Dr Evil’s Move Legality Engine for VASSAL — tech demo

Dr Evil’s Move Legality Engine for VASSAL — tech demo. This is an early, in-development proof of concept, not a finished
product — I’m sharing it so people can see where this is heading.

Three classic hex wargames (SPI Arnhem, AH Tobruk, ASL) with movement rules fully enforced, running entirely in your
browser. Unzip, double-click START_HERE.html, drag counters — the engine computes every legal destination: terrain costs,
rivers and bridges, zones of control, walls and hedges, vehicle facing. No install, nothing to configure. Expect rough edges
— each game is one scenario, movement only, and some unit values are still provisional.

Download: https://drive.google.com/file/d/1Gxjkx3myzGKjuww4NgXyanX9Ld3jylnI/view?usp=drive_link
Source (MIT): GitHub - DrEvil-TitaniumHelix/vsav-engine: Game-state API for VASSAL 3: read, rule-check, and write .vsav save files directly. No VASSAL modifications. By DrEvil / Titanium Helix. · GitHub

This is step one toward the real goal: an AI opponent for VASSAL games. The full engine reads and writes real VASSAL
.vsav files — so VASSAL, this browser UI, and an AI player become three interchangeable clients of the same save. Move
legality was the foundation; combat resolution and the AI itself are next.

This is exciting

The beta (or alpha??) version of the Legality engine is now available for testing and feedback. You can download a simple .exe or get the git repo for your own experimentation. Two games are included (I am not distributing per se, since this is an internal test platform). Tobruk is a very simple tank battle, while Afrika Korps (by cholmcc Afrika Korps (3rd Ed.) - Module Library - Vassal ) is a complex test case.

The UI should be intuitive but may have a learning curve. Let me explain the Tier selection. Tier 0 is meant to replicate current VASSAL. Tier 1 imposes movement restrictions. Tier 2 resolves combat and imposes combat rule restriction. Tier 3a, which is the goal of the system, is an AI opponent. 3a is a relatively simple foe and teacher. Tier 3b - for later - will let you pipe in your favorite AI model and should play brilliantly.

I’d recommend trying Tobruk first, and if you understand chomcc’s AK then give it a try later. There will be bugs, and in fact creation of these encodings revealed numerous bugs in the original rules (as I am sure grognards are aware).

What is this?: its a computerization based on existing modules, with Claude Code ingesting all the rules, cataloging them, then imposing them in a UI inspired by VASSAL.

Why did I do this?: so I can play any existing module solo against an AI opponent and learn from an AI teacher with infinite patience while enjoying the game. Plus, its a step toward preservation of these masterpieces, making them even more accessible to a wide audience.

I’ll create a video walkthrough. PM me to get access to the exe, git repo, and video, or I can post here if approved by moderators

@JoeFromRhodeIsland I asked Claude code to encode SPI the blue and grey. It completed the job in 3 hours and I will add it to the beta tomorrow. So the system produced tobruk in 5 days then afrika korps in 3 days then blue and grey in 3 hours. Safe to say the system has learned and can likely encode any hex game with the requisite source documents.

I’ll try to post a link to the git and exe in the next post here, but they may get spam filtered.

Git: GitHub - DrEvil-TitaniumHelix/vsav-engine: Encodes printed wargames whole - map, counters, rules, combat tables - into computerized versions that enforce their own rules, with rulebook citations. Deterministic legality gate, engine-owned seeded dice, replay-verifiable logs. Ships Afrika Korps + Tobruk; bring your own module to encode yours. · GitHub

The exe only: https://drive.google.com/file/d/1FuJlt54Mb2FIAunbKrEXBKCpHCOngCpH/view?usp=drive_link

New a final beta version pushed - same links as above. You can encode your own games - see the encoding guide.

4 scenarios total: added SPI’s Blue & Gray: Chickamauga and Arnhem full scenario

I’ll stand down now pending feedback. I estimate about 1 month of work with Fable to encode the top 250 games in Vassal if there is interest. Interest will also drive other upgrades like full LLM AI opponent (its also possible to create a custom RAG modeled and retrained model that could be an expert just for hex wargames), and a module submission and configuration management function.

Bruce I am loving this, but as a computer programming novice, I need to take a deep breath and check this out.

I’m a computer engineer… don’t waste your time. What he’s saying isn’t credible to anyone familiar with even the general specifications of IBM’s “Deep Blue” machine, and the simplicity of the rules of Chess, compared to merely the movement rules of ASL, and the MUCH MUCH MUCH larger storage storage tree(*)

(*) a tree is a type of data structure which is recursively self-referential in nature, and generally grows at the rate of x^N (x to the Nth power), and therefore has memory and disk storage requirements which grow at such a rate. Deep Blue was juggling the potential moves and counter-moves only 32 pieces, of which 16 (8 on each side) are pawns, and without such counfounding things such as terrain features (roads, grain fields, walls, hedges, elevation changes, steep embankments / cliffs, stream beds, canal/river hexes, enter/exit pillboxes, bridges, etc. Not to mention going up and down staircases, and let alone figuring out the logic of how much or little, and which particular part of the map any particular unit should want within its line of sight (units being used for ambush doesn’t necessarily want to be in a position with large amounts of terrain within its line of sight – only the trip point for the ambush, and the overall killzone.)

I do not for a moment believe that he, who admits at the beginning that he’s NOT a programer, has even considered the sheer number of possiblities that an AI agent would have to be trained on, let alone the hundreds of examples of each that an AI agent needs in training data (as well as BAD examples to score negatively)

Dr Evil… interesting name he chose for himself… When someone tells you what they are, believe it.

He’s shilling for FABLE… which is an AI made for narrative fantasy games like D&D, not tactical historical map-oriented games.

Dr. EVIL (believe it when he tells you what he is) says earlier that one of his sub-projects burned up 3 hours worth of time on Claude Code.

That’s enough tokens to buy a decent used car.

FOR ONE early 1970’s game with very simple rules (about a dozen pages), compared to ASL’s rules which are hundreds of pages long, and just the rules for LEGALITY of moves cover dozens of pages (let alone everything else which helps a player distinguish between a move that improves his tactical situation and one which does not). So, assuming the cost of tokens for Claude Code to run numerous processors in parallel for 3 hours… to come up with code to figure out movement for Tobruk… How many hundreds of thousands of dollars would it take for the several WEEKS needed to do the same for ASL?

And what exactly would be the point of playing ASL, or even Afrika Korps, against an AI rather than a human opponent? I don’t know of any wargamers who are just salivating for the chance to play against anything other than a human opponent.

This entire thread smells like Dr Evil = Shill for FABLE AI.

Fable 5 is the most recent frontier model from Anthropic. It is a consumer version of Mythos. I am not familiar with the Fable you are referring to. I have the max plan on claude code, so all of this has been free.

You must be the only customer still getting the subsizidzed, under $1000/month per user rate. All the commercial customers, when they were switched over to “per token” billing, were using their monthly allotments in less than 3 work days (so, somewhere between 24 and 36 compute-hours)

Fable. Mythos.

Fantastic stories that are mostly fictional.

They’re revealing EXACTLY what they’re selling is. Unbelievable fantasy. The biggest fantasy being “You can fire all of those highly skilled employees with advanced degrees in engineering, analysis, etc. and replace them by typing prompts at an AI.” Of course, the finance bros don’t even know enough about engineering to even know WHAT prompts to give to an AI, EVEN IF the LLM’s could actually do what Sam Altman, et al claimed. Which they can’t.

What CAN LLM’s do? Follow the herd. Because they are nothing more than STATISTICAL TABLES of what word/term Y follows term X in the context of words W with the last one being X. That’s ALL LLM’s are. Nothing more and nothing less. In other words, HUGE predictive text databases, just like the Google Keyboard on any android-based phone.

As I said before, there’s a reason why Sam Altman left the country in the middle of the night on a private, chartered aircraft to go HIDE in South America.

I guess prompting is already a thing of the past. But your point about AI being incompetent while replacing all these employees is quite scary.