Building the RtW2 Combat Simulator

[tortugapower]

Hi all,

I'd like to discuss and get feedback on ideas for the RtW2 combat simulator mod which I am developing. oaktree is my main consultant, and he's absolutely top-notch for naval knowledge and insight into programming challenges. At this stage, however, it's close enough to completion that I'd like to open it up to more people.

Prototype for this is Python. In my head, I pretend I'll go back and redo it in C++ when it's all finished, but... unless Fredrik wants to include this officially, that's unlikely.

Let me start with the high-level design. Flow of simulator right now (with approx. completion):

1. Read .SAC (combat) file and create relevant objects (ships, fleets, etc.). [100%]
   
   
2. Initialize starting conditions (weather, distance of fleets, etc.). [90%]
   
   
3. Spotting and long-distance air combat phase, if either fleet has carriers. [90%]
   
   
4. Combat phase, similar to my RtW1 simulator but with carriers. [70%]
   
   
5. Combat resolution (including to update the .SAC file). [90%]

---

The first thing I'd like to review is #3, "scouting and long-range air combat": the additions needed before the surface fleet (RtW1-style) combat begins. That is, the carrier task force engagement at range -- the Midway-style battle.

In my simulator, fleets close distances as they approach normal surface (RtW1) combat. Scouting/LR air combat currently goes as follows:

1. At every discrete closing distance, if enemy fleet has not been detected, scouts are sent out.
   
2. Detection chance for scouts is approximated based on circumference of a circle. It assumes an out-and-back (line) or out + 90-degree turn + new line back. Fleet visibility is increased per ship linearly (should probably increase as only square root).
   
3. If enemy fleet has been detected, bombers + escorts are sent out. First air combat (vs. any CAP) then air strikes (bombers vs. ships) are resolved.
4. Repeat A-C until the distance has been closed to surface-combat range.

Presently, I have not implemented a CTF-only battle scheme, where there is no surface battle.

Here are sample air-to-air combat results for two identical fleets. mostly battleships with 2 fleet carriers:


The fighter escorts of Fleet 1 (Japan) encounter the CAP of Fleet 1 (Japan).
6 of the 8 attacking airplanes are shot down.
2 of the 20 defending airplanes are shot down.
The bombers of Fleet 1 (Japan) encounter the CAP of Fleet 1 (Japan).
6 of the 52 attacking airplanes are shot down.
7 of the 18 defending airplanes are shot down.

46 bombers approach the fleet, and encounter long-range flak (DP guns).
9 bombers are disrupted, 2 are destroyed on approach.
3 wings begin a bombing run.
-> Wing 1 (12 bombers) approaches Japan Light Cruiser Div 12.
Jintsu shot down 0 bombers, disrupted 1, and was attacked by 2: 2 hits.
(CL) Jintsu is sinking below the waves!
-> Wing 2 (12 bombers) approaches Japan Battle Div 6.
Azuma shot down 0 bombers, disrupted 0, and was attacked by 9: 4 hits.
(BB) Azuma is sinking below the waves!
Aso shot down 0 bombers, disrupted 0, and was attacked by 3: 1 hits.
-> Wing 3 (11 bombers) approaches Nisshin *.
Nisshin shot down 0 bombers, disrupted 0, and was attacked by 11: 3 hits.
I wonder, do the above numbers seem to make sense? Take, for instance, 18 fighters attacking 52 bombers, does a 6:7 kill ratio seem too bomber-favored? (That was one random roll, but let's pretend it's representative.)

I guess I'll treat this as part blog, part opportunity for feedback and questions. Pinging garrisonchisholm in particular 

[garrisonchisholm]

Well, there's just too many variables to say. That could happen, yes. If aircraft and crews are of equal quality and there is no surprise? It feels a titch steep on CAP casualties, but if they were Really good attackers with tight gunnery arcs that stayed in formation, maybe.

If I could take a step back though, what you are attempting to do is maintain Ship/Plane performance data "usage", so that the units the player has designed/ordered maintain their distinctive performance even in the sim? That sounds like a lofty goal, but attainable with suitable attention and work. My approach would have been a 'dumbed down' one with a "user beware" sticker on it, essentially simply randomizing winner/loser & losses off of a pure tonnage comparison, but if the specific units approach you're taking could work I am sure everyone would love it. :]

Back to your original question, i.e. are the combat results plausible, yes, but all things being equal I would expect 18 CAP to gain something closer to a 2-1 kill ratio, because presumably they would be able to use altitude and hence speed to their advantage, which feels to me like the biggest factor. Run 100 samples of the same encounter, and then let's see what our averages are.

[tortugapower]

Oct 21, 2019 17:40:44 GMT -6 garrisonchisholm said:

Well, there's just too many variables to say. That could happen, yes. If aircraft and crews are of equal quality and there is no surprise? It feels a titch steep on attacker casualties, but if they were Really good attackers with tight gunnery arcs that stayed in formation, maybe.

...

Back to your original question, i.e. are the combat results plausible, yes, but all things being equal I would expect 18 CAP to gain something closer to a 2-1 kill ratio, because presumably they would be able to use altitude and hence speed to their advantage, which feels to me like the biggest factor. Run 100 samples of the same encounter, and then let's see what our averages are.

Thank you, that's valuable feedback. I only have a vague sense of what's realistic/historical as far as fighter vs. bomber (or fighter vs. fighter) losses. I know that depends a lot on the fighter crew and bomber crew and the types of planes involved. This is a generic fighter with average crew versus a generic fighter (or bomber) with average crew.

If I could take a step back though, what you are attempting to do is maintain Ship/Plane performance data "usage", so that the units the player has designed/ordered maintain their distinctive performance even in the sim? That sounds like a lofty goal, but attainable with suitable attention and work. My approach would have been a 'dumbed down' one with a "user beware" sticker on it, essentially simply randomizing winner/loser & losses off of a pure tonnage comparison, but if the specific units approach you're taking could work I am sure everyone would love it. :]

Essentially, yes: I'm doing things a very detailed manner. There will be many dice rolls for each phase of combat (even for each round of each phase). You and I have discussed this a bit before, and I think your system (a single percent chance for the whole result) would be very quick and probably easy to implement.

My complicated system has a number of benefits over a one-off roll of the dice -- if used correctly, complexity should be a tool to better inform the calculations. I think the main disadvantage is the system is very complicated, so designing it is a pain. But I'm here already, so that time has already been sunk. =)

[dizzy]

Oct 21, 2019 17:40:44 GMT -6 garrisonchisholm said:

My approach would have been a 'dumbed down' one with a "user beware" sticker on it, essentially simply randomizing winner/loser & losses off of a pure tonnage comparison...

Please, no! An auto-calc battle should take Ship Crew experience, tonnage, gun size, # of guns, fire control, etc., into account before auto-calc anything. That goes for airstrikes as well. Crew quality is huge, speed of the airplane is huge as it will determine how many passes the fighter can make on a bomber before it reaches the target and things like toughness of the aircraft and firepower are all important. If we could add them all up, take the square root and use the difference of one another in percentage to determine the modifier to the dice roll, or however you decide to do it, I'd be confident knowing the auto-calc would do an adequate job. But just remember that tonnage means nothing when encountering carrier strike groups. Often battle groups don't come into contact with one another, as an inferior group with higher speed can maintain it's distance from a superior group and only use aircraft to engage.

If this sim takes the above into account, then it will have so many variables that battles will often be dissimilar in results. I looked at your numbers, but this is only one example and there's not enough to know about the aircraft involved. Sometimes the aircraft, crew and technology are so lopsided, you'll get situations like the Great Marianas Turkey Shoot that happened during the Battle of the Philippine Sea. I really hope your auto-calc sim can take those into account.

As a kid, my dad used to watch the Black Sheep Squadron with me about F4U corsairs that were stationed on an island somewhere in the Pacific and he helped me build ww2 model planes. I'm a total enthusiast in this era and I'm so thrilled you're thinking of adding as much detail as you can because it will turn out to be fantastic if you're able to do that.

[garrisonchisholm]

Fear not Dizzy, not only is Tortuga long embarked upon his particular plan, I have no technical aptitude to execute anything else myself. Bear in mind though the issue I was attempting to address in my previous consultations; a way to take a pass on the tedium sometimes experienced when your war is well in hand and you're just waiting for it to end. If the user has a battle where they simply don't care about the losses because the odds in the player's favor are so high that they must be negligible losses, then I just thought an "Easy" button would be the answer. I never meant to imply that my methodology was anything more than a time-saver, meant to abstract a result in 5 seconds rather than 20 minutes.

To the OP though, if you have something close to a working model, I think the key is to run a large number of simulated encounters and present those results. That will give us something to analyze. :]

[tortugapower]

Yes, it's a fair point garrisonchisholm . The added complexity will add almost nothing in the use-case of not caring much about the result. I wanted this to be robust enough to simulate any combat with a fair result, even when the war is still in doubt.

Statistics forthcoming are here! For a fighter vs. fighter dogfighter:


Losses (Left | Top) 5 10 15 20

4 (1.49 | 1.14) (3.1 | 0.9) (3.82 | 0.85) (3.99 | 0.81)

9 (1.21 | 2.9) (3.03 | 2.58) (4.92 | 2.28) (6.88 | 2.16)

14 (1.06 | 4.29) (2.68 | 4.49) (4.46 | 4.09) (6.47 | 3.73)

19 (1.07 | 4.86) (2.53 | 6.28) (4.15 | 5.88) (6.03 | 5.6)

(edit: so this is a way of determining if numerical superiority is being properly modeled by the simulator)

I'll post similar for fighter vs. bomber.

[tortugapower]

Here are some statistics for bombers (left numbers) being attacked by CAP (top).

Losses (Left | Top) 5 10 15 20

5 (1.44 | 0.3) (3.3 | 0.26) (4.54 | 0.23) (4.89 | 0.19)

10 (1.2 | 0.67) (2.95 | 0.59) (4.76 | 0.58) (6.85 | 0.49)

20 (1.08 | 1.57) (2.5 | 1.38) (4.06 | 1.24) (5.85 | 1.16)

30 (0.98 | 2.31) (2.23 | 2.18) (3.72 | 2.09) (5.34 | 1.88)

40 (0.92 | 3.14) (2.1 | 3.02) (3.44 | 2.91) (4.92 | 2.81)

While I think the fighter vs. fighter looks okay, I'm not satisfied by this (cap vs bombers).

[dorn]

Oct 22, 2019 14:55:49 GMT -6 tortugapower said:

Here are some statistics for bombers (left numbers) being attacked by CAP (top).

Losses (Left | Top) 5 10 15 20

5 (1.44 | 0.3) (3.3 | 0.26) (4.54 | 0.23) (4.89 | 0.19)

10 (1.2 | 0.67) (2.95 | 0.59) (4.76 | 0.58) (6.85 | 0.49)

20 (1.08 | 1.57) (2.5 | 1.38) (4.06 | 1.24) (5.85 | 1.16)

30 (0.98 | 2.31) (2.23 | 2.18) (3.72 | 2.09) (5.34 | 1.88)

40 (0.92 | 3.14) (2.1 | 3.02) (3.44 | 2.91) (4.92 | 2.81)

While I think the fighter vs. fighter looks okay, I'm not satisfied by this (cap vs bombers).

It is even more comlicated as main task of CAP is to deny attack from bombers, it is not needed to shot them down. Even fighters without ammunition was able to bomber to cancel bomb run.

Another thing is that numbers plays important role.

[tortugapower]

I'm definitely glad you're here dorn because I'm sure your feedback will be invaluable. I could add a dispersion factor to CAP attacking bombers -- is that what you're suggesting?

Right now each plane only attacks once, and I think I will get more fair results (at least for CAP vs. bombers) if I allow a plane who achieves a victory to roll another chance for a kill. That way, with high kill %, less fighters can get more kills than even their own numbers, if they are advanced enough compared to their prey.

[dorn]

Oct 22, 2019 15:26:44 GMT -6 tortugapower said:

I'm definitely glad you're here dorn because I'm sure your feedback will be invaluable. I could add a dispersion factor to CAP attacking bombers -- is that what you're suggesting?

Right now each plane only attacks once, and I think I will get more fair results (at least for CAP vs. bombers) if I allow a plane who achieves a victory to roll another chance for a kill. That way, with high kill %, less fighters can get more kills than even their own numbers, if they are advanced enough compared to their prey.

It is really complex and most difficult thing is how to make it simple and still taking all values which are important in RTW into considaration.

Some brainstorming about CAP fight vs. bombers.

1. CAP interception check - each of these tech should improve chance to intercept bombers. It should not be automatic and especially early it was difficult (you need to be in right position, right altitude etc)

a) Early air search radar

b) Improved air search radar

c) VHF radio in aicraft

d) Air intercept control

e) CIC

f) Deck launched interceptors

2. CAP combat

It could be CAP vs. bombers alone but bombers can be escorted by fighters and this does not mean that CAP need to fight first escorts

From this there should be:

CAP ... some damaged, shot down, empty ammunition, ready. "Damaged" is for tactical part same as "Shot down" but not for strategic layer. "Empty ammunition" means unavailable for some time (I think we can abstract from fact that these fighters could still disrupt enemy bombers).

Escort .. same as CAP

Bombers ... destroyed, damaged, disrupted, make bomber run

3. AA effectivity

Heavy AA

Medium AA

Effects on bombers: bomber shot down, damaged (continue run with lowe accuracy), damaged and disrupted, disrupted (lowered accuracy by same range), unopposed

4. bomber run

With experience there should be higher chance for torpedo bombers doing attack from different side highly increasing chance of hit

[tortugapower]

dorn In theory I agree. However, contrary to my discussion with Gary, I'm not planning to go into *that* level of detail. I'm not even sure if RtW2 goes into such detail.

My current system is to project all of that into bins of either "destroyed" and "dispersed", and I think we can still have a good (or "good enough") simulator at that level. Plus, it's fun to think of a clever way of abstracting the minutiae that you detailed.

However, if you want to write the code to do this at that detailed level, I'm happy to include it.

[dorn]

Oct 22, 2019 16:05:55 GMT -6 tortugapower said:

dorn In theory I agree. However, contrary to my discussion with Gary, I'm not planning to go into *that* level of detail. I'm not even sure if RtW2 goes into such detail.

My current system is to project all of that into bins of either "destroyed" and "dispersed", and I think we can still have a good (or "good enough") simulator at that level. Plus, it's fun to think of a clever way of abstracting the minutiae that you detailed.

However, if you want to write the code to do this at that detailed level, I'm happy to include it.

I agree with you but at least I will take technology at least of some effects. There is need to have better result if you have better plane, better technology (doctrine) or better experienced crew.

[dizzy]

I think an important note about bombers that we mostly all know is that they will drop their ordnance when being shot at and damaged will then attempt to evade. After that, the CAP fighters are off to harass another attacking bomber and the former makes their best effort to limp home if they can or they ditch. A lot of the times, no one gets shot down. So hopefully the proper 'weight' can be placed on disruption and dispersion when dealing with CAP. But I like dorn's other variable mentions like AA and such which all needs to be a factor in whether the bombers make a successful run.

There's also something to say about the dual attack factor, and this is something that was seen in the battle of Midway. Torpedo bombers can pull off CAP flying at higher altitude and clear the way for unopposed dive bomber attacks. So there might be a combined attack bonus present.

Also, there's the matter of speed. This reduces reaction time which makes bombers so much more effective and the only real way to deal with that was the complexity of dorn's mentioning all the components of detection, communication and response which factor in a lot of myriad techs to work in conjunction to counter faster bombers.

GL, tortugapower it's a lot of stuff. I have low expectations and high hopes!

[samweston]

This would be nice to have, and it will be interesting to see how well it ends up handling air combat.

Sadly I know that it won't be able to simulate the real situations, such as a squadron breaking into two flights to engage escorts and bombers during cap. But it stands to remember, fighters are the primary tool for actually bringing down aircraft. Flak and high volume guns such as 40mm, and 20mm are to make ships less appealing as a target, and something as small as .50 is only when you have no other option. What is going to get the most murky is when SSM and SAM get introduced, as they act in a hybrid role as a weapon for damage and denial.