The AI used GL Defense 3 times in the 221 formation (Strong I Big TE Post). In my opinion, this is perfectly reasonable.
Conclusion: The AI uses GL Defense when it makes sense, provided that full-block plays (30/30) are activated without any rules.

My reason for asking that was to determine if AI routinely used it against standard offences; clearly not.

Grandad's concern is therefore solely about human owners doing so.

Since your test was with no rules present, maybe when human abandons team, then delete any rules set for that team, leaving a clean slate? Would prevent new owners keeping any GL inappropriate usage, would prevent AI using those rules. Would reduce cases.

If a human owner wants to run GL offence all day long, then that should be their choice - with consequences. Question is how? Artificially penalise the attributes of players when the play isn't in a situation where the offence would be reasonable - say X & 1 ytg anywhere on pitch, X & 2 or less ytg by opposition goalline?

Should similar occur for GL defence then - if not against 23, 31 or 22 offensive sets, then heavily penalise?

Rules don't remain when a user abandons their team - the AI generates a fresh gameplan, depth chart, etc.

The defense weighs each play but always tries to find a play to match the offensive formation - so a goal line offense will try to choose a goal line defense to counter.

Good to hear - always thought rules remain for some reason.

So the issue Grandad raised would only apply to human owners then.

What is not reasonable is the Ai using GL offense in non GL situations, or better put, not more of a simulation of RL NFL football, its less of a simulation. 1st and 10 on own 20 is not a GL situation, or short yardage. I can see using the 2rb3te (GL) on very short yardage situations, but, that will auto the GL defense every time if there is a GL defense play activated. From most recent game vs AI managed KC Chiefs...........

1-10-KCY 46 (8:33) 40-Stephen Winget ran to KCY 45 for -1 yards. Tackle by 94-Moises Camacho .
Offensive Play: Goal Line Normal HB Dive
Defensive Play: Goal Line Attack #3

3-5-KCY 30 (5:13) 37-Thomas Prickett ran to KCY 34 for 4 yards. Tackle by 33-Deandre Peck .
Offensive Play: Goal Line Normal HB Dive
Defensive Play: Goal Line Attack #3

2-10-KCY 7 (9:37) 37-Thomas Prickett ran to KCY 6 for -1 yards. Tackle by 94-Moises Camacho .
Offensive Play: Goal Line Normal HB Outside Strong
Defensive Play: Goal Line Attack #3
Run Key

Here's the only time it was used in a 0-2 yard situation ...

3-1-KCY 36 (7:45) 40-Stephen Winget ran to KCY 34 for -2 yards. Tackle by 2-Robert Carter .
Offensive Play: Goal Line Normal HB Weak
Defensive Play: Goal Line Attack #3
Run Key

GL #3 was the only GL defense that I had active.

If a GL offensive play is active in your playbook, and is not zero'd out in the personnel for offense gameplan, it will be called/used by the AI coach in non GL situations.

The GL offense run was used 10 times by the KC AI coach
in the game, to the detriment of the KC offense, which is always the case for an AI team.

The GL defense is automatically used against the GL 2rb3te, the 311 Power I (which is really not a good idea), and the 2 TE formations. You have to use a rule to prevent the GL from being automatically used against the other non GL personnel/formations. Or, you have to not have any GL defensive plays active, which most do.

That is why Im suggesting improvement in the use of both the GL offense and defense by the AI Coach. Prefer that it not be used other than what it was intended for and how it is normally used in RL NFL play. The Eagles dont use the "Tush Push" on other than 0-1 yard situations, as one example, and that might be ruled out by the NFL rules committee this coming season. Had a recent game that was ruined by overuse of the GL defense, due to a mistake in the gameplan. There was no overuse penalty, the GL and a few other defensive plays that are used heavily, do not get the overuse penalty. Maximum # of play use would solve that problem better than the overuse penalty. The overuse penalty is inconsistent at best and not applied to all the plays. As is the "Hot Read" which does not provide any advantage for the "Read" 99% of the time.

The overuse of the GL offense by the AI coach in non GL situations and triggering the GL defense is one of the problems with the RO leagues, if you have participated in one and have watched a RO game, you will not be seeing more simulated RL NFL football, its less. That in addition to several other play calls and gameplan goofs that are painfull to watch in both RO and regular league games, have made a list.

Youre much more tech savy than I am Martin, but Im going on search topic, "can AI be used to write and modify game code?"

https://www.google.com/search?q=can+AI+be+used+to+write+and+modify+game+code%3F&oq=&gs_lcrp=EgZjaHJvbWUqCQgAEEUYOxjCAzIJCAAQRRg7GMIDMgkIARBFGDsYwgMyCQgCEEUYOxjCAzIJCAMQRRg7GMIDMgkIBBBFGDsYwgMyCQgFEEUYOxjCAzIJCAYQRRg7GMIDMgkIBxBFGDsYwgPSAQk1NjAwajBqMTWoAgiwAgE&sourceid=chrome&source=chrome.rb&ie=UTF-8

Based on my initial scan, quick read, it can be used but has limitations. However, as AI continues to evolve, those limitations may be reduced and/or eliminated.

I've used AI heavily, and quite frankly, it's a factor of how precise you are with the questions you ask.

It's mostly trained by scanning Stack Overflow, Github repositories, massive noise to signal ratio.

Used AI on some things not too familiar with... disaster. Took longer to fix than code.

I've actually had extremely good success with AI - I use it regularly on my day job (my company heavily encourages it) and can only credit the speed at which I've been able to do updates here to the help from AI. But I agree, it's all about how you interact with it, because it can be very helpful or it can completely destroy your code base 😂

Did you change anything in display features?

1. Highlighting own team: Standings are barely visible highlighted, for previous and next game the highlighting is completely missing (global problem as it effects Beta and all leagues as well)
2. Line break doesn't work properly on Beta only. You have to scroll left and right to read everything.

System: Windows 11, Chrome (newest updates)

Oh absolutely the latter. I've been known to swear at it, knowing I'll not get punched.

Knocking up a proof of concept, sure, accept massive limitations and problems. Usually re-write heavily.

Work is starting to lean towards limited AI - due to the sector involved, not good idea to rely heavily on it.