Player Editor Dysfunction

[luckymann]

I understand a massive set of changes have been made to player ratings for v25 and this is not a comment or report on them.

This is purely connected to players edited using the in-game editor in v25.

I present four players, all edited using the same input stats into the exact same environment (1920 MLB) using the same settings.

The first set of stats are from v24; the second from v25.

The main areas of concern here are:

1. edited pitchers coming in with unduly low STUFF ratings
2. an obscene number of triples in lieu of home runs
3. incredibly low K% for hitters

Again, this is not an analysis or commentary on the players' performances other than by way of highlighting these fairly obvious issues with how they are being generated in v25.

FWIW, other than a mild argument for Bullet's pitching being marginally too strong in v24, their results in the earlier version are where I expect and want to see them.


POINT 1: LOW STUFF RATINGS

It seems that there has been a trend with OOTP's ratings in punishing non-power pitchers in a modern context, with STUFF ratings moving lower over the past couple of programming resets. I have no real issue with that but clearly the reading these players are getting via the editor is problematic.

Bullet Rogan's K9 in v24 is 3.6; in v25 it is 1.6. Webster McDonald's drops from 4.1 to 1.5 although he is used so infrequently in v25 (almost certainly due to his super weak STUFF ratings) and virtually always as an RP that this comparison is not really valid.


POINT 1: HIGH 3B / LOWER HR

Check out those almost impossible 3B stats across the board in v25 as opposed to v24 and how they eat into HR production:

• Rogan: 364 triples in 5961 AB v 136 in 5065; 53 HR v 97
• Moore: 297 triples in 4321 AB v 111 in 3540; 20 HR v 39
• Johnson: 314 triples in 5259 AB v 107 in 5830; 49 HR v 114

For Moore, some of this can be explained away by his speed; but Johnson was a lead-foot and Rogan only moderately fast.


POINT 3: SUPER LOW HITTER K%

Again, the numbers being generated in v25 as opposed to v24 are simply inexplicable and totally unacceptable:

• Rogan: 1.2% v 4.9%
• Moore: 1.0% v 3.9%
• Johnson: 1.1% v 4.1%

Giving this the eyeball test, what seems to be to be happening is that the edits are being applied in a modern-day MLB context against an environment where strikeouts are way up for both hitters and pitchers and triples at record lows with HR at record highs.

I have no idea if that is the case or if it has something to do with the new 550 scale or the BABIP "enhancements".

All I know is that as it stands the player editor is virtually unusable in v25.

And for any "small sample size" nonsense I can only say that in a second batch of players introduced into the same environment (1920 historical MLB) I saw exactly the same thing happening.

I understand that this facet of the game isn't used by many people and could therefore be considered "niche". But all the same, for many of us it is the only way to get the NeL players how they should be. On that basis alone, given the sluggish "progress" in this regard, I would hope this issue gets treated with the commensurate level of attention by the devs and higher-ups. Check around the boards and you'll see an increasing amount of chatter about the NeL and so I don't think this is as "niche" as it may seem.

If you need any other contextual info let me know. These issues only seem to be affecting edited players, not game-created ones and I have only done this in the historical environment I mention so I can't speak for custom games or the like.

Thanks

G

[Lukas Berger]

You really don't want to look at the default editor stats for this, as it will tell you basically nothing useful in a league with as extreme values as the Negro Leagues.

If you look at the new lines in the editor for 25 that give expected output stats in the actual league the players are in, that will give a much better picture of what to expect, and you can then edit the players based on those values.

The triples being too high in some cases issue is a separate thing though, we'll fix that up in one of the next patches.

[luckymann]

Originally Posted by Lukas Berger

You really don't want to look at the default editor stats for this, as it will tell you basically nothing useful in a league with as extreme values as the Negro Leagues.

If you look at the new lines in the editor for 25 that give expected output stats in the actual league the players are in, that will give a much better picture of what to expect, and you can then edit the players based on those values.

The triples being too high in some cases issue is a separate thing though, we'll fix that up in one of the next patches.

Hey bud, appreciate the reply as always although I'm kind of lost.

I'm talking about the results I am getting from the edited stats I am entering, not the performance of these guys as they come into the game raw.

If I am entering the same stats and the environment containing these players is exactly the same, I don't see how this explains this massive variation in ratings generation and, more importantly, player performance from 24 to 25.

I understand that the new programming results in different ratings than would be those generated for the v24 version of a player, but isn't the goal of all this the statistical performance? Aren't the ratings just the means to an end, with that end being how the player(s) perform in game?

Or am I missing something?

G

[luckymann]

OK I am going to persist with this because, well, I'm annoyed and keen to get to the bottom of it.

Stepping away from players that have in any way been edited, let's look at Dizzy Dean in the same save, which sits in the offseason at the end of 1937.

So the 1938 initial recalc has been done. This save is a 1-year recalc historical save with basically all default settings applied.

Here is how Dizzy Dean looks - scouting at 100% - ratings / real stats / editor:

So from what has been said above, Dizzy should be expected to perform as per the "expected current" output in the editor.

Those stats are miles off his historical stats for 1938. Not even close. Why would this be? Why, in a 1-year recalc historical save with almost complete default settings, wouldn't his "expected current" stats be his historical stats for the year in question, give or take a bit for variance?

While we're at it, why the regression in his CURR / POT ratings for his various pitches? That has bothered me for years.

Thanks

G

[Lukas Berger]

Sorry for the slow reply, been sick.

One thing that will affect what you're seeing with the expected output of the stats, as well as the real performance, is the league environment as a whole.

So Dean for example, if you import using a regular historical league with one-real recalc, you would expect to see the expected performance (and the actual performance as well) somewhat around what he did in real life, depending on exactly what import settings you used when creating the league.

But, now say you take a regular 1938 league and then you add in all the Negro League players to that league as well. That massively inflates the overall talent of the league, and thus Dean's expected performance will be much worse than in real life, because now he's in a much better league than real-life.

Or the same even if you're not adding in extra players, but are just rerating players that are already in the league. As you adjust the ratings of guys in the league, that will raise of lower the baseline, and thus affect the expected performance of even guys you haven't edited.

The same is true even if you don't edit anyone, but do something like significantly adjust the make worse settings when you're setting up the league.

So basically, if you have a league with much higher or lower talent levels than the default league would have, for any reason, then players will not perform to real-life expectations without your adjusting their ratings to match the overall raised/lowered level of the league as a whole.

Does that make sense? Hope I'm explaining that well enough, still a little foggy after being sick, so if I'm not clear, please just let me know

[luckymann]

Originally Posted by Lukas Berger

Sorry for the slow reply, been sick.

One thing that will affect what you're seeing with the expected output of the stats, as well as the real performance, is the league environment as a whole.

So Dean for example, if you import using a regular historical league with one-real recalc, you would expect to see the expected performance (and the actual performance as well) somewhat around what he did in real life, depending on exactly what import settings you used when creating the league.

But, now say you take a regular 1938 league and then you add in all the Negro League players to that league as well. That massively inflates the overall talent of the league, and thus Dean's expected performance will be much worse than in real life, because now he's in a much better league than real-life.

Or the same even if you're not adding in extra players, but are just rerating players that are already in the league. As you adjust the ratings of guys in the league, that will raise of lower the baseline, and thus affect the expected performance of even guys you haven't edited.

The same is true even if you don't edit anyone, but do something like significantly adjust the make worse settings when you're setting up the league.

So basically, if you have a league with much higher or lower talent levels than the default league would have, for any reason, then players will not perform to real-life expectations without your adjusting their ratings to match the overall raised/lowered level of the league as a whole.

Does that make sense? Hope I'm explaining that well enough, still a little foggy after being sick, so if I'm not clear, please just let me know

Sorry you're feeling poorly mate, hope you feel better soon and thanks for replying.

The issue here for me is this. I see what you are saying but the scenario I have explained in the post above is none of the things you mention. It is a straight up, mainly default settings historical save with no minors nor NeLers currently in the offseason before the 1938 MLB season.

In which case, as you yourself are saying, his "expected" rate in the editor should fairly closely align with his historical stats and then generate stats for that season when it is simmed at or very close to that mark.

Clearly the screenshot above shows this simply isn't the case.

Which suggests to me something in the editor - not in the players as the game creates them (I bet he ends up going close to his IRL stats when it is simmed), but rather in the editor itself - is broken. I highly suspect is has something to do with the partial change to 1-550 ratings and perhaps also the BABIP being now parsed out instead of hidden and calculated out of sight.

I could well be wrong in this assumption but that notwithstanding I can't believe something isn't going haywire in there to be causing this massive disconnect - both that screenshot and what I saw and have explained in great length (ignoring the triples problem) in the OP would back me up in that regard, I think.

G

[Lukas Berger]

Originally Posted by luckymann

The issue here for me is this. I see what you are saying but the scenario I have explained in the post above is none of the things you mention. It is a straight up, mainly default settings historical save with no minors nor NeLers currently in the offseason before the 1938 MLB season.

In which case, as you yourself are saying, his "expected" rate in the editor should fairly closely align with his historical stats and then generate stats for that season when it is simmed at or very close to that mark.

Clearly the screenshot above shows this simply isn't the case.

Which suggests to me something in the editor - not in the players as the game creates them but in the editor itself - is broken. I highly suspect is has something to do with the partial change to 1-550 ratings and perhaps also the BABIP being now parsed out instead of hidden and calculated out of sight.

I could well be wrong in this assumption but that notwithstanding I can't believe something isn't going haywire in there to be causing this massive disconnect - both that screenshot and what I saw and have explained in great length (ignoring the triples problem) in the OP would back me up in that regard, I think.

G

I may be misunderstanding you here, and if so, my apologies.

Is Dean in a different league from the Negro League guys you initially posted?

Because I see that they're on MLB teams and you also mention in your post from yesterday that Dean is in the same save as the NeL guys.

Maybe the easiest thing here would be if you could upload or post a link to the league and I can take a look at it so I can see exactly how things look in it rather our have to chat at length clarify everything about the environment. (Will probably be early next week before I can get to checking it out).

[luckymann]

Originally Posted by Lukas Berger

I may be misunderstanding you here, and if so, my apologies.

Is Dean in a different league from the Negro League guys you initially posted?

Because I see that they're on MLB teams and you also mention in your post from yesterday that Dean is in the same save as the NeL guys.

Maybe the easiest thing here would be if you could upload or post a link to the league and I can take a look at it so I can see exactly how things look in it rather our have to chat at length clarify everything about the environment. (Will probably be early next week before I can get to checking it out).

It is the same save but there were only ever four NeL players in the league (two pos players, one two-way player and one pitcher) and they have now all retired, so their effect would have to be minimal overall and approximately zero now, no?

Look, I think you know me well enough by now to know that I wouldn't be pushing back on this so hard if I didn't honestly believe there is, one way or another, a serious issue here. And, all due respect, nothing you have countered with has so far allayed my fears in this regard.

It might not be that there is a problem at your end, I am completely open to that possibility. Perhaps the new ratings system needs a change in input methodology and some tinkering with my way of going about things.

What I do know is that the process for editing imported players that I have been using without any serious issues and with great performative in-game success since v23 now no longer seems to work for me in v25.

So all I am asking is that someone on the dev team go in there and try to edit an imported player and see if they are able to get the result they are seeking.

If the answer is "no" then I guess you guys need to go in there and fix it.

If the answer is "yes", then I'd love some insight into why I can't get it to do the same and, more importantly, what I need to do in order to do so.

Thanks bud.

G

[Lukas Berger]

Been chatting with Matt and there might be a few things going on here.

One, the editor will be a bit less reliable for predicting ERAs the further you get from the modern era, as some things, like error rates, are not adjusted per league environments.

With Dean, there's also kind of a bit of weirdness in his real-life stats, in that Dean had a higher ratio of unearned runs in 1937 than expected. League-wide the gap is 18% (12.5k ER vs 14.8k R). Dean was 28% (59 vs 76).

EDIT: Just realized you might be looking at Dean with 38 ratings, not 37. Anyway, if that's the case, that's even more a situation where we'll never be able to give him the 'correct' ERA along with the correct rate stats, because his actual rate stats in 38 were pretty bad to mediocre, even for the era, but his ERA was insanely low. Haven't dug into things to see why that was, but I would guess he got even luckier with sequencing of earned runs or something like that than he did in 37).

So if we're "correct" on Dean, based on his other rate stats, which we more or less seem to be, we're still pretty much usually going to be high on his ERA because it's highly unlikely any given sim will recreate his real-life sequencing that lead to those 'extra' unearned runs and thus the lower than expected ERA.

So to some extent you may need to make allowances here. Some pitchers may not be able to hit both their real-life ERA and their real-life rate stats. So you might need to adjust what you're entering so that if f.e., you want a player to hit a target ERA that in real-life was lower than he is being projected for in the editor, you might need to boost his rate stats above what they really were. Or vice-versa.

There may be some more stuff playing into this on our end too, as well as some more you can change when entering. This is just a quick response. We'll try to look at things here a little more and see if there's more going on here.

[Lukas Berger]

Overall though, I would not recommend editing historical players necessarily based on them the input stats you want them to have, because the actual league environment will affect and change the results.

Ultimately, I think you'll get results more like what you're would want if you look at the expected stats line, and then simply adjust the ratings until you get an expected stats line that is in the ballpark of what you're looking for, rather than using the stats input to reverse engineer and create the ratings. That is unlikely to work as well in 25 as it did in previous versions, given how the engine has changed.

[luckymann]

Originally Posted by Lukas Berger

Overall though, I would not recommend editing historical players necessarily based on them the input stats you want them to have, because the actual league environment will affect and change the results.

Ultimately, I think you'll get results more like what you're would want if you look at the expected stats line, and then simply adjust the ratings until you get an expected stats line that is in the ballpark of what you're looking for, rather than using the stats input to reverse engineer and create the ratings. That is unlikely to work as well in 25 as it did in previous versions, given how the engine has changed.

OK, really appreciate the detailed responses Lukas, so thanks for taking the time to do so. For now, I'm going to leave it at that then have a bit of a play around with importing more players on 25 and come back to you with what I am seeing.

I will be honest with you, however, I'm not sure what you're saying here covers the thing even close to entirely. And the first comment above makes little sense if you are merely trying to add a historical player to a straight-up MLB environment. I just don't understand how if you use stats from the historical year in a universe that is trying to replicate that same year as closely as possible why there's anything other than rounding-error slippage. It is completely counterintuitive to how I see things.

If you go back over all the evidence I have presented here - the wacky low K%, the 3B in lieu of HR, the really poor stuff ratings being generated and the effect that has on usage - I feel confident you'll find a breakage in there of some sort, with my best guess being the application of modern standards on historical stats from earlier eras or something along those lines; or perhaps as I mention some sort of transcription error with the new ratings scale or BABIP settings.

But like I said, I'll leave this for now and come back when I have more evidence one way or the other.

Have a good weekend

G

[Lukas Berger]

Originally Posted by luckymann

OK, really appreciate the detailed responses Lukas, so thanks for taking the time to do so. For now, I'm going to leave it at that then have a bit of a play around with importing more players on 25 and come back to you with what I am seeing.

I will be honest with you, however, I'm not sure what you're saying here covers the thing even close to entirely. And the first comment above makes little sense if you are merely trying to add a historical player to a straight-up MLB environment. I just don't understand how if you use stats from the historical year in a universe that is trying to replicate that same year as closely as possible why there's anything other than rounding-error slippage. It is completely counterintuitive to how I see things.

If you go back over all the evidence I have presented here - the wacky low K%, the 3B in lieu of HR, the really poor stuff ratings being generated and the effect that has on usage - I feel confident you'll find a breakage in there of some sort, with my best guess being the application of modern standards on historical stats from earlier eras or something along those lines; or perhaps as I mention some sort of transcription error with the new ratings scale or BABIP settings.

But like I said, I'll leave this for now and come back when I have more evidence one way or the other.

Have a good weekend

G

Yeah, it's certainty possible something is off here, I don't discount that at all.

The 3b's I think are a separate issue which this patch has hopefully adjusted, but even with the other stats, there's enough change between the eras, and usage does factor into things enough that it's entirely possible we're making a error here or something is not fully adjusted correctly.

We'll definitely take a closer look here and see what things look like, and if there's a need to make some changes, we'll do that.

[luckymann]

Originally Posted by Lukas Berger

Yeah, it's certainty possible something is off here, I don't discount that at all.

The 3b's I think are a separate issue which this patch has hopefully adjusted, but even with the other stats, there's enough change between the eras, and usage does factor into things enough that it's entirely possible we're making a error here or something is not fully adjusted correctly.

We'll definitely take a closer look here and see what things look like, and if there's a need to make some changes, we'll do that.

That's all I can ask for.

Many thanks mate.

G