Definitely Something Wrong with the Simulation

[Painmantle]

Originally Posted by Lou Gehrig

In all due respect, "definitely" is a pretty strong word. If it was "definitely" don't you think the experienced players in this forum would be yelling from the highest roof about it?

Absolutely not, OOTP is rigged and there is a conspiracy by everyone who plays the game to ruin the OP's experience and make him look crazy. It's happening to us too, we just can't say anything about it because of the "non- disclosure agreement"

[Otrex]

Originally Posted by greenOak

No offense, but nobody here is taking you seriously nor should they. You don't seem to understand how far variance can and does swing in one direction or the other. You're evidence is severely lacking sample size. Run ~1000 seasons and if you still have your problem, then maybe we can start talking.

Hmm... the obvious troll is well, obvious, but I'll indulge you for a moment.

My background is in economics and I've done a bit of work in data regression models. My thesis demonstrated, with 95% confidence, the differences between reported and actual results of osteoarthritis in the Canadian workforce. It called on a health study in which there were about 22,000 valid responses... but truth be told, I obtained nearly the same result with just 500 responses randomly picked from the data field. Truth be told, scientists will sometimes accept a sample size of n=30 as sign that further research is necessary. I might not completely understand variance, but I wonder how much you do as well.

When you say "run 1000 seasons" are you basing that on experience working in data regression, or are you trying to provide me with an impossible task so as to prove the validity of your point? I'm going to say it's the latter. You might as well have said "Hey everyone, look at his silly face! Let's not allow him on the jungle gym until he stops looking so silly!"

I've claimed a statistical turnaround within a short period of time that is repeatable and with the only changing factor being the declaration to the program that I am running a particular team (and perhaps a 10% change to rosters, selected at random). That includes taking a 118-game loser and making them into a World Series champion within two seasons with no input of any kind. I believe that is unprecedented historically. Therefore, even with the small sample size of just 9 seasons, I continue to claim that there is a reason to suspect an issue, given that, upon demand, I can jump to the worst team in the league and duplicate those results.

Maybe it's a result of my settings? Maybe it can be fixed with a check-mark being placed in a box that I have not noticed? Great, let's find a way to have that addressed. But if the best you've got is the adolescent version of "Don't listen to him! He's silly!" along with the mis-use of the word "you're", I think I know who I'd pay attention to if I were the developer.

[Otrex]

Originally Posted by Lou Gehrig

In all due respect, "definitely" is a pretty strong word. If it was "definitely" don't you think the experienced players in this forum would be yelling from the highest roof about it?

Let's just say I've played perhaps more than a couple games of OotP. You can reference my join date if you wish.

[BIG17EASY]

Originally Posted by Otrex

Hmm... the obvious troll is well, obvious, but I'll indulge you for a moment.

My background is in economics and I've done a bit of work in data regression models. My thesis demonstrated, with 95% confidence, the differences between reported and actual results of osteoarthritis in the Canadian workforce. It called on a health study in which there were about 22,000 valid responses... but truth be told, I obtained nearly the same result with just 500 responses randomly picked from the data field. Truth be told, scientists will sometimes accept a sample size of n=30 as sign that further research is necessary. I might not completely understand variance, but I wonder how much you do as well.

When you say "run 1000 seasons" are you basing that on experience working in data regression, or are you trying to provide me with an impossible task so as to prove the validity of your point? I'm going to say it's the latter. You might as well have said "Hey everyone, look at his silly face! Let's not allow him on the jungle gym until he stops looking so silly!"

I've claimed a statistical turnaround within a short period of time that is repeatable and with the only changing factor being the declaration to the program that I am running a particular team (and perhaps a 10% change to rosters, selected at random). That includes taking a 118-game loser and making them into a World Series champion within two seasons with no input of any kind. I believe that is unprecedented historically. Therefore, even with the small sample size of just 9 seasons, I continue to claim that there is a reason to suspect an issue, given that, upon demand, I can jump to the worst team in the league and duplicate those results.

Maybe it's a result of my settings? Maybe it can be fixed with a check-mark being placed in a box that I have not noticed? Great, let's find a way to have that addressed. But if the best you've got is the adolescent version of "Don't listen to him! He's silly!" along with the mis-use of the word "you're", I think I know who I'd pay attention to if I were the developer.

Or maybe it's an anomaly.

[Ruthian23]

Originally Posted by Otrex

Hmm... the obvious troll is well, obvious, but I'll indulge you for a moment.

My background is in economics and I've done a bit of work in data regression models. My thesis demonstrated, with 95% confidence, the differences between reported and actual results of osteoarthritis in the Canadian workforce. It called on a health study in which there were about 22,000 valid responses... but truth be told, I obtained nearly the same result with just 500 responses randomly picked from the data field. Truth be told, scientists will sometimes accept a sample size of n=30 as sign that further research is necessary. I might not completely understand variance, but I wonder how much you do as well.

But unless I'm missing something, your sample size isn't 500 or even 30. It's five, so not even at the threshold you yourself indicate is the minimum for scientists to suggest that further research is necessary.

[Otrex]

Originally Posted by I♥Pepper

I think that the sample size is nearly impossible to determine in this case. To determine sample size, one has to know the size of the population. What is the size of the population in the o.p.'s example? The number of total decisions the a.i. can make over the course of a season? Nearly impossible to determine me thinks.

Instead, it becomes more of a subjective determination based on personal preference. Would you allow your star hitter 1000 hitless at bats before trading him? Probably not. More like about 50, maybe. Or maybe more or less based on other factors.

This line of thought can continue into the OP's example. I think there are way to many variables involved in the starting conditions of the example given. Before I made a personal judgement on whether or not my presence was influencing results even without any input from me, I would need to stabilize the simulation. Or, at least define parameters. For example, under what starting conditions should I expect my team to win the division 9 times out of 10? What conditions should I expect my team to loose 9 times out of 10. Then apply those starting conditions and run tests. Examples of factors to consider are : Can players evolve? Can players be traded? Do coaches evolve? Can coaches be hired and fired? Are injuries allowed? Rainouts? Team strategy settings? Substitution patterns? Fatigue? And many more. It's a complicated process that requires a little more control than just 5 random seasons of results, at least it would for me.

I agree with your well-reasoned line of questioning. Yes, it's difficult to prove my assessment in a statistical sense without somehow freezing all results in time and running the exact same simulation again with the only change being the manager switching from team-to-team.

If my experience is not consistent with others playing the game, then I suspect it's a settings issue, though in this case I cannot ascertain where the issue might be because I have used settings suggested by other veterans of the random debut feature. You would think that using historical numbers would add EXTRA stability to the simulation results, but so far, that doesn't appear to be the case since, as I noted briefly, Ray Jablonski managed to hit 52 homeruns in one of the simmed seasons despite a historical career high of 21.

I am running with a 16 team league, split up into two leagues of 8 with 1st and 2nd place having a best-of-7 playoff before entering the world series. I allow players from 1901 through 1981 to participate and I have "Neutralized Stats" turned on.

Speaking of neutralized stats, my understanding is that is meant to be compensation for ballpark differences. Is that correct? I also notice that neutralizing adds in missing seasons for players, which is a very good idea for those WWI and WWII players whom deserve better career totals (it still doesn't give Christy Mathewson the career he deserves, but that's understandable).

I am indeed hoping that someone can show me the one checkbox I've not checked that would fix this issue, but with results this peculiar I stand by my claim of "definitely wrong" realizing that yes, even at 1 in 30 million, someone has to win the lottery (though you won't catch me playing).

[Otrex]

Originally Posted by BIG17EASY

Or maybe it's an anomaly.

I would like to think so, but I have obtained the same results more than once, where the worst team in the league becomes the champion very quickly. At some point an anomaly, if repeatable, has to be considered to be a "ghost in the machine".

[Otrex]

Originally Posted by Ruthian23

But unless I'm missing something, your sample size isn't 500 or even 30. It's five, so not even at the threshold you yourself indicate is the minimum for scientists to suggest that further research is necessary.

That's fair, assuming that n = "the number of seasons simmed". I would suggest that with so many players in the league, the value of a simmed season is quite a bit higher than n=1.

That being said, I ran more seasons last night, selecting the worst team in the league again, and they too won the Series within 2 years. The team I just left sunk back to a 100 game loser within 2 years. Perhaps not statistically significant, but a cause for concern none-the-less.

I am up to 10 seasons now, for what that's worth. There have been 3 batting triple crowns in that time. Lefty Grove joined the league, which is playing to 1993 statistics, and fanned 477 batters in 274 innings.

[Ruthian23]

Originally Posted by Otrex

That's fair, assuming that n = "the number of seasons simmed". I would suggest that with so many players in the league, the value of a simmed season is quite a bit higher than n=1.

That being said, I ran more seasons last night, selecting the worst team in the league again, and they too won the Series within 2 years. The team I just left sunk back to a 100 game loser within 2 years. Perhaps not statistically significant, but a cause for concern none-the-less.

I am up to 10 seasons now, for what that's worth. There have been 3 batting triple crowns in that time. Lefty Grove joined the league, which is playing to 1993 statistics, and fanned 477 batters in 274 innings.

For the record, I haven't been seeing this at all in the current (or any past) version. In one season, my team was projected in the pre-season to win 90+ games and, without any significant injuries, started the year like 12-30.

Your Lefty Grove striking out 477 in 274 innings example using 1993 statistics does seem a little wonky, but perhaps that's something with your settings?

[Zorro]

Originally Posted by Otrex

Lefty Grove joined the league, which is playing to 1993 statistics, and fanned 477 batters in 274 innings.

Wow, He is a keeper. Did he lead the league?......................................... ........Z

[Ruthian23]

Originally Posted by Zorro

Wow, He is a keeper. Did he lead the league?......................................... ........Z

That's actually a good question. Are you seeing other strikeout numbers similarly high, or did Grove lead the league by like 200+Ks?

[Otrex]

Originally Posted by Zorro

Wow, He is a keeper. Did he lead the league?......................................... ........Z

Fortunately, yes, and by a wide margin. Next closest was Steve Carlton at somewhere in the 279 range with several stacked up right behind him in a more predictable pattern (Addie Joss, etc).

Also interesting, a 22-year old Bob Feller has a career ERA of almost 5.00 and cannot seem to make things work. The AI has placed him in the bullpen now.

And Bob Horner, for several seasons, was nearly unstoppable. He won 2 of the 3 triple crowns I mentioned earlier. Yes, that Bob Horner.

[David Watts]

[QUOTE=Otrex;3841225]That's fair, assuming that n = "the number of seasons simmed". I would suggest that with so many players in the league, the value of a simmed season is quite a bit higher than n=1.

That being said, I ran more seasons last night, selecting the worst team in the league again, and they too won the Series within 2 years. The team I just left sunk back to a 100 game loser within 2 years. Perhaps not statistically significant, but a cause for concern none-the-less.

I am up to 10 seasons now, for what that's worth. There have been 3 batting triple crowns in that time. Lefty Grove joined the league, which is playing to 1993 statistics, and fanned 477 batters in 274 innings.[/QUOTE]

Just curious, did you edit the "make bad" options when you set up your league? I'm talking about the options to make players with limited at bats/innings pitched bad.

[Ruthian23]

Originally Posted by Otrex

Fortunately, yes, and by a wide margin. Next closest was Steve Carlton at somewhere in the 279 range.

Also interesting, a 22-year old Bob Feller has a career ERA of almost 5.00 and cannot seem to make things work. The AI has placed him in the bullpen now

You might want to report the Grove thing as a bug. They might want to know all of your settings and if you've made any changes to Grove, but I can't believe a guy striking out almost 500 batters in 1993 and leading the league by almost 200 Ks is working as intended unless you maxed out his abilities in commish mode or something.

As for Feller, I don't think that one's an issue. As long as you have player development on, guys will boom or bust. He probably took some ratings hits and just isn't going to become Bob Feller.

[Otrex]

Yes, I increased both the "adjust" and "weaken" numbers for the make-bad assessment. I can probably look up the exact figures if that would be helpful.

[Otrex]

Originally Posted by Ruthian23

You might want to report the Grove thing as a bug. They might want to know all of your settings and if you've made any changes to Grove, but I can't believe a guy striking out almost 500 batters in 1993 and leading the league by almost 200 Ks is working as intended unless you maxed out his abilities in commish mode or something.

As for Feller, I don't think that one's an issue. As long as you have player development on, guys will boom or bust. He probably took some ratings hits and just isn't going to become Bob Feller.

Agreed on Feller, though when he first came up (at 20y/o I think) he was projected to be a two-star. 4 seasons in, when he still wasn't doing well, he had a major injury, and now he's terrible. BUT, I think that's actually a reasonable outcome as you suggest.

No modifications to Grove at all. Let me know if I can check any league settings to help narrow this down.

[Otrex]

Can anyone confirm if there is any normalizing of era stats going on in the game, or is it just neutralization of park bias?

I feel like it's the former, since Wildfire Schulte hits 40+ each year as you would probably expect under a normalization routine.

The Bob Horner thing though, I cannot explain.

[David Watts]

Originally Posted by Otrex

Yes, I increased both the "adjust" and "weaken" numbers for the make-bad assessment. I can probably look up the exact figures if that would be helpful.

There's your issue. I messed with those one time in the past and ended up with video game like results. I think in my game Herb Score whiffed 400 batters.

[Otrex]

So, you're saying that I should just leave those numbers as default? That seems strange because I made them "tougher", though only perhaps 20% so.

Just wanted to avoid having situations where a pitcher with 10 innings gets to be rated 5-stars, but you're saying it breaks the stats? Just so I understand, what is your theory behind that?

Speaking of which, I assume the weaken and adjust settings are expressed by the season? That is to say, if a player pitches less than 10 innings as a reliever in one season, he will be weakened? As opposed to 10 innings in a career.


EDIT - I assume of course, that by "there's your issue" you're referencing the Lefty Grove issue as opposed to the original complaint? Or do you think that's somehow related too?

[Padreman]

I made the playoffs 23 times in my league but currently in a 13 year skid in last place so like realm life it comes and goes.
Oh I made the playoffs 23 times but one won the title 4 times. This in my long running fictional league