• Major upgrades completed! Please report any issues you spot in here

I used ChatGPT to develop a Gameplay Mod and what I achieved is amazing! - Give it a try

amartingil

amartingil

League 1
Joined
20 September 2016
I've never been much into AI until this year, when I started using it for learning purposes. After being impressed by what ChatGPT is capable of, I decided to see if it could help me dive into EAFC modding and better understand how the parameters and values inside the gameplay files work.

I should say that I already had modding experience, so I wasn't starting from scratch, but AI has allowed me to go thousands of times deeper. I won’t go into much detail for now, just watch the videos and give it a try.





The modifications I’ve made so far reflect my vision of how the game should feel:
  • Realistic ball physics, with real-time energy transfer and less reliance on scripted patterns.
  • Realistic body movement, with noticeable inertia and weight shifts.
  • Dribbling that respects physics, where the ball naturally separates from the carrier.
  • Defensive shape that functions as a unit.
  • Intelligent AI in defense, that predicts rather than reacts, stays active, and aggressively presses for the ball.
  • PES 5/6 vibe

I don’t have time to go into full detail now, but I’ll share a more in-depth explanation of the tweaks and process when possible.

The mods are made for the TU14.1, I do not know how it changes when using it in TU16.

The mod is divided into separate fifamod modules for development purposes. To experience the full effect, all gameplay-related files must be used, including the player downscale fifamod file:

Gameplay fifamod files:
https://www.mediafire.com/file/72wxnfam6xehwbx/Attack_dribbling_AI_-_Evoweb.fifamod/file (Modifies Cpu dribbling decision making)
https://www.mediafire.com/file/sa0jv49sj79oxc6/Defense_Positioning_and_AI_-_Evoweb.fifamod/file. (Changes team shape during defensive phases - modifies Cpu defensive decision making)
https://www.mediafire.com/file/eb74aob2yb1n85d/Physics_and_dribbling_-_Evoweb.fifamod/file (Changes ball physics, player's movement with and without the ball)
https://www.mediafire.com/file/jd8vse9h3mnh5jr/Players_downscaled_%2B_Shorter_shorts_-_Evoweb.fifamod/file (Downscales player size)

Graphics fifamod files:
https://www.mediafire.com/file/ihfydhqlttxryh3/Color_grading_-_Evoweb.fifamod/file (Changes color grading in gameplay)
https://www.mediafire.com/file/b9qab7gh2lkifim/No_undershorts_-_Evoweb.fifamod/file (removes undershorts-uses legacy editing, works only in 14.1)

I just wanted to share this because I believe there’s a lot of frustration nowadays with the state of football games, and the results here are so good that everyone should be able to enjoy them.





Using the downscale introduces a visual issue: players appear tiny in many cinematics. This is the price to pay. However, if you prioritize gameplay over everything else, I’m sure you’ll be able to overlook this flaw.
 
Last edited:
raul_ole said:
IT looks good tu16?
Click to expand...

wiiuchamp said:
can you update to latest version?
Click to expand...

The gameplay files should work for TU16, as there are no changes in the legacy files.
When you apply then in Fifa Mod Manager simply ignore the incompatibility message.

However, if EA has done any changes after TU14.1 in files that are not part of the mod, the gameplay feeling might differ sightly.

Unfortunately, I'm away holiday for a week tomorrow, so I won't be able to update anything till then. Nevertheless, you should give it a try.

The setup I developed the mod for is:
Simulation mode disabled (all sliders at 50)
Weather disabled
Wind disabled
Cpu custom (all sliders at 50)

All semiassisted.
Analog sprint activated.
Contextual dribbling disabled
 
This looks like an interesting gameplay!
I'm not too much into AI either, but for these type of gameplay and algorithms, it does have it uses. It does look like the gameplay is improved and more realistic! Will give it a try when an updated version comes out! :)
 
amartingil said:
I don’t have time to go into full detail now, but I’ll share a more in-depth explanation of the tweaks and process when possible.
Click to expand...
I would like to get a in depth explanation. Maybe to a later point of time. I often heard Modders here and there using AI for creating amazing mods for many Games around. I have no clue how AI can help, so yeah would be pretty interesting you do some day an enlightning report. :D

Well. it´s of course worth a try and I am curious what you and ChatGPT have created. :-)
The not updated Mod to TU 16 can be for some reason change some things with your mod. I´ll play first on tu 16 and after that I maybe downgrade to TU 14.1. Easy enough through the FC downgrade Tool.
 
Byronic said:
I would like to get a in depth explanation. Maybe to a later point of time. I often heard Modders here and there using AI for creating amazing mods for many Games around. I have no clue how AI can help, so yeah would be pretty interesting you do some day an enlightning report. :D

Well. it´s of course worth a try and I am curious what you and ChatGPT have created. :-)
The not updated Mod to TU 16 can be for some reason change some things with your mod. I´ll play first on tu 16 and after that I maybe downgrade to TU 14.1. Easy enough through the FC downgrade Tool.
Click to expand...
What I've done, basically, is upload gameplay-related files to ChatGPT and ask it to analyze their content. Each of these files contains key names whose functions you can only guess by tweaking and testing their impact. Doing this manually is a never-ending task. With AI, however, you can reach conclusions thousands of times faster, as it can guide you in a precise (though not perfect and often flawed) way. It can tell you the potential impact of changing a value, an array, or a curve, and even suggest specific tweaks aligned with your vision of gameplay.

For example, I uploaded the file that controls ball control parameters after extracting it with FET, and asked it which entries affect frequency of touches, touch length, and turning time, etc. Then, I asked for suggestions to make dribbling feel more realistic and weighty. AI provided a list of modifications, which I applied and tested. I then gave ChatGPT feedback on the actual effects of its recommendations and repeated this cycle, fine-tuning things further each time. Throughout, I remained critical and used my own judgment to determine the final outcome.
 
amartingil said:
What I've done, basically, is upload gameplay-related files to ChatGPT and ask it to analyze their content. Each of these files contains key names whose functions you can only guess by tweaking and testing their impact. Doing this manually is a never-ending task. With AI, however, you can reach conclusions thousands of times faster, as it can guide you in a precise (though not perfect and often flawed) way. It can tell you the potential impact of changing a value, an array, or a curve, and even suggest specific tweaks aligned with your vision of gameplay.

For example, I uploaded the file that controls ball control parameters after extracting it with FET, and asked it which entries affect frequency of touches, touch length, and turning time, etc. Then, I asked for suggestions to make dribbling feel more realistic and weighty. AI provided a list of modifications, which I applied and tested. I then gave ChatGPT feedback on the actual effects of its recommendations and repeated this cycle, fine-tuning things further each time. Throughout, I remained critical and used my own judgment to determine the final outcome.
Click to expand...
Cool stuff! Thanks for taking the time to explain the AI part in depth.
 
You must log in or register to reply here.
Back
Top Bottom