Pretty interesting. I actually really like the idea of a steep platform with notches you can climb on. A Rockwall end game would be quite a sight. Did you factor the game theme into any parts of the prediction?
I didn’t necessarily factor in any traditional surface level game “theme” but many of the finer details are certainly determined around deliberate focused game theory
Stacking boxes… the recent elevator releases were not a coincidence… 
Two-speed shifting elevators. High gear for stacking, low gear for climbing.
Swerve PTO?
This is actually pretty close to what I had drawn a bit back cause I was bored, I drew this a few months ago 
I’ll ask the AI.
So… recycle rush with a different endgame? Haha
This reveals the ultimate question:
Is defense possible in a stacking game? Or will we just have a whole lot of penalties?
I don’t enjoy competing in or volunteering for games in which onerous technical fouls are used to enforce the style of play that designers intended. 2003 and 2015 were both bad games by that standard.
That depends on two things.
- Are robots allowed to interact?
- Are stacks in any zone of the field protected?
If the answer to 2 is “no”, it’s not a stacking game. Question doesn’t apply.
If the answer to 1 is “no”, defense isn’t possible (theoretically).
Ideally, the answers are both “yes”… but that hasn’t happened yet.
hmm that’s an interesting prediction. Are there any other iterations of the guess? Maybe other estimations with the same parameters?
What was the nature of the input provided to ChatGPT which generated the game prediction shown?
Sure it is. 2003 was the best game ever!
I’d say that this is as good as any other prediction. Has about as much chance of being correct too, AI or no AI. Of course, as we all know, the real prediction is that this will finally be our water game.
I second this question. The information fed into it could drastically impact its results if you only included recycle rush and lunacy
ChatGPT is trained on the “internet” up to 2021ish. It seems from my interaction to have strong idea of the macro concepts of FRC game theory.
After getting it to give me a very abstract rundown of the “2023” comp I had it go into detail on game pieces, scoring, and field layout. The contents of the document are my condensed and weeded interpretation of the output.
When prompting the program I used neutral language, only asking it further detail on what it had previously generated. Only after is when I added my own spin to the output!
The AI point values varied somewhat from my final balancing. In addition the small game objects were not color coded to an alliance, that was a balancing decision I made after.
However, the model was quite confident with small irregular game objects placed in the stacked containers, and an overall “stacked box” themed comp.
The endgame is exactly as described by the program, sans point values and the 2 bot on top limitation.
Parts of this remind me more of an FTC game. (variety of shapes, moving things to the side or corner of the field)
I see you’ve played GDC spoony before!