Professor Luckymann’s Lab NeL Project

[luckymann]

[luckymann]

[luckymann]

[luckymann]

[luckymann]

Think I've finally got the formatting how I want it - only took twelve goes at it!

[luckymann]

[luckymann]

[luckymann]

[luckymann]

[Makonnen]

Somehow missed this. It's hard, right? To really integrate, you pretty much have to expand very quickly. Or else there are just too many historical players who never reach MLB.

I do think https://i9s.org/wp/available-players-chronological/ should be helpful, especially if you revisit the early years.

My plan is to integrate these into the actual historical data the game uses--I had very good success with that on a limited run, and had begun to write the scripting that took the raw stats and converted them into OOTP compatible data. Those could then be merged into the game database and reimported.

That has to, I think, be the ultimate goal: at that point, the game engine is working off reasonable MLE's as it does its thing.

But I plan to do it in "sets," with the first either covering just the pre 1900 folks or, more likely, the players from pre 1900 to 1910. I've been lax on updates for a couple weeks, but if you look at https://i9s.org/wp/most-recent-player-updates/ , you'll see a much greater rate of progress than previously on this stuff. Fingers crossed.

[luckymann]

Originally Posted by Makonnen

Somehow missed this. It's hard, right? To really integrate, you pretty much have to expand very quickly. Or else there are just too many historical players who never reach MLB.

I do think https://i9s.org/wp/available-players-chronological/ should be helpful, especially if you revisit the early years.

My plan is to integrate these into the actual historical data the game uses--I had very good success with that on a limited run, and had begun to write the scripting that took the raw stats and converted them into OOTP compatible data. Those could then be merged into the game database and reimported.

That has to, I think, be the ultimate goal: at that point, the game engine is working off reasonable MLE's as it does its thing.

But I plan to do it in "sets," with the first either covering just the pre 1900 folks or, more likely, the players from pre 1900 to 1910. I've been lax on updates for a couple weeks, but if you look at https://i9s.org/wp/most-recent-player-updates/ , you'll see a much greater rate of progress than previously on this stuff. Fingers crossed.

Yeah, once you get the player stats right the two major considerations are usage and overall statistical output.

No doubt, as we're going to see in this test from here on in with the big influx just about to get underway, for NeL players to be included the 16-team MLB structure is insufficient to hold them all without some pretty sizeable flow-on effect for the MLB guys.

To expand with 2 or maybe 4 teams solves this but creates other problems in turn, especially with the well-documented issues of AI roster management.

Plus, if the NeL players are eventually able to be used without having to invoke full minors, as they should be at least from 1920 onward, the standard MLB setup simply doesn't provide enough players at a number of positions - SP / C / SS / CF mainly - to support the extra teams under full functionality and allowing all of the various game options. Even with the addition of NeL guys.

So you're damned if you do (expand to 18-20 teams) and damned if you don't.

The other thing is that a whole new set of LTMs MUST be designed to allow the MLB guys' performances not to be hit heavily by the introduction of the NeLers. Whatever shine one most would get from seeing these guys in the game would be dulled greatly if Babe Ruth ends up hitting 400 career HR.

So I continue to advocate a totally holistic approach to this. Just introducing the NeL guys will create as many new problems as it will solve old ones if they are simply chucked in without these other factors taken into consideration.

All this exercise is trying to do is first and foremost check my methodology for the NeL players in a game setting but also look at the effect their introduction has on the standard MLB OOTP game. Whether or not the devs and those who might at some point hopefully be tasked with finding a way to do this even look at this thread is out of my control and certainly not something I am taking into consideration.

G

[luckymann]

[Makonnen]

All makes sense. Seems like you're more posting results here than your background work, but if you ever want to chat about a particular player, I'll follow here, or eml me.

I love Eric's work, and it has some very deep flaws in particular use cases.

How are you handling 2-way players? It's a pretty tricky question, imo, culminating with Rogan who (more than Radcliffe) really was, I think, Ohtani-before-Ohtani.

[luckymann]

Originally Posted by Makonnen

All makes sense. Seems like you're more posting results here than your background work, but if you ever want to chat about a particular player, I'll follow here, or eml me.

I love Eric's work, and it has some very deep flaws in particular use cases.

How are you handling 2-way players? It's a pretty tricky question, imo, culminating with Rogan who (more than Radcliffe) really was, I think, Ohtani-before-Ohtani.

Thanks bud, I will take you up on that sometime over the next few steps to fill some gaps and get your thoughts on a couple things.

2WP aren't much of a hassle for me here as there are only a handful EC treats as such. Buckner has been in from the start and Pedroso for a while now. That leaves just Rogan, Dihigo and the great Lazaro Salazar.

Radcliffe doesn't make the cut here and is the one I've had issues with previously as OOTP cannot quite get its "head" around someone pitching and catching, so he usually just becomes a pitcher pretty quickly and stays that way.

[luckymann]

[luckymann]

[luckymann]

This is of course, the sort of "quantum-leap" point in the timeline for NeL players as their historical stats for the most part start being assessed as MLB-level, and there are more of them so that the "make-bads" have less of an effect.

For this exercise, there are also a number of considerations. Usage is a big issue - there are only so many positions and teams across the MLB, so with this big influx of quality all players will be hit for how they are used.

Statistical output is the other thing to keep in mind. LTMs mean there are only so many counting stats to go around and that makes for compressed leaderboards with all these "extraneous" players now in the league.

The way around this is, of course, to add more teams. The LTMs pro-rata out as per how many teams are in the league. But with historical minors on, I'd have to reset this every year and I don't want to do that. So I'll just have to tolerate the inevitable regression to the mean.

For anyone who is manually adding NeLers to the historical MLB setup, however, I recommend you add 2 teams sometime before 1920 if just using the major league level and even up to 4 if using historical minors.

G

[luckymann]

[luckymann]

Just a quick look at how all that talent affected the overall performance of the league.

Here are the respective hitting and pitching leaders and stats for the two tests:






To be honest, the opposite of what I expected happened with outlier performances more extreme in the Treatment. Normally the application of LTMs has a compression effect that regresses stats to the mean to varying degrees. That hasn't happened as much here, with Ruth the obvious standout.

Comparing this to the historical, it seems that the NeLer counting stats are being taken from the second-tier players. Cy Williams only had a handful of HR in the Treatment vs 17 IRL, while Tillie Walker and George Sisler also saw their totals decrease a bit.

Then again, Babe did lose 14 HR in the Control sim, which equates to 26% so perhaps somewhere in the midpoint of the two groups is where we should be looking. That gives him 46 HR, or a drop of 15%. Career-wise that would knock him down to a smidge over 600, fairly noticeable.

The other big counting stat for consideration is pitcher strikeouts. Clearly, with the influx of so many hotly-rated pitchers at once decimates the K leaderboard for white guys. Pete Alexander led the MLB IRL in 1920 with 173. In the sims, he had 95 in the Control and 102 in the Treatment. That is a massive drop. (Oddly, Walter Johnson's total in both increased on the historical.)

Of course, small sample size yada yada yada but I think you'll find these results will hold in the macro sense, even if the various players change, over a larger number of iterations.

Will be very interested to see what effect, if any, the new LTM protocol has on this sort of situation.

[luckymann]