Effects of Attack Value

[Rhyfelwyr]

I always thought attack value was how likely a unit was to make a kill when it hit an enemy unit, with a higher attack value making it more likely to succeed. However, I recently read something about a hidden lethality value which is in fact what determines how likely a hit is to kill an enemy unit. If this is the case with M2TW, what exactly does the attack value do?

[IrishArmenian]

I have no idea and have been wanting to know what exactly attack values do, too.

[Zarky]

I don´t know the mechanics but it´s attack value vs. enemys defense value (maybe separate rolls: against defense skill, armor and shield.)
If anyone knows more, feel free to tell

[Agent Smith]

Well, I can tell you how it worked in MTW, but I'm not sure if they carried over the same system to RTW and M2TW.

[Budwise]

Well, I can tell you how it worked in MTW, but I'm not sure if they carried over the same system to RTW and M2TW.

and that is?

[sapi]

I always thought attack value was how likely a unit was to make a kill when it hit an enemy unit, with a higher attack value making it more likely to succeed. However, I recently read something about a hidden lethality value which is in fact what determines how likely a hit is to kill an enemy unit. If this is the case with M2TW, what exactly does the attack value do?

Animations also have a huge effect on kill-rates.

[Miracle]

and that is?

Something like this:

Attacker's Chance to Hit = 1.9% + (Attack - TargetDefense)*20%

Keep in mind that in MTW attack speeds are a lot faster and the CtH is subject to a wide range of modifiers.

Reference:
https://forums.totalwar.org/vb/showp...31&postcount=9

There's a lot of stat discussion in that thread as well.

[econ21]

Something like this:

Attacker's Chance to Hit = 1.9% + (Attack - TargetDefense)*20%.

I've always been vague on the exact formula - there were some old posts went into this. But I would modify the above formula to:

Chance to hit = 1.9% * (1.2)^(Attack - TargetDefence)

where * = multiply and ^ = raise to the power

ie with attack=defence, there is a 1.9% kill chance. With one higher attack, it becomes 1.9% plus 20% of the 1.9% = 2.28.

The mechanics chapter in the MTW strategy guide (written by a CA staffer) gives the following table:


The combat factor is (attack - overall defence) where overall defence depends on defence skill, shield and armour to varying degrees.

Puzz3D reckons the kill chance increments in RTW are smaller than in STW/MTW, so they should rise by about 10% rather than 20%

What I think "lethality" does in RTW and M2TW is turn the "kill chance" into a "hit chance". So whether you hit is based on the combat factor and lethality is the % chance that the hit kills. So a hitting with a spear with a lethality of 0.73 would kill 73% of the time.

Whether M2TW/RTW really works in a similar way to STW/MTW only CA knows, but I think it is a decent working assumption. If it does, we have a pretty good idea of how attack, defence and lethality work. Animations and the swing speed stat seem much harder to factor in and modders seem to have to balance them by trial and error.

[Agent Smith]

and that is?

Econ just posted it. I didn't think it applied to Medieval 2, though. It seems like something different is going on under the surface.

I wish they would give us more detail as to how it works out. Especially because there seems to still be debate as to what the defense values actually are, if they are showing up on the unit cards right, etc.

[alpaca]

Well I believe it's still a function of attack-defense but the "hidden lethality factor" is probably not so hidden but a result of the animations.
In M2TW, we have a 3d collision model between soldiers which means for example that it's very hard for archers with their small daggers to hit mounted units, and at the same time makes a lot of 2-handed units fare worse than you'd expect against other infantry because they are too close to them to hit them effectively with their large weapons sometimes.
This also applies to shields where the defense value is in fact only applied if the defender actually plays the block animation.
So all in all, it's very complicated and devilish to balance.

[Kobal2fr]

Well I believe it's still a function of attack-defense but the "hidden lethality factor" is probably not so hidden but a result of the animations.
In M2TW, we have a 3d collision model between soldiers which means for example that it's very hard for archers with their small daggers to hit mounted units, and at the same time makes a lot of 2-handed units fare worse than you'd expect against other infantry because they are too close to them to hit them effectively with their large weapons sometimes.
This also applies to shields where the defense value is in fact only applied if the defender actually plays the block animation.
So all in all, it's very complicated and devilish to balance.

I don't think the kill rates are based on a 3d collision model. Soldiers die fairly often because someone swung a sword or a spear 5 inches away from them .

In fact, it's a problem with user movies, sometimes you can witness the moviemaker trying to do a close-up on the more stylish moves (like that quick horizontal slash that brutaly turns the enemy around, then lunge through the back) and come out utterly ridiculous because the models don't actually collide at all, or the lunge hitting thin air.

The animations affect attack speed, but there's more - 2 handers are weak against cav, for example, because in melee knights have a superfast left/right/left downward stroke, while most of the 2h strikes involve a long "arming" process, during which the 2H user is not "invincible". But by comparison, a spearman who does the long "shield to the face, move back, then spear the fallen enemy in the back while he's prone" move is not vulnerable during his finisher, as far as I've seen. Same for swords : I've never seen a DFK die while he was doing the aforementionned slash-then-lunge. So long anims are sometimes good, sometimes bad.

Also, I think you have it backwards : the shield is not "only factored in when the model blocks", I rather think the model blocks *because* a hit roll was parried/blocked/missed

[alpaca]

Nope, what you describe is because the collision mechanism uses a much less detailed model than is actually shown on screen (and probably because calculations aren't exactly exact).

Try it out if you want, use a halberd unit against swordsmen and you'll see that a lot of the time instead of actually killing anybody the halbierdiers will try to push back the enemy with the back of their stick.

[Kobal2fr]

Nope, what you describe is because the collision mechanism uses a much less detailed model than is actually shown on screen (and probably because calculations aren't exactly exact).

Try it out if you want, use a halberd unit against swordsmen and you'll see that a lot of the time instead of actually killing anybody the halbierdiers will try to push back the enemy with the back of their stick.

You would have a point but... what coder on Earth would make a collision model extend further than the chrome (that is, the graphical eyecandy extravaganza model) ? Makes absolutely no sense.

I certainly agree with you that the "animations" used in battle calculations must be very simple stick figures and skeletons and so on, and not the complex polygons that make up the models we see, but there's no point in making the skeletons extend further than the polygons strapped onto them. It doesn't make the calculations any easier, and it would make "ghost fighting" even more prevalent than it is right now... No, I think the air swording is a testimony that either a collision model is used and buggy, or that there is no collision at all, just models told to do this and that when attack rolls go this way or that.


By the same token, I think the push back thing is just the animation halberdiers use when their attack roll fails, or the opponent attack rolls fail maybe, the equivalent of the parry/block for shielded units, not a symptom that they need more space for their anims... 2handers are on the whole unconcerned about slashing through the guts of 3 of their comrades to hit the one guy they're targetting, for example . Another example of this is pikemen when you take away their secondary attacks : they poke-poke-poke through enemies 1 inch away, looking like idiots, when your spacing reasoning would have them move back the length of their pike to strike efficiently.

Admitedly, my theory doesn't explain why some animation sets are better than others, or the former 2H bug for that matter... I guess we'll have to organize a meeting between Mr. Lusted and Mrs. Scopolamine in a few months to know the truth

[PutCashIn]

Has anyone witnessed/recorded a soldier dying for what appears to be a totally numbers, non amination based situation? as in, no one was around the soldier to hit him, or its unit was under missile fire but the soldier didnt appear to be 'hit'....?

I could swear I (might) have seen it once...and only once, in a repelled seige attempt, an enemy unit was routing but got chaught in the gate, resulting in single file unit spreading accross half the map, under arrow fire and in close combat in parts, I'm (pretty) sure some dude in the middle fell over simply from 'too much damage' or whatever.

Then again, the unit was under fire from the city's towers, and I could also swear that I've seen ram's catch on fire in between salvos from the towers...?

Other games that use damage per second with a 3D model seem to drop amination accuracy in favor of numbers crunching, TW has units with only 1 HitPoint so the reverse is achievable...although I tire of having to guess weather a soldier in plate (8 armour) with a battle axe (20 attack) is gona get undone by a 3/3 (or whatever) peasent simply because hes big and slow and the peas'y is light and quick.

[atheotes]

I think the game uses collision models to determine hits (as opposed to kills) at some level/extent. I have never seen anyone die from arrows unless they were hit... and not all times the arrows hit do the soldiers die.

Kobal's theory would be easier to implement though it doesnt explain the 2-hander bug and their general ineffectiveness.

[pike master]

just after a cav charge you will see many soldiers fall who are not close enough to have gotten injured.

the front row of pikemen fall regardless if any sword and shield infantry get past the pikes.

[Husar]

I agree mostly with Kobal, the only thing that might have a 3D collision model are missiles(a fast unit that starts moving after a shot was fired can evade it).
And the twohander animation really is a block animation, I have never seen a twohander actually push another guy back. Since 1.2 most twohanders also seem to use only one stabbing animation against cavalry, making them more effective, however, their often really low attack stats are what makes them ineffective since the shield bug was fixed. I increased all twohander and halberd attack stats to numbers between 15 and 24 or so and that makes them quite lethal, giving german zweihanders more defense(that melee-only stat) seemed to show a lot more blocking animations which leads me to think the animations are the result of calculations, which is also the only thing that makes sense. Most onehanded units do in fact have a very high defense, which also explains their effectiveness, for a guy with 10 attack it can take a while until he gets a kill against 21 defense, however, if you give that guy 23 attack...
This was obvious to me when twohanded knights with hammers would always lose against their sword and shield counterparts, if the twohander has only 5 attack to break 21 defense while the sword guy has 13 attack to break the 6 defense of the twohander, the result should be obvious. This was a bit different before the shield bug was fixed and animations do apparently play into attack speed(faster animations = more attacks per minute possible) but if one tweaks the stats accordingly, there is no need to edit any animations IMO.(that's with 1.2 where the twohander animations are fixed).

Let me underline my point with a quote from a CA dev as given in this interesting interview:

The second area would be balancing, and that's purely because we shipped a game with 21 factions and close to 400 units. That's a ludicrous amount of stats to tune, and so more time would translate to better results. It's something we've examined again with Kingdoms.

[Kobal2fr]

I agree mostly with Kobal, the only thing that might have a 3D collision model are missiles(a fast unit that starts moving after a shot was fired can evade it).

Hollywood deception . Or : what the eyes see is not what the numbers crucnh. The "to hit" roll is made first, factoring the fact that the target is moving or static, tight or loose etc..., then if Archer A got a hit, his arrow will be displayed as striking true, no matter what the target does in the meantime.

It's a pretty easy thing to ascertain : if a missile unit is ordered to shoot at another unit that's on the very edge of its range, and said target moves away before the shooter has finished his animation, the animation will go through and shoot, and hit/miss, even though the target might well be very much out of range by then (it's even easier to see with javelins shooting at cavalry). Same goes for targets who run behind walls/hills between the start of an missile unit's cycle and the time they actually launch the arrows/bolts/bullets/javelins/tallywhackers.

[Didz]

It's a pretty easy thing to ascertain : if a missile unit is ordered to shoot at another unit that's on the very edge of its range, and said target moves away before the shooter has finished his animation, the animation will go through and shoot, and hit/miss, even though the target might well be very much out of range by then (it's even easier to see with javelins shooting at cavalry). Same goes for targets who run behind walls/hills between the start of an missile unit's cycle and the time they actually launch the arrows/bolts/bullets/javelins/tallywhackers.

I can see the logic of what you are saying, but I'm not convinced its entirely true.

The reason I have doubts is quite simply the fact that collateral damage is done by missile fire.

Logic would suggest that if hits were determined in advance of the firing animation then they would be calculated on the selected target unit and consequently the missiles that hit would fly straight and true to those men from the target unit which were pre-determined to be hit.

But they don't....

If you happen to be pursuing that unit frequently you see your own men go down because the target unit moved out of the field of fire and your cavalry ran straight into it.

Likewise, frequently you see men falling in other units alongside or behind the target unit which one would not expect to have been included.

Indeed with musketry fire, certainly in STW bridge battles, it was possible to see men drop in units hundreds of yards beyond the target unit, who are presumably being hit by spent balls fired at the unit crossing the bridge.

So, if what you say is true, the calculation of hits from missile fire would have to be pretty sophisticated to work out the odds of collateral damage to units who at the time of firing may not actually be in the field of fire but are predicted to move into it during the flight of the missile.

A classic test for instance would be to target a cannon at a unit and then just before the cannon fires move a friendly unit in front of it. If the hits were determined prior to the cannon firing then the ball ought to fly straight through the intervening unit and hit the target it was destined to hit. On the other hand if the ball ploughs through your hapless test unit it must be doing so because of collision detection as it could not possibly have predicted that you were going to be dumb enough to move that unit prior to you actually moving it.

Again in STW it was actually possible for archers to shoot the archer in front of them in the back if you got the unit badly deployed on a reverse slope, which seems to be a very unlikely scenario if the hits were being precalculated.

As far as melee combat is concerned I'm less certain.

Personally, I watch a lot of close up melee combat as I like looking for the 'funny's' that sometimes occur like the guy getting kicked in the balls.

So, far I've never noticed anyone go down with no logical reason for doing so. What I have noticed is that frequently the blow that kills them doesn't seem to connect perfectly, but I just put this down to the fact that every character in the game must have a collision zone which extends like an aura around the figure itself and that this coupled with animation lag sometimes results in the animation being triggered and not completing before the figures become seperated. Thus a sword strike appears to hit thin air but the target still goes down, that doesn't necessaily mean that the blow was not triggered by a collision, merely that the animation failed to keep pace with the result.

[Husar]

Well Didz, there is a certain collision detection working of course, but once a decision for an attack is made, the result will be calculated and then the animation will be played. That does not mean that the animation decides whether an attack is calculated or not. Also noone attacks an anemy who is currently doing a finishing move because he is excluded from the routine which determines who to attack next, once he is finished, they will start attacking him. To test this further, giving someone a defense value of 30 might be a good idea to see how often they will defend against a blow, might test that now.

[Kobal2fr]

Yup, I thought about that during a boring (but missile heavy) battle. I think you have noticed as well as I did that archers ordered to fire at a unit behind another unit will often do much more damage to the front unit than to the back one. But if you order them to fire at the front one, they'll do less damage to them than when they were firing at the unit behind them.

So there's probably some form of collision going on after the shot is made, you're right. Just, not between a missile unit and its target, in this case I believe it's all attack rolls.

Or maybe a check is made at the location of a failed arrow's landing point ? In fact, I had dreamed up the exact same test you did : order crossbows to fire, and move knights right in front of them before they actually fire. If the knights get shot to pieces, then there's collision.

Oh, and STW/MTW is not the same engine at all, so what applied may not be valid still.


EDIT : oh, and as to the original issue, I just did a test the results of which... puzzled me. I used billmen (whose animation set is "known" to suck) vs. Noble Swordsmen (the "overpowered" DFK anim). I first gave the billmen enough armor to have the same overall defense as the swords, but kept their attack/charge value which is much higher, even without factoring AP in. The upped billmen promptly trounced the knights, who were reduced to 45 on the charge. During the protracted melee, while the knights attacked on the whole more often, the billmen very often just took the hits, flinched, and attacked right back.
But I figured the result was flawed by the charge. So I did another test, this time putting their armor back to 0, and upping defense all the way to "Noble Swordsmen Total Defense". This time I let the knights charge and the billmen brace. They took a heavy hit on the charge, but then evened the score during the melee, and beat the knights in the end (higher attack and AP, remember ?). Again, the swordsmen attacked more often on the whole, but this time most of their attacks were parried/deflected by the billmen using the butt of their weapon to do so.

OK, I know two tests alone are not conclusive but... I wonder if this whole "some animations are better than others" idea isn't plain ol' wrong. I know it's generally accepted as being true, but it also dates back to the days when the shield bug roamed unknown... Hmmm...Mebbe it's just the big discrepancy in defense values that make 2handers suck this bad in long melees...

[phonicsmonkey]

On the other hand if the ball ploughs through your hapless test unit it must be doing so because of collision detection as it could not possibly have predicted that you were going to be dumb enough to move that unit prior to you actually moving it.

I am dumb enough. I once charged three units of hospitallers through a line of my own culverins just as they were firing - OUCH

I have also charged my general at an enemy unit just as three volleys of my own javelins were landing - bye bye general

not sure if there's any explanation for friendly fire unless there's a collision model at work for missiles...of course this proves nothing for melee

has anyone ever seen a soldier kill one of his own with a misplaced pike jab or reckless sword swing? I haven't...again however this proves nothing, as collision might only be possible, in melee, with hostiles

[Mad Mac]

[QUOTE][OK, I know two tests alone are not conclusive but... I wonder if this whole "some animations are better than others" idea isn't plain ol' wrong. I know it's generally accepted as being true, but it also dates back to the days when the shield bug roamed unknown... Hmmm...Mebbe it's just the big discrepancy in defense values that make 2handers suck this bad in long melees.../QUOTE]

Those are some veeeeery interesting test results. Especially the idea that blocking animations are possibly tied to defense actually kicking in and parrying attacks. And it does make a fair bit of sense, as even a good two hander typically has less overall defense than a basic armoured sarge with no upgrades. Units like that are going to drop like flies.

Still, two-handers are known to improve in peformance with faster animations, so I think attack speed is going to be an issue, regardless.

[FactionHeir]

Two things that I found peculiar with current animations / hidden rolls.

1. Note that if an AI unit routs on the walls, you can have the entire wall packed with fast swordsmen and hardly any of them will ever hit the routers who run through them.

2. When you walk a unit slowly in one direction and enemy routers run in the same direction against your unit, many of them die even though none of your units are facing them or even making any attempt at an attack animation. They just bump into the back of your unit and perish.

Those two need to be explained before a conclusion can be drawn on animation vs rolls.

[Kobal2fr]

Mwaaahahaha. I love prodding ignorance's fat buttocks. I did another round of tests with billmen vs. noble swords (just in case you wondered why those two : because England vs Scotland is the default custom battle).

This time, I gave the billmen exactly the same attack values as swords, and also exactly the same defense (including shield value, on a lark, wanted to know wether the unit model mattered with it. It doesn't). Fired a couple battles, the billmen got owned. Severely.

I was about to mumble the usual "aaah well, back to the drawing board, swords really *are* overpowered by nature", when I realized I had missed something : their collision mass wasn't equal. Oooh, and I had missed delay and "skeletal compensation factor", whatever this is. And heat values. So, not wanting to completely bork my EDU, I copy/pasted the swordsmen entry, renamed it billmen, gave them billmen models and anims, and commented the original billmen out of the file. And this time around, lo and behold, both units were *exactly* the same in the field. Did 8 tests. Each battle took a *long* time to complete (hey, high morale, high armor, no ap, weakish attack...), there were a lot of parries, dodges and whatnot, but in the end both units were down to single digits when one or the other eventually ran.


So... in short : neener neener ! I was right ! Animations are not really factored in combat ! It's all pure, mathematical and statistical goodness using factors the effect or meaning of which I have ABSOLUTELY no idea and didn't even know existed .. The numbers control the animations, not the other way round !
So, who's buying me the first beer ?

[Mad Mac]

So swords are superior because of their superior bone structure, heat tolerance, and pants budget? You know, it was a lot easier to just blame the animations...

[Husar]

So, who's buying me the first beer ?

I've been agreeing with you but take that beer anyway.

[Didz]

Well I've been paying even more attention to close combat since reading the earlier posts, particularly during pursuits and I've noticed a number of things.

1) Units die without being subject to a hit animation. The most common is the instance where a horseman bumps into a routing unit and one of the routers just drops dead on the spot. There was still a collision but the attacker sometimes didn't go through any animation.

2) Routing units frequently run right through a formed unit whilst only taking a few casualties.

3) I've still not seen anyone die for no apparent reason, there is always at least a bump, even though no sword is swung.

4) I've still not seen anything that looked like a 'friendly fire' event in close combat. Two units of horse archers ordered to pursue the same routed unit sometimes inflict casualties on each other but that is almost certainly due to stray arrows due to their continued use of the bow even when in melee combat.

[econ21]

I was about to mumble the usual "aaah well, back to the drawing board, swords really *are* overpowered by nature", when I realized I had missed something : their collision mass wasn't equal. Oooh, and I had missed delay and "skeletal compensation factor", whatever this is.

I don't suppose we could prevail upon you to test which of those "forgotten" factors was the culprit? Is "delay" the "swing speed" factor? If so, my money is on that as the villain.

But one thing I don't get is why 2Hers animations were thought to be the source of their weakness in the infamous 2Her bug? And why switching them to, say, JHI animations, fixed it? How can we reconcile that with your "animations don't matter" conclusion?

[Kobal2fr]

@Didz : "horseman bumps into foot soldier, footman instantly keels over" happens all the time when horses charge - I call this the Aura of Death. Lots of static electricity in those platemails . It's *very* noticeable in this video (sorry Foz if I'm ruining your bandwidth )

I believe that what happens is that units that are considered to be "charging" don't need their hits to be followed by an animation at all, which kinda makes sense. If it wasn't the case, then a charging cav unit would only kill 20 men tops on the charge (considering a 40-strong unit charging in two rows) instead of ploughing through infantry as they're supposed to. It also happens with charging inf, but since those have a collision mass roughly equivalent to the other inf they're charging, the charge stops as the first ranks clash (and since their charge bonus is weaker, they accrue less killing hits, so the effect is less noticeable).

As to point 2), maybe units are counted as perma-charging when hitting routers, which would explain why routers fall in one brush to cav but can take several hits from lowly (but fast) peasants ?

3) and 4) agreed. In fact, I've seen a lot of melee fighters happily swinging right through their comrades without hurting them. Cue humorous screenshot with charging knights sticking their lances into each other.

I don't suppose we could prevail upon you to test which of those "forgotten" factors was the culprit? Is "delay" the "swing speed" factor? If so, my money is on that as the villain.

Well, delay is commented as being "minimum number of seconds between attacks", but I rechecked and it's 25 tenth of a second for every unit. So it's all in mass (which only affects charges AFAIK) and skeletal thingamajig.

The problem with testing it further is that while I can ascertain wether making it higher or lower is better, it still wouldn't tell me *why* it is so. Or why the hell anyone would need to compensate for his bloody skeleton in the middle of a goddamn battle .
But I will test it some more, yes. I hate stuff I can't figure out.

But one thing I don't get is why 2Hers animations were thought to be the source of their weakness in the infamous 2Her bug? And why switching them to, say, JHI animations, fixed it? How can we reconcile that with your "animations don't matter" conclusion?

I've thought about this, and this is the scenario I came up with :

Billmen and Axemen never tried to hit cavalry units because...

Billmen lacked an animation for "attack someone up there". Say a billman attacks a horseman, does his attack roll, hits, kills. The game engine takes this info in, blows its whistle, turns to the fighters and says : "OK, you swing like *this*, and you, you fall down like *that* when he's n milliseconds into his animation. Make it look good people, we're only doing one take". Only because of the missing "swing like *this*" anim, the horseman was forever waiting for his cue to fall down.

Soldiers are probably not counted as "dead" until they've finished their dying anim, that much can be deduced from a simple missile volley : archers shoot, hit, enemy goes "aaargh, they got me ! tell my wife...", sag, lie down, and only after they've been down for a little time will the unit count actually go down and the dead man's green/red circle will vanish.
So, cav didn't die because they were waiting to start their disabling death anim. In the meantime they weren't dead and could attack.

The JHI animation set didn't lack an anim for "attacking something higher up", so didn't have this problem. 2H swords didn't either, IIRC, only 2H axes (which billmen share).

Sounds plausible ?