pandakraut

Here are some of the behind the scenes mechanics I've learned about while modding. I've posted about several of these previously but wanted to get them all compiled into a single location. Weapon Recovery Rates Colonel difficulty rates Win - supplyCapturePercent: 0.5 - weaponCapturePercent: 0.3 - weaponCaptureUnitPercent: 1 - weaponSavePercent: 0.5 Draw - supplyCapturePercent: 0.5 - weaponCapturePercent: 0.15 - weaponCaptureUnitPercent: 1 - weaponSavePercent: 0.3 Loss - supplyCapturePercent: 0.5 - weaponCapturePercent: 0.05 - weaponCaptureUnitPercent: 1 - weaponSavePercent2: 0.1 Brigadier General difficulty rates Win - supplyCapturePercent: 0.5 - weaponCapturePercent: 0.2 - weaponCaptureUnitPercent: 0.75 - weaponSavePercent: 0.5 Draw - supplyCapturePercent: 0.5 - weaponCapturePercent: 0.15 - weaponCaptureUnitPercent: 0.75 - weaponSavePercent: 0.25 Loss - supplyCapturePercent: 0.5 - weaponCapturePercent: 0.05 - weaponCaptureUnitPercent: 0.75 - weaponSavePercent: 0.1 Major General and Legendary Win - supplyCapturePercent: 0.5 - weaponCapturePercent: 0.1 - weaponCaptureUnitPercent: 0.5 - weaponSavePercent: 0.3 Draw - supplyCapturePercent: 0.5 - weaponCapturePercent: 0.05 - weaponCaptureUnitPercent: 0.5 - weaponSavePercent: 0.15 Loss - supplyCapturePercent: 0.5 - weaponCapturePercent: 0.05 - weaponCaptureUnitPercent: 0.5 - weaponSavePercent: 0.1 - weaponCaptureUnitPercent is currently not used in the base game so no weapons are recovered from captured units. - You will still get the weapons from the men that were killed in a unit that is captured. Shatter Mechanics - infantry: hp% < .25 and morale% < 0.1 = shatter. hp% < .1 = shatter - artillery: hp% < .3 and morale% < 0.2 = shatter. hp% < .1 = shatter - cavalry: hp% < .4 and morale% < 0.2 = shatter. hp% < .1 = shatter - skirmishers: hp% < .35 and morale% < 0.2 = shatter. hp% < .1 = shatter - Shattering is worse than capturing in terms of the AISize. Capture is the only way to fully remove an entire unit from the AIsize. - Shattered units do allow you to recover their weapons. - AI tech and training are also affected by reducing the AISize. Scaling The scaling algorithm itself is rather complex but I'll try to summarize it. For every battle each AI unit has a default size. This size will increase or decrease based on several different factors. The following is not a complete list but is generally sufficient for most players purposes. 1) Difficulty: As seen at the campaign start MG and Legendary will cause sizes to go up. 2) Randomness: There is a random size modifier that is generated for each battle that will apply to all AI units. 3) Manpower Pool: There is an AI Size variable which is where the intelligence report size value comes from. This is where snowballing can happen if you aren't killing enough units in each battle. I'll provide some detail as to how this can be controlled later on. 4) Your army composition: In every battle your army makeup is compared to the AI's army makeup. This includes any allied units you get in the battle and any units you do not deploy. This comparison basically uses the average size of each type of unit though infantry size can also affect the size of other units. In side battles this allows you to lower your average unit size by creating temporary ballast units of the minimum size in corps that don't deploy. This does not work in major battles, but can sometimes appear to work because there are minimum floors in the algorithm. If ballast units appear to be lowering scaling in a major battle you just need to add more men in any way. This is calculated at battle start so on multiday battles you can wait to increase men until day 2 to limit scaling as well. 5) Historical force size factor: Each battle has its own factor that tries to enforce a specific ratio of forces. For example CSA Antietam leans toward the player being outnumbered 2:1. Union Antietam is closer to 1:1. Other factors can definitely override this. Tips: You can have a significant effect on the AI's average unit size by keeping your own small. For example, I play most of either campaign with units around 1k. Units will get a bit larger as it goes own but rarely do I go much about 1.3k. Even without ballast units this will largely keep AI units as small as they can be given other factors. You can also go the completely opposite direction and make your units max size to boost AI scaling as much as possible. The more men the AI has the more weapons you can recover and the more xp you can farm. When fielding a majority of max size infantry units scaling will not be able to keep up even on legendary due to unit size caps and the AI fielding other unit types. This means that it's possible to outnumber the AI in nearly every battle with max size units even on legendary. The main thing to compare when looking at other peoples campaigns is the intelligence reported size and their average unit size. The AI Size minimum is 50k for most of the campaign past the 1st 3rd. Here are the numbers for the Union Campaign minimum AISize, valid for all difficulties: https://old.reddit.com/r/ultimategeneral/comments/95stw1/what_does_army_size_actually_mean_and_is_it_bad/e4499mf/ From battle to battle, the only damage you can deal to the AI is reducing the AI Size, indicated by the intelligence report size percentage. This is accomplished by killing its units. Capturing a unit is the only way to completely remove its numbers from the AI Size. Men remaining in a shattered unit escape. Lowering the AISize has an impact on training and technology. However, all three values have a minimum amount based on the current timeline of the campaign. For example, if the AI size is 60k prior to 2nd bull run. You inflict 40k casualties and then post battle the AISize is now 65k. The resulting size is lower than it could have been, but limited by the minimum. On the harder difficulties this means that it is not always worth it to kill everything if you take to many losses to do it as the AI always gets 'free' troops back. It is also possible to not get enough kills and then the AISize will start to grow beyond the minimum. Also note, the AISize is not actually a limit on what you will face in battle. A variety of scaling factors can push that number higher or lower as described above. Optimal Unit Sizes for Damage These vary slightly based on weapon damage. The benefit of adding men/guns decreases significantly prior to these limits so using a slightly smaller size will be more optimal if money is a concern. - max artillery size for ranged damage ~14 - max skirmisher size for ranged damage ~375 - max infantry size for ranged damage ~1845 - The damage penalty for infantry is much smaller than for other unit types. This means that 2500 man units with detached skirmishers will deal more total damage when their volleys are combined. - mounted cavalry ranged attacks never fall off due to unit size. Larger units will always deal more damage. - The benefits of adding more men to mounted cavalry starts to yield decreasing returns between 500-600. - dismounted cavalry have the same optimum size as normal skirmishers at ~375 - melee damage will always continuously go up based on size though for some unit types it will do so at a diminishing rate. Surrender Mechanics Surrender checks occur when a certain damage threshold is met on a given frame. This means during melee or during a specific volley dozens of checks can occur. Melee produces checks more consistently than firing. Each check generates a random value that is compared against a surrender probability. It is nearly impossible for a close to full strength unit to surrender, but it is possible to get lucky. Artillery units will never make a surrender check. A unit will never make a surrender check if an allied unit that is not routing is nearby. This is part of the reason why splitting units up is very important. Surrender probability increases based on - morale - hp / hp max - hp ratio of all nearby allied units that are not routing vs all nearby enemy units. This is one of the larger factors. - enemy coverage in 360 degrees. Hard to quantify this more, but that's what the code appears to doing. - current ammo / max ammo - officer is dead(a wounded officer has no impact) Condition has no effect beyond reducing a units melee damage so that they can't inflict casualties and thus their morale drops faster. Low condition does reduce unit speed which makes it easier to stay on top of the unit. Map edges also do weird things with melee contact points and can cause units to stop doing the normal amounts of damage. Sometimes this results in both units just exhausting and routing themselves and they'll just slowly grind each other down. Battle Mechanics - Higher elevation gives increased accuracy/damage and morale regeneration. - Weapons always hit but can do < 1 damage with enough negative modifiers. - Base Damage Calculation is Weapon damage * Random value between AccuracyLow / 100 and Accuracy High / 100. - Artillery Shell shot modifies damage by 1.2 and accuracy low by .75. - Artillery Canister shot modifies damage by 1.75 and accuracy low by .5. - Canister shot is out to .25 of maximum range. - Shell shot is from .25 to .5 of maximum range. The spreadsheets attached below show the damage degradation as range increases for each weapon. You must be logged into the forum to access them. Rifles have a RangeHint column which contains the range value displayed in game UI for each weapon. This value frequently does not match the weapons actual range. The RangeHint value helps show the player which rifles are more effective at longer ranges, but to me are misleading in some cases. Keep in mind that some of these stats in isolation can be misleading. Nothing replaces actually trying out weapons yourself. For artillery I would also highly encourage you to read this guide first https://steamcommunity.com/sharedfiles/filedetails/?id=1105446690. The conclusions within match up with the data very accurately and are probably a better starting point. ArtilleryWeaponCurves.xlsx CavalryWeaponCurves.xlsx InfantryWeaponCurves.xlsx SkirmisherWeaponCurves.xlsx WeaponStats.xlsx DamageFalloffBySize.xlsx @TechnoSarge has also done some additional analysis on price effectiveness and dps which you can find here. Summary: http://forum.game-labs.net/topic/26142-hidden-mechanics-and-weapon-damage-degradation/?page=2&tab=comments#comment-574298 Analysis spreadsheets: http://forum.game-labs.net/topic/26142-hidden-mechanics-and-weapon-damage-degradation/?page=2&tab=comments#comment-573686 Some nice average damage spreadsheets put together by @Sir Galahadn't https://www.dropbox.com/s/adm7wm62aja9g53/Infantry Weapon Curves.xlsx?dl=0 https://www.dropbox.com/s/yk0fm8e86m9olh4/Artillery Weapons Analysis.xlsx?dl=0

That would make sense. Thanks for clearing it up. With getting all the Texas rifles, there are limiters in place that prevents you from going above the battle defaulted weapon selections. I would guess this in place to prevent you from getting later period weapons early. So if the tech level is below the battle default there is some variance, otherwise you get the battle defaults. I would have to do a bunch of specific testing to comment in any more detail.

I've seen this idea in various places but I think most of what you are seeing related to weapon scaling is coincidence. If you setup a 10 recon campaign you can see that the AI weapons do not change when you change your own. I took a Union pre Antietam save where most of my army was equipped with Harpers Ferrys and most of my artillery was 24pdrs and 20pdrs. CSA was equipped with Harpers Ferrys and an assortment of 3in, 12pdrs, and napoleons. Switching out all of my units to 1842s, and as cheap of cannon as were available I saw no change in the AI's equipment. I repeated this test on a save prior to 2nd bull run. Replaced all my weapons and retreated from the battle immediately. I again saw no change in the AI's equipment at Antietam. In both cases I also sold all of my weapons in case the armory was checked. As best I am aware, each AI unit has a defined weapon which will influence the starting point of the AI's weapon selection. Then based on the AI's tech level, a cost coefficient(making the AI more likely to select certain weapons), and a slight random factor the unit's weapon can change. I think this calculation only occurs at the end of the previous battle.

Yes, we reduced the max at some point.

As far as I can tell the AI doesn't scale based on your veterans or weapons. This is an area I've researched less however so I could be missing something. I think you're safe to just build the best army you can though.

Artillery prices were increased somewhat in the latest version to increase the cost of the player stockpiling the best cannon very quickly. Investing in economy should mitigate that fairly well though.

The same concepts apply on lower difficulties, but because of the difficulty factors scaling is much less impactful. I would agree you can completely ignore scaling on BG.

Just wanted to add that the intelligence report values do not actually limit the AI's final size. The displayed numbers are more of a starting size for the AI and if your army is large enough the numbers you actually face in the battle can be significantly larger. This usually only starts being noticeable on Legendary with max sized units though. Depending on various factors you will often be outnumbered at least 2:1 as the CSA. In many battles it takes time for the AI to actually bring all of those units onto the field though, so your smaller numbers can defeat each wave before the next arrives.

I messed up and the surrender weapon recovery changes didn't make it into 1.2. Will have a new version up soon with that corrected.

In terms of corps restrictions I would probably recommend self imposed restrictions. Could go even further and limit yourself to historical ratios of unit types and such. Coding this kind of restriction for the supply wagon is a bit more work than it might initially appear and this feels like the kind of feature that everyone will want a slightly different version of.

This is an existing issue from the base game. When you have veterans selected the game calculates how much a veteran for that unit would cost based on the unit's experience and charges you for that cost. No comparison is made against whether or not you could achieve the same result with rookies.

Glad you're enjoying it

You can start with 7 training. This gives the union nearly 1 star recruits and the CSA ~1.5 star recruits. Though currently starting with econ and logistics and just buying cannon and fielding 24p and 20p as soon as possible is probably better

The size shown on the intelligence screen gives you the starting point of the AI army. Scaling from various factors can raise and lower the actual number you face. I would probably just look at the prebattle amount. I enabled it for all values of recon specifically so you get some warning before heading into a battle.

Should get 6 units instead of 5 there, will get it changed in the hotfix patch. I'm not sure that matches what I've seen. If only generals or supply units are left it should trigger the no units remaining flags. Did you have recon 4 to see what the remaining strength was? My recommendation for clearing generals is park an arty unit near them and don't chase them around. Don't directly target them as that can cause them to fall back. Spotting changes in the mod are only for the player. Sometimes you can get in a scenario where a unit will flip in and out of cover. If a unit triggers its firing animation that will continue even if you immediately stealth again.

I think this is just really bad luck. At a quick glance the formula doesn't care about unit sizes or cover. If a unit takes damage, roll a dice to see if an officer gets wounded. If they are wounded roll a second dice to see if they are killed. I'll add it to my list to check on what the actual rates are, though fair warning I think Jonny would rather those rates go up than down I've lost almost no officers on skirmishers, I usually just drop captains or majors on 250 man units, ignore any efficiency penalties and let them rank up. I think if you're switching to infantry I would consider it as just holding a forward position rather than delaying like skirmishers would. I've found that infantry have a far harder time disengaging than the skirmishers will and will take more casualties performing the same delaying action. With skirmishers I basically just force the enemy to stay out of column and waste condition on reloading while continually falling back. There are probably a few to many places where it's just more convenient to use maybe half or 2/3 the extra slots on skirmishers and just bring more of other unit types instead though. Feedback on the positives or negatives of this aspect is appreciated. Which side? I mostly updated the low number(<6) of deployment slots based on memory so I'm sure I missed a few. Union Cold Harbor was definitely missed. Getting through that with only 5 units and no skirmishers is likely no fun. Probably have to hold off the point and try and retake it or something. If you completely destroy all enemy units you will always immediately win most battles. There may be some exceptions, but you can skip part 2 of Phillipi for example. Adding or significantly changing objectives is still out of reach unfortunately. It's probably possible but I haven't figured it out yet. 2nd bull run and Antietam have issues with this on both sides. Given that it's only 5k max for a capture I'm not to worried about it given the comparative effort to remove them. Base damage values Ignoring cost, ammo, reload, collateral radius, etc 20p: 110 30-50 = 33 - 55dmg 3in: 70 45-60 = 31.5 - 42dmg nap: 80 42.5-65 = 34 - 52dmg 24p: 100 25-55 = 25 - 55dmg So the 20p will deal more damage per shot and has longer range. However the main difference is how much damage they do at various ranges. I'm working on getting those damage curves available for viewing, but all I can say for now is that the 20p and 24p have performed significantly better at the same roles as the 3in and the Nap for me. The other thing to note is that in general we attempt to present trade offs with the weapons. You should be able to go super heavy siege batteries and eat the extra cost and reload times. But you can also go with light to medium artillery that is cheap, reloads quickly, and still deals decent damage in comparison.

We can do quite a bit more than that. I could rewrite the entire damage algorithm and add diminishing returns to accuracy if I wanted to for example. I think any benefit from doing so is outweighed by the downsides though. Adding new mechanics is tricky because there is no UI support for them and they would largely have to be hardcoded as there is no space in the assets file. There are other limits but I don't have any concise way of describing them currently. Maps, UI elements(other than text), objectives, timers, mission order are the types of things we definitely can't change so far.

Accuracy does not have diminishing returns which is partially why it's causing problems. For Infantry I use a mix. I have some full accuracy units, some full melee, and some lvl1 accuracy/lvl 2 melee units. The full melee units usually back up the full accuracy units since they will get overrun if a movement speed brigade charges them.

Add some heavy cannon to snipe the enemy artillery and it would probably work just fine. I'm using almost nothing but long range artillery and leaving battles with barely any casualties. Requires a 2 star skirmisher to scout though. We made some major breakthroughs in terms of modding the weapon damage on so we should be able to get some better results in the next major version. I think this means that accuracy perks will be able to be toned down without affecting giant units ability to reliably damage each other. Probably will have a hotfix for 1.11 before that with only minor changes, though that's all dependent on real life schedules.

Without spoiling to much, make sure to pay attention to your units performance as you as you progress further and increase their size. There are some hidden mechanics to discover. If winning Antietam seems out of reach, the other tactic you can try is a fighting retreat to inflict losses while still ultimately leaving the battlefield before the majority of reinforcements arrive and you take to much damage. This can be hard to execute correctly though as there are lots of opportunities for things to go wrong.

Cavalry will always be more effective in open terrain. as any kind of cover will drop melee kill rates considerably. A lot of the value of melee cavalry is in their morale shock. They inflict kills very quickly, which drops morale, which in turn causes the melee value to drop further. The faster you break the unit the fewer casualties the cavalry will take. In cover the kill rate gets slowed down enough that they trade kills more evenly, the longer the battle lasts the worse it will go for the cavalry. The melee value of palmettos was reduced in the last patch so you will need to upgrade to better pistol/saber weapons for the same results as before. Most rifles have melee values around 65 though they are some specialized rifles with higher values.

I have to ask, do you mean 25 guns across 3 brigades or all three brigades are at maximum size?

Presumably once we get the mod updated for 1.11 the achievements should start working again. The current plan is to just copy paste the previous version changes in instead of doing a balance pass so we should have it ready in less time than the update to 1.10 took.

I'm going to look into changing the x3 fast forward into around an x8. Will probably be to fast to play on, but it'll help with waiting for timers to run out or early parts of battles where you are just watching a column march somewhere.

Excellent, thanks for writing up the instructions as well. We'll include those and the mac friendly dll in the next version.