall career lengths "right" in v2007, using MLB data as our baseline. From an OOTP perspective, Sandy Koufax's retirement is completely unjustified. I mean--no immediate injury, no degradation of performance from whatever injuries he's sustained, just 320 brilliant IP followed by a retirement.

A secondary question is: How does the OOTP community react to a solid pitcher leaving for such an apparently unjustified reason? The answer is: unpleasantly. (note: justifiably so)

All these are just ramblings of a madman...but at the end of the day, I'm very comfortable with our overall aging algorithm. These nuances are a great example of what makes modelling career development in a baseball management simulation so interesting, and what makes Markus pull his hair out.

[Curtis]

RonCo, I have no evidence for this myself, but in debating an unrelated issue a couple of weeks ago someone raised the point that in the pre WWII era it was common for players to 'retire' (more likely be released) from the majors and go on to have a multi-year career in the minors.

Is there any way to tell if any of your 40 examples above did that? If they did, then it would make the number per decade who match what's being reported in the game even lower.

[Frank Delarocca Jr]

This isn't a problem I was having with just players before WW2. I've seen Nolan Ryan retire at age 23 after just a handful of appearances and a 2.33 ERA; Tom Seaver decide to 'hang'em up' in 1972 after winning nearly 100 games in his career to that point, winning a Cy Young with 25 wins in 1969, sent to the pen by 71 and then done; Steve Carlton just disappear after teh 73 season; etc etc

I've kind of given up on the issue for now. I'm glad to see it's TT'd and all and that this thread made it to Logged Issues, but it's just too much work to try and keep those guys from disappearing for now. I've tried weighting the values every which way possible, started new leagues, what have you, and just no dice.

Until a marked improvement, it's just a fictional game for me.

-Frank D.

[RonCo]

Originally Posted by Curtis

RonCo, I have no evidence for this myself, but in debating an unrelated issue a couple of weeks ago someone raised the point that in the pre WWII era it was common for players to 'retire' (more likely be released) from the majors and go on to have a multi-year career in the minors.

Is there any way to tell if any of your 40 examples above did that? If they did, then it would make the number per decade who match what's being reported in the game even lower.

I'm certain the root of the problem is a form of a bug rather than meeting a design goal like moving through the minors at an older age. My look at data above was an attempt to quantify the issue, not to explain it away. I've already explained what I believe are the forcing functions of this process, and I know Markus is aware of it. These dynamics have been discussed in the past in the beta areas. Markus has a couple options at his disposal. If I were to chose only one, I would properly tune the Lahman import to the newer development curves. Other betas would prefer a different set of aging curves alltogheter. Another option is to work on the AI (esp. the "stats vs. ratings" feature). What Markus will do is up to him, of course.

In the end, every user is right in listing the things they feel are the most bothersome or best things about the game. What doesnt concern you may make the game painful to me, and visa versa. Blah, blah, blah...that's all good.

[Curtis]

Originally Posted by RonCo

I'm certain the root of the problem is a form of a bug rather than meeting a design goal like moving through the minors at an older age. My look at data above was an attempt to quantify the issue, not to explain it away.

My attempt was for the same reason. What I was thinking was that if many of the real life players really did go to the minors rather than retiring, that would mean that the effect the thread originator was seeing was even more pronounced than your analysis above seemed to indicate (ie. occuring roughly two or three times as often by your analysis, but maybe four or five times as often if the alternative is taken into account), since the players in his game ARE retiring instead of demoting.

[Markus Heinsohn]

Originally Posted by f.montoya

Now, forget the fact that in real life he retired on top...but what GM in his right mind would release Koufax after this kind of season:


Koufax Sandy 1966 30 27 9 41 41 1.73</pre>

I'd take Koufax with twice as bad of ratings over almost any other pitcher available.

These are real life examples, not OOTP ones...

I checked the code and I see no issues.

In the early era, when the stuff drop is only a few points it makes a big difference because almost everyone has low stuff, so the margin is thinner, thus cases of players being released.

[TheBishop]

Originally Posted by Markus Heinsohn

These are real life examples, not OOTP ones...

I checked the code and I see no issues.

In the early era, when the stuff drop is only a few points it makes a big difference because almost everyone has low stuff, so the margin is thinner, thus cases of players being released.

If you're not seeing an issue, than you're turning a blind eye.

[Ksyrup]

Yeah, I just came back to dig this thread up because I'm seeing issues still. And this isn't with Deadball Era pitchers. I created a new historical league and simmed through 1975. Just in reviewing some recent history, I found a couple of examples where guys were either demoted or released and went untouched and then retired. The guy who was demoted (Dean Chance) was coming off of a Cy Young year (26-4, 2.57 ERA), and the next year he was in the minors. He threw 56 innings in the majors that whole year after posting a 1.97 ERA in AAA. He came back up for 2 years, was very good, got released, went untouched, and then retired. The other (Bob Gibson) had a 21-win season, got released before the next season started, went untouched, and retired.

Gibson went from 9(9)/15(16)/11(11) to 9(9)/15(15)/10(11). Chance went from 10(10)/17(17)/12(12) to 10(10)/17(17)/11(12) when he got sent down, and then from 10(10)/14(16)/10(10) to 10(10)/13(13)/9(9) when he was released. Gibson was 37; Chance was 29 then 31. Neither retired on his own; they were both released outright, sat in FA for a year, then retired.

But here's the other issue - I had the ratings weights set for 0/50/35/15. That means ratings were supposed to account for NONE of the AI's evaluation of these players. It was supposed to be totally based on stats. So if you can't see an issue with how these guys were demoted/released in general, you HAVE to see an issue with the fact that ratings are apparently playing a role in evaluation when we don't want them to.

There are issues here. Please look into this.

[Ksyrup]

Originally Posted by RonCo

The kicker here is that we've gone to great pains to get the overall career lengths "right" in v2007, using MLB data as our baseline. From an OOTP perspective, Sandy Koufax's retirement is completely unjustified. I mean--no immediate injury, no degradation of performance from whatever injuries he's sustained, just 320 brilliant IP followed by a retirement.

A secondary question is: How does the OOTP community react to a solid pitcher leaving for such an apparently unjustified reason? The answer is: unpleasantly. (note: justifiably so)

Just wanted to bump this to see if anyone is still willing to look into this issue, or if Markus' comment has essentially ended the conversation.

And also, to respond to this point...I'm all for the game including guys who hang it up at a relatively young age, on top of their game, for no good reason. It shouldn't happen too often, but it should happen. The problem is, that situation is not what we're experiencing, at all. Koufax is irrelevant to the problem that we're describing. The only common element here is that you have a pitcher whose last season is a very good one. That's it. The crucial difference is that these guys aren't leaving of their own accord, they're being released/demoted or just going into FA after their contract expires and not signed by anyone else. They sit in FA looking for a job, and then they retire. If you wanted to model the Koufax scenario, they should just announce their retirement after the season ends. But these guys do not want to retire - they want to pitch.

And finally, here's a point that I hope doesn't get lost in all of this. And that is, regardless of whether you think this is an issue or not, the results I have seen have occurred with my ratings weights set at 0/50/35/15. That means the AI is supposed to be determining what to do based 100% on stats. These moves, and a bunch of others that aren't just limited to pitchers, are not consistent with that at all. How can the AI drop a good pitcher from a team before he's thrown a pitch one year, immediately following a productive season? And how can no other team want this guy? Teams are losing 100 games each year and throwing their starts away on crap pitchers, yet a guy who won 15-20 games the year before sits in FA? That's a problem!