Question
If every state of the USA declared war against each other, which would win?
Answer
LOL! Lovely question. I've actually thought a lot about this as part of some idle speculation for a sci-fi near future dystopian story.
I think there is no way to imagine a true 50x50 conflict. You'd have a domino-effect war where neighbors would fight and winners would consolidate and take on the next neighbor etc. If you eliminate the idea of a "Kantian peace" stabilized by balance of power dynamics, and exclude the possibility of alliances, it is a simple simulation, but still not easy to guess the outcome without computer modeling. I'll offer a guess at the end, but this is how I'd actually attempt to answer the question for reals.
BIG ASSUMPTION #1: Military resources end up allied with, and loyal to, the states they are based in. This assumption is WRONG, but it still creates a simpler subset of scenarios to think about. If you allow a preliminary military switching/defection phase, the model gets very messy indeed. So here is the simple and wrong first-order model.
BIG ASSUMPTION #2: Strategic nukes are out of the picture.
Script setup (a basic knock-out model with some elements of a regional overlapping round-robin structure to things).
Lanchester is easiest to simulate (I've actually built a simpler version of this type of model), but maneuver warfare (the other end of the spectrum) wouldn't be too hard to simulate either, at a high level (it's just a set of weights at the highest level of abstraction, to advantage the smarter maneuvering side). The model above would take me about a 2-3 days to build. A simple maneuver model would take maybe a month.
This is low-hanging fruit. But it is not even zeroth-order accurate obviously, because the distribution of resources (armories, naval bases, level of training in population) at the start of conflict, and anisotropic geography (mississippi, rockies, metropolitan regions etc.) would shape conflict far more decisively than neighbor relationships alone would predict. Modeling this is a matter of detailed gruntwork and extracting lessons from the Civil war to encode as weights for various parameters etc. I'd replace the basic geographic-neighbor graph for modeling this conflict with a more sophisticated assets-and-fronts model. Dimensionality of the model would go up about 10x.
If you allow action-at-a-distance (air strikes take out 2nd-degree neighbors etc.) then things get way more complex to model. This would be a proper R&D project of about a year. Model dimensionality would go up about 100x.
If you allow alliances etc., then this gets messy as hell, and I'd abandon computational modeling and go with narrative analysis (basically, get a panel of a few dozen sci-fi writers to write a story each of the war, and look for robust predictions at their intersection). I'd then extract a few of the most interesting scenario clusters for more detailed computational modeling.
But if I REALLY had to just guess based on what I know now, I'd bet on California leading a winning conquered alliance out of the southwest, against a Texas-led conquered alliance from the east. This may surprise those with fantasies about red-state superiority at this sort of thing.
Why? Ultimately, resources other than nukes are about evenly distributed today, so population effects will ultimately determine the outcome. At over 25% of the nation, California can simply punch in a weight class that no other state can. It can easily roll up all the territory west of the Mississippi very rapidly and then patiently wait there to fight whatever exhausted adversary emerges out of the East as last state standing. The only real weakness of the West, assuming no nukes, is water. So long as the Colorado basin river system is protected, the East doesn't have a chance in hell.
In a way, this is just the Civil War rotated about 90 degrees counter-clockwise, and with tactical air power thrown into the mix on both sides. And sad to say, it will likely again ultimately boil down to a Red v. Blue war, and the Blue side will prevail for very similar reasons (better industrial base, bigger population, more financial skill in isolating the economy of the other side...). The Reds may have some minor ground-war tactical advantages (presumably more of them have guns, know how to use them, and have training via militias), but in the context of modern warfare, these are fairly useless.
I think there is no way to imagine a true 50x50 conflict. You'd have a domino-effect war where neighbors would fight and winners would consolidate and take on the next neighbor etc. If you eliminate the idea of a "Kantian peace" stabilized by balance of power dynamics, and exclude the possibility of alliances, it is a simple simulation, but still not easy to guess the outcome without computer modeling. I'll offer a guess at the end, but this is how I'd actually attempt to answer the question for reals.
BIG ASSUMPTION #1: Military resources end up allied with, and loyal to, the states they are based in. This assumption is WRONG, but it still creates a simpler subset of scenarios to think about. If you allow a preliminary military switching/defection phase, the model gets very messy indeed. So here is the simple and wrong first-order model.
BIG ASSUMPTION #2: Strategic nukes are out of the picture.
Script setup (a basic knock-out model with some elements of a regional overlapping round-robin structure to things).
- Everybody attacks everybody along shared borders at T=0. Iteration proceeds at a roughly weekly time-step (though I'd refine the timing model...)
- Each state assesses whether it is winning or losing and picks a defense/offense model for each time step. We'll damp the switching so that the D/O switch only happens if there are gains or losses for n consecutive steps or so.
- The offenders concentrate almost all resources on the weakest neighbor, leaving just enough on other borders to prevent nasty surprises. The defenders concentrate resources on the weakest border (more precisely, they try to make all borders equally weak)
- Lanchester dynamics weaken/strengthen states at each time step.
- When a state falls, it is absorbed into winner state and simulation is restarted with new geography.
Lanchester is easiest to simulate (I've actually built a simpler version of this type of model), but maneuver warfare (the other end of the spectrum) wouldn't be too hard to simulate either, at a high level (it's just a set of weights at the highest level of abstraction, to advantage the smarter maneuvering side). The model above would take me about a 2-3 days to build. A simple maneuver model would take maybe a month.
This is low-hanging fruit. But it is not even zeroth-order accurate obviously, because the distribution of resources (armories, naval bases, level of training in population) at the start of conflict, and anisotropic geography (mississippi, rockies, metropolitan regions etc.) would shape conflict far more decisively than neighbor relationships alone would predict. Modeling this is a matter of detailed gruntwork and extracting lessons from the Civil war to encode as weights for various parameters etc. I'd replace the basic geographic-neighbor graph for modeling this conflict with a more sophisticated assets-and-fronts model. Dimensionality of the model would go up about 10x.
If you allow action-at-a-distance (air strikes take out 2nd-degree neighbors etc.) then things get way more complex to model. This would be a proper R&D project of about a year. Model dimensionality would go up about 100x.
If you allow alliances etc., then this gets messy as hell, and I'd abandon computational modeling and go with narrative analysis (basically, get a panel of a few dozen sci-fi writers to write a story each of the war, and look for robust predictions at their intersection). I'd then extract a few of the most interesting scenario clusters for more detailed computational modeling.
But if I REALLY had to just guess based on what I know now, I'd bet on California leading a winning conquered alliance out of the southwest, against a Texas-led conquered alliance from the east. This may surprise those with fantasies about red-state superiority at this sort of thing.
Why? Ultimately, resources other than nukes are about evenly distributed today, so population effects will ultimately determine the outcome. At over 25% of the nation, California can simply punch in a weight class that no other state can. It can easily roll up all the territory west of the Mississippi very rapidly and then patiently wait there to fight whatever exhausted adversary emerges out of the East as last state standing. The only real weakness of the West, assuming no nukes, is water. So long as the Colorado basin river system is protected, the East doesn't have a chance in hell.
In a way, this is just the Civil War rotated about 90 degrees counter-clockwise, and with tactical air power thrown into the mix on both sides. And sad to say, it will likely again ultimately boil down to a Red v. Blue war, and the Blue side will prevail for very similar reasons (better industrial base, bigger population, more financial skill in isolating the economy of the other side...). The Reds may have some minor ground-war tactical advantages (presumably more of them have guns, know how to use them, and have training via militias), but in the context of modern warfare, these are fairly useless.