Time Warp With Two Ancient Armies

[mejobo]

Is it possible to expand Time Warp to allow ancient armies to face each other? It seems like it could be nice to access some of the scenario and terrain options only available in Medieval (siege scenarios come to mind in particular) without resorting to modding.

[kronenblatt]

It's feasible but probably not desirable for Slitherine because that would allow for playing Ancients in Medieval without actually having to buy Ancients.

[mejobo]

Is there not a restriction already for what DLC are owned across both games? I assumed that would be the case and Time Warp as such would only be usable already if you own both. My assumption being incorrect would certainly explain why Time Warp is a bit limited!

[Paul59]

Is there not a restriction already for what DLC are owned across both games? I assumed that would be the case and Time Warp as such would only be usable already if you own both. My assumption being incorrect would certainly explain why Time Warp is a bit limited!

No, you don't have to own Ancients to play Time Warp in Medieval. That is why you cannot have two Ancients armies playing against each other in Time Warp, it would make buying Ancients largely redundant, just as kronenblatt said.

[Antaine22]

Sorry for necroing the thread, but this thing keeps revolving in my head.
I own both Ancients+DLCs and Medieval+DLCs. I'm more into the medieval period but like to fire an ancient scenario from time to time. And for some reason I don't really want to have both FoG2 and Medieval on my drive (crazy I know).
It'd be really great if we could "install" the Ancients DLCs to Medieval to unlock the Ancients vs Ancients in the time warp. Or get another DLC that unblocks it (idk "Battles of old" or something). I'd be readily throwing another $20-30 into it.

[MarkShot]

I just read the Medieval manual last night. The open battle types are richer, but honestly I am more an Ancient period person. But I bought this on the reputation (all DLC) based on Ancients (all DLC). I would actually pay for an engine upgrade option to just allow ancients to be played with this engine.

So, people could buy the upgrade or just stick with the original; up to them. I would pay for back porting these features to Ancient and have Ancient Classic and Ancient Plus.

Since you guys have this warp option available already, then the base code necessary is probably almost there. Finish the implementation, and let those who want it pay.

Go ahead and do some of that programming cut and paste stuff, and make some money. I understand I am not paying for new code, but for porting/UI ergonomics. I am fine with that.

PS: I noticed there is way more user content for Ancients. I just think there is less interest in this period, despite TW players advocating for MTW3 for 10 years now.

[kronenblatt]

Yes, I’d very much like that too, especially the Frozen map type but of course also the Medieval style of graphics and its scenarios. And would be happy to pay for it.

Off-topic, but somewhat related, I’d love to get a more intricate “AI” for both Ancients and Medieval in SP games (to make them more challenging), paid-for of course. I know there’s Rise of the AI Mod for Ancients but sadly not for Medieval and it would be awesome too have such an AI integrated into the base game.

[MarkShot]

Yes, I’d very much like that too, especially the Frozen map type but of course also the Medieval style of graphics and its scenarios. And would be happy to pay for it.

Off-topic, but somewhat related, I’d love to get a more intricate “AI” for both Ancients and Medieval in SP games (to make them more challenging), paid-for of course. I know there’s Rise of the AI Mod for Ancients but sadly not for Medieval and it would be awesome too have such an AI integrated into the base game.

Forget protein folding, this a project for AI. Understand the data formatting differences for the two games and rewrite all the Antiquity data files as mod for Medieval. But personally, I don't think you will get that done so easy with AI. I was playing with QWEN3 30B/3AB MOE q=4 on a 4090. I had it write a C program of intermediate complexity which it did well. I then showed it two WinBatch (defunct language from the 90s) complex script programs (well written and documented; readable). It explained the code well. It now had all the examples it needed for the key statements, conditionals, looping, exception handlers, and routine/function calls. I said, "rewrite that C program in WinBatch". It totally bombed at a task which most programmers could have worked out. It was not a lack of context as I gave it a 32K token window. So, perhaps Apple and META's Yann LeCun is right. These things cannot reason; only manipulate patterns. Because if this could be done by a mod, QWEN3 can hit 110 t/s on a 4090. I am thinking of picking up an RTX 6000 Pro next year more for the 96Gb VRAM than GB202 chip ... maybe I will revisit the idea of data port/mod at that time.

[MarkShot]

That would be quite a news story. "The AGI barrier broken by modding games!" I always new the secret to general intelligence was play.

[kronenblatt]

… I’d love to get a more intricate “AI” for both Ancients and Medieval in SP games (to make them more challenging), paid-for of course. I know there’s Rise of the AI Mod for Ancients but sadly not for Medieval and it would be awesome too have such an AI integrated into the base game.

Forget protein folding, this a project for AI. Understand the data formatting differences for the two games and rewrite all the Antiquity data files as mod for Medieval. But personally, I don't think you will get that done so easy with AI. I was playing with QWEN3 30B/3AB MOE q=4 on a 4090. I had it write a C program of intermediate complexity which it did well. I then showed it two WinBatch (defunct language from the 90s) complex script programs (well written and documented; readable). It explained the code well. It now had all the examples it needed for the key statements, conditionals, looping, exception handlers, and routine/function calls. I said, "rewrite that C program in WinBatch". It totally bombed at a task which most programmers could have worked out. It was not a lack of context as I gave it a 32K token window. So, perhaps Apple and META's Yann LeCun is right. These things cannot reason; only manipulate patterns. Because if this could be done by a mod, QWEN3 can hit 110 t/s on a 4090. I am thinking of picking up an RTX 6000 Pro next year more for the 96Gb VRAM than GB202 chip ... maybe I will revisit the idea of data port/mod at that time.

Could you, with your knowledge in programming, consider porting the Rise of AI Mod from Ancients to Medieval?

[MarkShot]

I will have a look. I am mainly interested in see what I can get AI to do. QWEN3 is pretty good at programming.

But I won't be getting a 6000 (workstation card) until 2026. With all the NVidia 5090 issues, I want to wait and see: 12VPWR cable, drivers, reviewer blackmail, missing ROPs, missing PHYSX, MFG, 4070=5090 for $549, ... I have seen Der Bauer's coverage and Gamers Nexus coverage of the 6000, but they are reviewing for FPS and not T/S. Watching and waiting ...

[kronenblatt]

I will have a look. I am mainly interested in see what I can get AI to do. QWEN3 is pretty good at programming.

That’s great! Looking forward to following your progress.

[MarkShot]

I will have a look. I am mainly interested in see what I can get AI to do. QWEN3 is pretty good at programming.

That’s great! Looking forward to following your progress.

Disclaimer: I have not done systems programming for like 30+ years, but I've been writing gaming utilities in WinBatch and AHK for maybe the last 25 years.

My first report on the first time looking at the code.

LANGUAGE
=========

A strongly typed block structured interpreted language which appears with argument passing (I didn't notice the use of global variables) which provides for constants and macro facilities which appears to be patterned on C. I believe this to be Slitherine's proletary engine called, Archon. I was looking at Ancient, but for this level it would not matter.

Okay, this is LUA as there are LUA DLLs in the main directory and LUA was based on C. There also appear to be Microsoft Class DLLs at the root level. So, LUA probably drives Archon's behavior with the low level UI and rendering done in compiled C or C++ using Microsoft .Net (since Matrix installs always include .Net as redistribable).

Code Base
=========

The game's code base which is exposed deals which much more than the game board/battle. However, the mod's scope is the board. AI.BSF seems to be file which tells the game what AI units will do, but there is not main entry point as the game must perform the simulation, and it would seem that eventually a flag is returned to the game telling it that the AI entry points have no further requests, and it may end the turn.

This code base seems to have no persistence (within the game engine itself), but RJ seems to be handling deployment and perhaps letting the game take its course from there. As he explicitly notes, this module is blocked for scenarios.

RJ clearly is more than the usual modder. SL's AI code is about 1500 lines, and RJ is double that. But he does have large segments of ideas he tried which he left in the code and later commented out.

Guesses
=======

My guess is the new game had added functionality, but much code reuse from the original including the above scripts. Thus, whatever RJ attempted is probably mainly valid, but insufficient for some new unit behaviors like dismounting and new battle types. It does appear that RJ approached the code from scratch as opposed to just tweaking the stock code. The main task would be to determine how new mechanics impact his heuristics, and what heuristics should exist for other game types.

Actually, in looking at the code both SL and RJ, I don't see any imposition of game type on the code. It may well be that the AI simply always attempts to route the most enemies, and that the game types are for the human player to add flavor.

So, I would assume most RJ's code is reusable. A text difference analyzer would highlight what SL felt was necessary to add or modify for the new game which would then point to the changes needed in RJ's code. I think the main volume of work for RJ's code is probably not coding but testing; not simply for bug, but does it yield and improvement over the SL stock opponent.

RJ appears to have made some optimizations to cache results for performance. Like he is determining routes which except for enemy units, the map is constant. This is somewhat similar to transposition tables in chess engines.

Additionally, he appeared to be making multiple passes over the data for optimal choices; setting 4 as a limit. Thus, increasing this number might also improve his AI behavior if one takes to be similar to chess engine look ahead. If this really worked, it would be nice to expose to users

<to be continued>

R

[MarkShot]

NEXT STEPS
==========

Well, as indicated testing is going to be where most of the porting work lies. One would hope that RBS has a built in auto play mode which allows the game to play itself like two sides of a PBEM. Then, you could have RJ-AI-M faceoff against SL-AI-M. If the port is any good, then RJ-AI-M should be the victor in most games as we already know SL-AI-M plays a decent game.

As for optimization, RJ's functions are optimizing priority and sequencing of movement, shooting, and sequencing. His code is fairly modular. A good coding AI, should have a base of optimization algorithms in its training which might be used for tuning RJ's functions and via the auto-tester determining if the win:loss ratio improves or the margin improves.

---

Comments anyone?

@RBS, is there an auto-test capability to run two instances of FOG2M (with one modded)?

PS: I don't think I would do more on this until 2026. At that point, I will have vastly improved AI card for programming models (as opposed to toy demos). I am not really motivated to code these days as I am old and just want to play games myself.

[MarkShot]

Oh, yes, one more RBS question.

If the Ancient data was reformatted as a mod to run under the Medieval engine would that be a EULA violation? And even if not a EULA violation, how would I or someone else avoid preventing future sales of the first game and DLCs if this is done? Ethically, I don't want to do that even if it not is precluded in the EULA for the first game.

[rbodleyscott]

One would hope that RBS has a built in auto play mode which allows the game to play itself like two sides of a PBEM.

Sadly not.

@RBS, is there an auto-test capability to run two instances of FOG2M (with one modded)?

No

[rbodleyscott]

Oh, yes, one more RBS question.

If the Ancient data was reformatted as a mod to run under the Medieval engine would that be a EULA violation? And even if not a EULA violation, how would I or someone else avoid preventing future sales of the first game and DLCs if this is done? Ethically, I don't want to do that even if it not is precluded in the EULA for the first game.

There would not be a problem with re-writing the Rise of AI Mod to work with Medieval, as long as it does not allow two ancient armies to fight in FOG2 Medieval!

Because obviously that could affect uptake of FOG2 Ancients.

[MarkShot]

Understood. I will go no further on data migration from the first game to the second, or publish any more ideas on how one might do it. It's not a question of EULAs, but the ethics of "doing wrong".

RJs work seemed to be parameter driven and not focused on specific units. So, there is nothing he did with precludes the second game being supported.

[rbodleyscott]

RJs work seemed to be parameter driven and not focused on specific units. So, there is nothing he did with precludes the second game being supported.

Correct. He just did not have sufficient interest in Medieval to do it himself.