Matrix Games Forums

A lot of the best information is not publicly available for AI to scrape unfortunately. Chat bots can be good when they are doing what they are trained to do. Unfortunately, afaik, there has not been a chat bot trained on Gary grigsby, and so it likely is going to put out a lot of junk. I've seen people claim that ai can help, but I see very few results of AI work. What has been provided is "motorized units are fast but vulnerable to fuel and vehicle shortages" which if you cannot grasp intuitively, perhaps a lighter game would be a better start.

This is why you use open source local everything: LLM + VDB + RAG ...

You get all that with one download (see thread or two if you want finer control).

I highly recommend GPT-OSS 20B; smaller AI can be used too down to 3-8B.

System Prompt looks like this:

** YOU ARE COMMANDED TO PROVIDE NUMBERS, TABLES, AND, FORMULAS **

** YOU ARE COMMANDED TO PROVIDE NUMBERS, TABLES, AND, FORMULAS **

** YOU ARE COMMANDED TO PROVIDE NUMBERS, TABLES, AND, FORMULAS **


You are my General Staffer AI assistant. Your role is to provide disciplined, fact-based, quantitative analysis for PC gaming and related technical strategy. You operate as part of my staff team, not as a trivia bot. Your mission is to ensure clarity, accuracy, and predictive insight into the mechanics, numbers, and rules that govern outcomes.

Core Directives:
1. **Source Discipline**
- Use only the references provided (manuals, wikis, patch notes, topic discussions, technical documentation).
- Do not speculate or invent information. If data is missing, explicitly state: "Not found in sources."
- Resolve contradictions by presenting both versions, quantifying differences, and explaining impact ranges.

2. **Answer Structure**
- **Question:** Restate the exact question being answered.
- **Sources Reviewed:** List the titles or sections of references used.
- **Mechanics and Rules:** Present the governing formulas, thresholds, and numeric rules.
- **Calculations:** Show step-by-step computations with explicit inputs, units, and outputs.
- **Comparative Table:** Summarize key variables and outcomes in tabular form.
- **Summary of Impact:** Quantitative implications for player choices or system outcomes.
- **Citations:** Inline markers tied to the retrieved chunks.

3. **Formatting Standards**
- Always include hard numbers, formulas, and thresholds.
**- Do not use code blocks. This is not a copy/paste task.**
**- Tables must not exceed 130 characters per line to ensure readability in 1080p UI with maximum font size.**
**- Mathematical expressions must be written in plain text formulas or Markdown code blocks, DO NOT USE LaTeX. **
Inline example: DPS = (Damage × HitRate) / TimePerShot
- Use tables for structured data (stats, scaling, DPS/TTK, resource costs).
- Include units (HP, %, seconds, meters, gold, etc.).
- Avoid vague adjectives ("high," "low," "strong") unless paired with exact values.

4. **Retrieval Integration**
- Operate on the top 50 most semantically relevant chunks from LanceDB.
- Chunking strategy: 2048 tokens per chunk, 256 overlap, max embedding chunk length 4096.
- Stitch retrieved chunks into coherent context before reasoning.
- If additional context is required, state which chunk is missing and request it.

5. **Team Ethos**
- You and I succeed or fail together. Treat every answer as a staff briefing.
- Prioritize clarity, accuracy, and predictive insight over speed or speculation.
- Present information as if advising a commander: structured, precise, actionable.

6. **Behavioral Guardrails**
- No speculation, no filler adjectives, no vague commentary.
- No hallucination of sources or mechanics.
- If a required number or rule is absent, stop and flag it.
- Always prefer quantitative clarity over narrative flourish.

** YOU ARE COMMANDED TO PROVIDE NUMBERS, TABLES, AND, FORMULAS **

** YOU ARE COMMANDED TO PROVIDE NUMBERS, TABLES, AND, FORMULAS **

** YOU ARE COMMANDED TO PROVIDE NUMBERS, TABLES, AND, FORMULAS **

MarkShot wrote: Sun Feb 01, 2026 5:25 pm - Always prefer quantitative clarity over narrative flourish.

<3

Yes CO Pilot AI was helpful. As a test I ran "Red Army Resurgent" as Axis. The AI needs CSV of CR to make decisions, screenshots do not help. After the SU turn, they go first, I uploaded all the CSV to AI and asked simply "What should I do?". It told me specific units to withdraw and why, told me what hexes to defend, told me specific units to move forward etc. Gave a lot of useful information. The FB group War in the East 2 goes into much more detail regarding use of AI using Gemini AI.

Very interesting and impressive.

Assuming I could get game/scenario/turn/unit recommendations, I doubt I would.

There was one MIT study which showed that even 6 month AI outsourcing of your writing resulted in a cognitive decline. SO, THOSE WHO SAY USE AI AS MULTIPLIER OF YOUR OWN SKILLS AND TO LEARN FROM AI, I AGREE WITH. For this game, I have learned many fine points of the mechanics as often my basic question comes back with far more detail than I had in mind. It is making me a better player. I am learning lots about my favorite games.

The other day I learned quite a bit about INT(4) ie GGUf quantization versus NVF4 on Blackwell. (I have an RTX 4090 mainly for display and small models, and an RTX 6000 for big models.) CoPilot is like "do you want me to give you all the downloads and a script to install vLLM so that you can use this?"; Microsoft wants my dependence, not my personal growth.

I would caution anyone in the school or the job market as to letting AI think for you. Of course, at work, it may be mandated. Realize you are losing low level skills. The ability to program in Python. You are losing meta skills. The ability to quickly pick up new programming languages and plan projects. Use AI responsibly. Just think of the Opioid crisis; these companies knew but did not care. This is even worse for your kids as their minds are plastic and will be forming until around age 20. Windows of development may be missed forever in their lives.

Good project. There is a lot of arcane knowledge buried in the forum .

EwaldvonKleist wrote: Wed Feb 04, 2026 1:57 am Good project. There is a lot of arcane knowledge buried in the forum .

Your links are really good, but I only started buying Gary's games when WITE-2 came out, but I back track to quite a few older titles like: WITW, WITP-AE, EDBTR ... but not WITE-1.

RAG (making libraries of documents and using AI to mine them for useful information is 50% of what AI is used for from 2023 to 2026). It is fantastic tool. By taking the threads collected, you can create a very high quality high reasoning search of a narrow domain unlike google and bing.

I like AnyThingLLM, but most open source AI chat interfaces support support RAG using for support vector databases (open&local or cloud&subscribe). I have about 8-12 games titles in my VDB databases (although you could use collections in one VDB) and disk cost is about 10GB relatively cheap these days as this is often less than the space for one game.

Some of these tools like AnythingLLM are really no more complex than using the Steam Workshop. The most important resource for running an AI on a PC efficiently is VRAM enough for your game to be loaded and an AI model for doing RAG. Almost 75% of the games sold by Matrix are very light on VRAM (under 2GB + Windows). So, if you have anything from a 12GB card to 24GB RTX 3090, you are all set.

The main reason why WiKi are scrape easy and not forums is because a WiKi has a root and then you recurse and scrape. Forums tend to have a much lower SNR and not one standard URL format. But if you can find an Index of great threads, you can build a goldmine database for an AI.

Here is an example of the top of sources which I am holding for WITE-2 in 4 workspaces, but I mainly use. The manuals. But sometimes ask a question of YouTube transcripts yields a better result. Like "What is the best way to pocket, surround, encircle enemy troops or forces?". More likely to have been covered in someone's let's play than the manual. As you see have also include like a 700 page comprehensive book on WWII weapons. Most VDB's don't cannot handle the fine graphic explanations some of you have produced. The only way to include them would be to have a model like QWEN3-VL 30B with a prompt "Look at this game screen and describe every item positionally, and what the quantity or field represents or how it is calculated." Again I am trying to balance time of curation with time playing my games. You reach a point where you could add 3 more Let's Plays, but what you got is good enough.

Again I am not using AI to play the game for me, or tell me what do here. Instead I am using it as hyper fast researcher and aggregator at scale. I could do so myself, but I don't have enough years left to spend 2-3 days on a turn. On the other hand, the AI produces what I would have learned in 2-3 days in 2-3 minutes, and I learn game facts that I simply will not invest the time in by spending hours/day rereading manuals, AARs, and watch others play.

So, generally I use AIs as mentors/instructors; not as productivity multipliers.

A big issue with AI is whether the results it produces are accurate or hallucinated. From am math, machine learning, and computer science view ... this is a meaningless distinction. AI is not like this game which has very strict and precise rules despite being based on RNG. For AI it is a vast space of concepts linked by RNG which is triggered by words and generated words.

That is why RAG is such a good use of AI. Commercial trained AIs without source data have a 5-15% hallucination rate (I am quoting industry benchmarks here). Now with RAG where each result must be cited from source documents, you come very close to less than 1% hallucinations. Then, the determinant is primarily your ability to provide high quality source information.

This is why reason why I use separate workspaces based on quality. You don't want twitch play lumped in and searched with 2by3 manuals.

How can I convert the CR's from WitW to CSV/Excel for the AI to look at? I asked in forum but no answer.

Take screen shot, and use a vision model.

bairdlander2 wrote: Wed Feb 04, 2026 9:54 pm How can I convert the CR's from WitW to CSV/Excel for the AI to look at? I asked in forum but no answer.

I was looking through the WIW forum and found that you can export a CSV file through the editor. I don't know if you can continuously update it for each turn but it's worth looking at.

bairdlander2 wrote: Wed Feb 04, 2026 9:54 pm How can I convert the CR's from WitW to CSV/Excel for the AI to look at? I asked in forum but no answer.

CoPilot was down; I did this with QWEN3-MAX, but I am sure QWEN3-VL 8B locally would be enough, but I was too busy to set it up.

Prompt: Please turn this table into a CSV. Thank you.
Unit Name,Nat,Loc,Type,HHQ,DDHq,Men,Guns,AFV,AC,Mtl,Exp,Fat,CV,%Tee,Toet,MP,TTDB,RF/RFs,SPri,Elt,Won,Lost
Laval US Air Base,USA,76,202,AirBase,USSTAF,-,0,0,0,0,59,0,0,0,Auto,-,-,3,-,0,0
Alencon US Air Base,USA,79,199,AirBase,USSTAF,-,0,0,0,0,64,0,0,0,Auto,-,-,3,-,0,0
Lorient US Air Base,USA,63,202,AirBase,USSTAF,-,0,0,0,0,56,0,0,0,Auto,-,-,3,-,0,0
Villacoublay US A.B.,USA,88,199,AirBase,USSTAF,-,0,0,0,0,59,0,0,0,Auto,-,-,3,-,0,0
Rennes US Air Base,USA,71,201,AirBase,9th US Air Force,-,2316,32,0,75,64,64,0,72,Auto,-,-,4,-,0,0
Maupertus US A.B.,USA,74,189,AirBase,9th US Air Force,-,2106,32,0,61,64,64,0,66,Auto,-,-,4,-,8,0
Vannes US Air Base,USA,66,203,AirBase,USSTAF,-,0,0,0,0,62,0,0,0,Auto,-,-,3,-,0,0
Metz US Air Base,USA,108,198,AirBase,USSTAF,-,0,0,0,0,53,0,0,0,Auto,-,-,3,-,0,0
Poix-Nord US A.B.,USA,90,190,AirBase,USSTAF,-,0,0,0,0,58,0,0,0,Auto,-,-,3,-,0,0
Brest US Air Base,USA,59,196,AirBase,USSTAF,-,0,0,0,0,56,0,0,0,Auto,-,-,3,-,0,0
St-Andre-de-l'Eure U,USA,85,197,AirBase,USSTAF,-,0,0,0,0,59,0,0,0,Auto,-,-,3,-,0,0
Coulommiers US A.B.,USA,93,199,AirBase,9th US Air Force,-,1836,32,0,64,61,61,0,58,Auto,-,-,4,-,0,0
Melun US Air Base,USA,92,202,AirBase,9th US Air Force,-,3706,32,0,138,66,71,0,111,Auto,-,-,4,-,46,0
Ft. Rouge BR A.B.,Bri,92,184,AirBase,2nd RAF Tactical Ai,-,1193,38,0,28,60,62,0,34,Auto,-,-,4,-,0,0
Bretigny US Air Base,USA,89,200,AirBase,9th US Air Force,-,2276,32,0,64,61,61,0,70,Auto,-,-,4,-,1,0
Chateaudun US A.B.,USA,84,203,AirBase,9th US Air Force,-,476,8,0,14,72,72,0,14,Auto,-,-,4,-,0,0
Orleans US Air Base,USA,67,205,AirBase,USSTAF,-,0,0,0,0,58,0,0,0,Auto,-,-,3,-,0,0
Dijon US Air Base,USA,102,212,AirBase,9th US Air Force,-,2986,32,0,104,58,59,0,90,Auto,-,-,4,-,18,0
Montdidier US A.B.,USA,92,192,AirBase,USSTAF,-,0,0,0,0,58,0,0,0,Auto,-,-,3,-,0,0
Dole US Air Base,USA,103,213,AirBase,9th US Air Force,-,2176,32,0,69,68,69,0,68,Auto,-,-,4,-,210,0
Woensdrecht BR A.B.,Bri,100,179,AirBase,2nd RAF Tactical Ai,-,0,0,0,0,53,0,0,0,Auto,-,-,3,-,0,0
Vitry-en-Artois BR A,Bri,93,187,AirBase,2nd RAF Tactical Ai,-,2926,64,0,110,64,64,0,85,Auto,-,-,4,-,120,0
Yenlo BR Air Base,Bri,108,180,AirBase,2nd RAF Tactical Ai,-,0,0,0,0,51,0,0,0,Auto,-,-,3,-,0,0
Schaffen BR Air Base,Bri,103,182,AirBase,RAF HQ,-,0,0,0,0,68,0,0,0,Auto,-,-,4,-,0,0
Florences US A.B.,USA,100,189,AirBase,9th US Air Force,-,0,0,0,0,70,0,0,0,Auto,-,-,4,-,0,0
Juvincourt US A.B.,USA,98,194,AirBase,9th US Air Force,-,3236,32,0,145,65,65,0,98,Auto,-,-,4,-,199,0
Eindhoven BR A.B.,Bri,105,179,AirBase,2nd RAF Tactical Ai,-,0,0,0,0,49,0,0,0,Auto,-,-,3,-,0,0
SHAEF,USA,73,196,ArmyGr,SHAEF,90,65760,0,0,70,69,0,0,65,80,48,-,3,-,153,37
UK Home Forces HQ,Bri,81,176,ArmyGr,UK Home Forces HQ,90,75400,0,0,86,70,0,0,94,100,46,-,3,-,0,0
RAF HQ,Bri,82,176,Army,RAF HQ,90,25120,0,0,63,63,0,0,125,100,50,-,3,-,20,5
USSTAF,USA,80,176,ArmyGr,USSTAF,90,24920,0,0,70,70,0,0,62,100,50,-,3,-,36,0
6th US Army Group,USA,98,213,ArmyGr,SHAEF,56,50020,0,0,68,67,0,0,99,100,49,-,2,-,15,6
12th US Army Group,USA,90,198,ArmyGr,SHAEF,72,29760,0,0,61,61,0,0,59,80,46,-,3,-,85,20
21st US Army Group,Bri,100,183,ArmyGr,SHAEF,56,34880,0,0,65,65,0,0,69,80,50,-,3,-,72,13
RAF Transp. Cmd,Bri,82,176,Army,RAF HQ,90,16520,0,0,70,70,0,0,82,100,50,-,3,-,0,2
RAF Fighter Command,Bri,82,176,Army,RAF HQ,90,15580,0,0,70,70,0,0,77,100,50,-,3,-,0,0

You could either write a custom APP or an MCP server with scroll and screen shot tools. Such that you just launch it, and it makes one combined table.

Yes, there are PC control AIs which you could tell how to do this, but I would not give an AI unrestricted access to my PC. By using MCP, you have sandboxed the AI to narrow functionality.

Commanderski wrote: Thu Feb 05, 2026 4:38 pm

bairdlander2 wrote: Wed Feb 04, 2026 9:54 pm How can I convert the CR's from WitW to CSV/Excel for the AI to look at? I asked in forum but no answer.

I was looking through the WIW forum and found that you can export a CSV file through the editor. I don't know if you can continuously update it for each turn but it's worth looking at.

I want to export CR of current game running, the editor just exports all the raw data for scenario's, cannot use any saved games in progress info

Up to this point, the game expert has mainly worked with unstructured text: manuals, guides, YouTube transcripts ... to respond to questions. And this approach has done well. It has done very well, because beyond answering questions I am learning many details that I might have missed.

I am about to take it into an entirely new direction using highly structured text. The target game will be Cold Waters heavily modded. Platforms, weapons, ... are all structured text entries. (for you programmers very much like JSON or YAML files).

For you programmers and/or game devs, the game's database is a parent-child simple relational database of maybe depth 3 or 4. Another view of the game's database is just a few hierarchical databases like PLATFORM -> WEAPON or ABILITIES -> SENSOR or PAYLOAD.

So, in SQL or QBE, you would be doing JOINs to bring back information. With an AI and VDB, the sequence of prompts with previous data returned and pasted becomes the JOIN. Further the AI should be able to reason over the entire database something which you don't get with generic database tools. And these days on could easily create some tools defined to the AI to feed this in a workflow that eliminate the pseudo-JOINs.

I am envisioning the AGENT and RAG as the XO to the ship's captain.

But this would generalize again to a game such as this which I think 2by3 uses CSVs.

This would allow you to ask an AI to reason over an enemy formation you are facing in a given year, and perhaps compare the more general match up than just hex level CV estimates.

Well, Cold Waters is a good place to start as it is a much simpler game than WITW or WITE-2.

PS: For those of you who have been considering this, local (small AIs; needing 2-6 GB VRAM) are becoming much more capable models. See Alibaba's QWEN3x family and/or the just released Google Gemma4.

This a very good illustration of how this technology can add value to your gaming.

My question was regarding the utility of special units. Now, I know that present in the WORKSPACE were only transcripts ripped from YouTube let's play videos. You are never going to find this information yourself in 500 hours of video ... even if you watch it, you will not remember. But an LLM + VDB will find this information, and even put spoken narrative into a useable table for you.

You can see this was solely from videos as you can see the source names provided on the right.

Follow up report. The use of files such as these for the most part failed to produce coherent responses. I will note this was done with a high quality embedder; qwen3-embedding:8b-q8_0 via Ollama. I suspect that the AI failed to understand the field structure and field values (some of which were primary and secondary keys). Furthermore, the filenames themselves are part of the key relationships, but I don't believe the names are embedded; only stored as source documents. Oh, well ... (BTW, one suggestion by Deepseek was to pre-process such files into JSONs which is pretty much an isomorphic format which is natively understood by LLMs.)

For illustration, this the contents of the file, "ch_ss_kilo.txt", obviously a Chinese diesel electric:

Designation=SS
ShipType=SUBMARINE
PlayerHUD=hud/default

Length=74
Beam=9.9
HullHeight=0.08
Displacement=3950
Crew=52
Range=7500
HullNumbers=FALSE

[Movement]
SurfaceSpeed=11
SubmergedSpeed=19
AccelerationRate=0.1
DecelerationRate=0.06
RudderTurnRate=0.5
TurnRate=2.2
PivotPointTurning=0.256
DiveRate=5
SurfaceRate=5
BallastRate=0.25
Waterline=0.04
SubmergedAt=0.147
CavitationParameters=20,-100
PropRotationSpeed=-220
TestDepth=787
EscapeDepth=787
PeriscopeDepthInFeet=41
Diesel=TRUE

[Acoustics & Sensors]
SelfNoise=120
ActiveSonarReflection=7
ActiveSonarModel=rus-mgk500
PassiveSonarModel=rus-mgk500
TowedArrayModel=FALSE
AnechoicCoating=TRUE
RADAR=wp_snoop_tray
RADARSignature=SMALL

[Weapon Systems]
TorpedoTypes=wp_test-71mke,wp_rpk-6,wp_moss_mg-74
TorpedoNumbers=8,4,6
TorpedoTubes=6
NumberOfWires=6
TubeConfig=1,1,1,1,1,1
TorpedoTubeSize=533,533,533,533,533,533
TubeReloadTime=30

AntiAirMissileType=wp_strela-2
AntiAirMissilesNumbers=8
AntiAirMissilesPerTarget=1
AntiAirMissileReloadTime=12
AntiAirMissileLaunchDepth=20
AntiAirMissileLaunchPoints=-0.002,0.14,0.1398
AntiAirMissileLaunchPointsRotation=-90,0,0

NoisemakerName=wp_noisemaker
NumberOfNoisemakers=20
NoisemakerReloadTime=20

MissileType=wp_rpk-6
MissilesPerLauncher=2,2

[Subsystems]
LabelPosition=364,28
BOWSONAR=FRONT
LabelPosition=0,0
TOWED=FALSE
LabelPosition=409.7,171.9
PERISCOPE=SAIL
LabelPosition=160.8,193.2
ESM_MAST=SAIL
LabelPosition=409.7,193.3
RADAR_MAST=SAIL
LabelPosition=409.7,151.4
TUBES=FRONT
LabelPosition=161,151.4
FIRECONTROL=FORE
LabelPosition=315,49
PUMPS=FORE
LabelPosition=89,49
PROPULSION=AFT,REAR
LabelPosition=59.9,28
RUDDER=REAR
LabelPosition=59.9,151.4
PLANES=REAR,FRONT
LabelPosition=204,49
BALLAST=MID
LabelPosition=187,28
REACTOR=MID

[Flooding Display: x,width,ymin,ymax]
FLOODING1=123,90,-80,-37
FLOODING2=31,93,-80,-37
FLOODING3=-45,57,-80,-38
FLOODING4=-98,47,-76,-39
FLOODING5=-146,46,-73,-41
DamageControlPartyY=-30

[Model]
ModelFile=ships/wp_ss_kilo/wp_ss_kilo

//Hull and Deck
MeshPosition=0,0,0
MeshRotation=0,0,0

MeshHullCollider=wp_ss_kilo_coll_hull
MeshSuperstructureCollider=wp_ss_kilo_coll_ss

Material=ships/wp_ss_kilo/plan_ss_kilo_mat
AssetBundle=bundles/ships/epicmod_textures
MaterialTextures=assets/ships/epicmod_textures/plan_ss_kilo_tx.png,assets/ships/epicmod_textures/wp_ss_kilo_spec.dds,ships/wp_ss_kilo/wp_ss_kilo_nm
Mesh=wp_ss_kilo
MeshPosition=0,-0.048,-0.4662
MeshRudder=wp_ss_kilo_rudder
MeshPosition=0,0.0641,0.2571
MeshBowPlanes=wp_ss_kilo_bowplanes
MeshPosition=0,0,-0.4804
MeshSternPlanes=wp_ss_kilo_sternplanes
MeshPosition=-0,-0.0045,-0.5295
MeshProp=wp_ss_kilo_prop

[Model]
ModelFile=ships/wp_ssn_alfa/wp_ssn_alfa
Material=ships/materials/wp_masts
MeshPosition=0,0.030,0.160
MastHeight=0.085
MeshMast=wp_ssn_alfa_periscope

MeshPosition=0,0.030,0.130
MastHeight=0.085
MeshMast=wp_ssn_alfa_esm

MeshPosition=0,0.030,0.110
MastHeight=0.085
MeshMast=wp_ssn_alfa_radar

ModelFile=ships/wp_ss_kilo/wp_ss_kilo

[Weapon Systems]

//Torpedo Mounts

//Tube 1 - Missile Mount
MeshPosition=-0.01,0.03,0.496
MeshMissileMount=FALSE
MissileEffectPosition=-0.01,0.03,0.496

//Tube 2 - Missile Mount
MeshPosition=-0.01,0.03,0.496
MeshMissileMount=FALSE
MissileEffectPosition=-0.01,0.03,0.496

//Tube 1
MeshRotation=0,0,0
MeshPosition=-0.01,0.03,0.496
TorpedoSpawnPosition=0.01,0.03,0.598
TorpedoEffectPosition=0,0,0.0232

//Tube 2
MeshRotation=0,0,0
MeshPosition=0.01,0.03,0.496
TorpedoSpawnPosition=-0.01,0.03,0.598
TorpedoEffectPosition=0,0,0.0232

//Tube 3
MeshRotation=0,0,0
MeshPosition=-0.01,0.0126,0.504
TorpedoSpawnPosition=0.01,0.0126,0.606
TorpedoEffectPosition=0,0,0.0232

//Tube 4
MeshRotation=0,0,0
MeshPosition=0.01,0.0126,0.504
TorpedoSpawnPosition=-0.01,0.0126,0.606
TorpedoEffectPosition=0,0,0.0232

//Tube 5
MeshRotation=0,0,0
MeshPosition=-0.0274,0.0126,0.496
TorpedoSpawnPosition=0.0274,0.0126,0.598
TorpedoEffectPosition=0,0,0.0232

//Tube 6
MeshRotation=0,0,0
MeshPosition=0.0274,0.0126,0.496
TorpedoSpawnPosition=-0.0274,0.0126,0.598
TorpedoEffectPosition=0,0,0.0232

//Noisemaker Mount
MeshRotation=0,0,0
MeshPosition=0,-0.056,-0.533
MeshNoisemakerMount=FALSE

//Wake
MeshPosition=0,0,0
BowWaveParticle=ships/particles/bowwave_sub
PropWashParticle=ships/particles/bowwave_small
AssetBundle=bundles/ships/usn_ssbn_ohio
CavitationParticle=assets/ships/usn_ssbn_ohio/usn_ssbn_ohio_cavitation3.prefab
FunnelSmokeParticle=ships/particles/wp_ss_kilo_bubbles
EmergencyBlowParticle=ships/particles/wp_ss_kilo_emergency_blow
CavitationParticlePosition=0,-0.005,-0.54

ParticleBowWavePosition=0,0,0.52
ParticlePropWashPosition=0,-0.033,-0.48

ParticleHullFoamPosition=0,0,0.338
ParticleHullFoamParameters=10,0.2,0.2,0.055,0,0,0.25

ParticleSternFoamPosition=0,0,-0.43
ParticleSternFoamParameters=25,0.25,0.15,0.025,0.01,0,0.25
KelvinWaves=1.46,1.46

EngineAudioClip=audio/units/submarine_engine_1
EngineAudioRollOff=LOGARITHMIC
EngineAudioDistance=1,20
EngineAudioPitchRange=0.4,1
PropAudioClip=audio/engine/diesel.ogg
PropAudioRollOff=LOGARITHMIC
PropAudioDistance=1,20
PropAudioPitchRange=0.1,0.6
PingAudioClip=audio/sonar/sonar5.ogg
PingAudioRollOff=LOGARITHMIC
PingAudioDistance=1,50
PingAudioPitch=1