Archer Accuracy Experiment

[Old Geezer]

I have heard and believed that archers do not gain accuracy from experience. Having run some tests I am convinced that increased experience does result in more kills.

I tested levels 0, 1, 4, and 9 longbowmen against highland pikemen on very hard setting on the Spanish plains (only clear weather was used as rain made the attacking longbows do strange things). The pikemen were defenders and controlled by me. The AI had the longbowmen. For the analysis I used only the first 5 volleys, because after that the AI advanced to point blank range at different times. I used at least 3 different custom battle loads for each level tested (except the level 1 which was one custom battle run 10 times) and ran 12 tests for the level 9 and level 4 longbowmen, 10 trials for the level 1, and 13 for the 0 level.

The level 0 left an average of 64 pikemen standing.
The level 1 left an average of 61.5 pikemen standing.
The level 4 left an average of 59 pikemen standing.
The level 9 left an average of 54 pikemen standing.

These results have convinced me that level of experience does count. There is enough spread that this might not be noticeable during game play. The spread for the level 0 longbowmen ranged from 67 to 60; the level 1 ranged from56 to 65; the level 4 ranged from 62 to 52; and the level 9 ranged from 61 to 47. So during any particular battle you could have all three levels getting close enough results that you might, as I did, assume that experience doesn't help accuracy.

I don't think that the level 9 has any better accuracy than the level 7, but I have not tested it yet.

I tried testing ballista but had problems because they will advance to point blank range in fearless disregard of my defending peasants (probably due to securerom ).

So, I advise not taking level 0 archers up against the Mongols level 6 archers in a shoot-out.

[Ratwar]

Very interesting, though I'm gonna have to disagree with you about level 9 having the same as 7. From level 1 to 4, there is a 2.5 increase, so if it is a linear function, it would make sense for level 7 to have the same 2.5-3 difference, but level 9 is roughly twice that.

Still, although interesting, looks almost useless. A difference on average of 2.5 after 5 volleys against something with no armor is small enough already, but if you take armored units, the difference will be even smaller. Plus, the standard deviation appears to be quite large. The level 9 longbows were actually worse than one group of level 0...

[rossahh]

Do archers get the same damage bonus (for their arrows) with experience like the melee infantry get with their weapons?

[Dead Guy]

Run a repeated measures ANOVA. Your null hypothesis would be that archers of varying experience don't differ in the amount of men they kill with their ranged attack under the same conditions, I presume. Then, if p<0.05 there is a statistically significant difference between the groups of archers and experience probably effects the accuracy of archers :p But there is always the chance that you made a type 1 error, the difference you detect is a coinceidence, and the null hypothesis is actually true.

I hate statistics... =)

[Old Geezer]

I just ran 10 trials with level 7 longbowmen and had an average of 56.1 men standing after the 5 volleys. So, it looks like every chevron may give some benefit, probably at least 1.

[hellenes]

This is one of the bizzare no brainers from CA....WHY REMOVE ARCHER EXPERIENCE BONUS? The more guys this men kill the better they become....
A big step back from RTW....

[ataribaby]

I reckon the experience bonus for missile units should benefit their Missile Attack attribute; e.g. for every three chevrons you get +1 Missile Attack value. As it is you've got a small boost to melee attack and the usual boost to morale.

The melee boost is usually neither here nor there as you don't deploy archers to fight other infantry hand to hand. The exceptions are dual-role archer/melee units such as Ottoman Infantry or Norse Archers who are designed for a bit of hand-to-hand action.

The morale boost is good as most archers have a morale of 3 - 2 below the average - so any gain helps in the unit not giving up immediately on contact with enemy cavalry until help can arrive from other infantry.

But as archers are there as "a woman's strategy, to throw things at us" surely it is their throwing ability that should get the experience boost.

p.s. I'm still pretty sceptical about the idea that accuracy is improved from the fact that accuracy is coded to be tied in with projectile-type.

[Rhyfelwyr]

Well I think there's definitely something in those results. I couldn't just be a coincidence could it?

[ataribaby]

Weeell... ah ok. It looks pretty convincing, and it would be really nice if it were the case; it makes that Woodsmens' Guild HQ all the more worthwhile.

[PBI]

I suppose it does make sense though that experience affects accuracy but not missile damage; if the archers are too far away to aim for specific body parts or weak points, then the damage would depend only upon the type of missile and its velocity, neither of which are improved by experience, but the chance to hit the target in the first place would improve with the skill of the archer.

Still, good to see experience does seem to have some meaningful effect for archers; I'd actually taken to not expending all my arrows on the units in the town center towards the end of a siege, and instead sending in the heavy infantry since I assumed the experience would benefit them more. I wonder if this is different for siege weapons, since I believe it was established that experience made no difference for them?

Also, it's still strange that weapon upgrades only affect melee abilities; they really should improve missile damage.

EDIT: I'm pretty sure accuracy was tied to projectile type in MTW also, and experience very definitely did affect accuracy there.

[Old Geezer]

One very interesting thing which I noticed on the last series of trials which for some strange reason had 6 rainy days was that the archers shot one to three volleys and then advanced to point blank range as soon as the rain got visibly heavy. They invariably had misreable results and got only a couple of hits if any in the heavy rain at extreme range. I strongly suspect that the talk that I have heard on the forums about weather not affecting accuracy is also wrong. This will be hard to establish because the intensity of the rain is hard to guage and the archers advance at different times depending upon factors which could be one or several. i.e., the number of hits in a volley, total hits, intensity of the rain, the forecast, and who knows what else. Why can't CA just tell us these things? Do the designers, developers, and playtesters ever communicate with one another? I guess they don't have a common language. I didn't mean to start a rant. I need to save that for "The Final Patch".

Really though, M2TW has a lot of staying power. I am still learning so much basic stuff that I'll not be ready for Empires for a couple of years.

Can anyone think of a way to test artillery so that we can get some reliable trials?

[PBI]

Can anyone think of a way to test artillery so that we can get some reliable trials?

I have an idea.

The problem is that to test artillery against units, it's impossible to get the units to stand still long enough or consistently enough to get off enough shots at roughly similar range. Perhaps we could mod the unit's movement speed down to zero for the purposes of the test? Alternatively we could try modding the terrain movement speed multiplier to achieve the same thing.

Similarly with the weather, it might be possible to mod the probabilities of rain etc for different climates, in order to create one with guaranteed constant rain of a certain volume.

[Old Geezer]

Great Idea! Mod the movement - to .0?

[Ramses II CP]

Or you could just play multiplayer and have one player stand still for it.

[Ratwar]

With Artillery, I'm gonna have to suggest that instead of measuring kills per volley, measure volleys till hit.

[Rhyfelwyr]

One very interesting thing which I noticed on the last series of trials which for some strange reason had 6 rainy days was that the archers shot one to three volleys and then advanced to point blank range as soon as the rain got visibly heavy. They invariably had misreable results and got only a couple of hits if any in the heavy rain at extreme range. I strongly suspect that the talk that I have heard on the forums about weather not affecting accuracy is also wrong. This will be hard to establish because the intensity of the rain is hard to guage and the archers advance at different times depending upon factors which could be one or several. i.e., the number of hits in a volley, total hits, intensity of the rain, the forecast, and who knows what else. Why can't CA just tell us these things? Do the designers, developers, and playtesters ever communicate with one another? I guess they don't have a common language. I didn't mean to start a rant. I need to save that for "The Final Patch".

Very interesting. Why do CA undersell themselves, couldn't some PR guy just stick a post here and there to tell us about all these less than obvious features?

[ataribaby]

They invariably had misreable results and got only a couple of hits if any in the heavy rain at extreme range. I strongly suspect that the talk that I have heard on the forums about weather not affecting accuracy is also wrong.

I can definitely say that rain (heavy especially) has an adverse effect on archers. They seem to fire as many arrows at the enemy but more of the arrows seem to just disappear into the enemy unit without causing injury.

There's always been a hint of archery + rain = nerf in the unit descriptions, especially where composite bows are concerned (they're meant to disintegrate when the rain dissolves the glue aren't they?).

But it's only recently I've noticed the effects of weather when experimenting by choosing rainy battles. Sieges are definitely easier as fire arrows are less able to burn rams and towers. Archers firing flaming arrows seem to have even less effect in heavy rain (so it's good to siege assault at night in a storm when the AI has flaming arrows switched on). I've also noticed a distinct effect on visibility; the AI tends not to engage your troops until you get very close, which includes archers as their range is effectively reduced.

[Old Geezer]

To test artillery accuracy would it work to set up a custom battle where I assault a castle gate with ballista which have various experience levels? (I am not able to run any trials for a while.) Some gates on the larger cities or fortresses seem to be very hard to hit from a distance and I have had to move closer. I think this is because the gates are recessed so far that they present a smaller target.

[ataribaby]

To test artillery accuracy would it work to set up a custom battle where I assault a castle gate with ballista which have various experience levels? (I am not able to run any trials for a while.) Some gates on the larger cities or fortresses seem to be very hard to hit from a distance and I have had to move closer. I think this is because the gates are recessed so far that they present a smaller target.

I tried a test against towers before:

Spoiler Alert, click show to read: 

I was intrigued to find this out once and for all because I've had a tendency to hold on to costly-upkeep, experienced arty because of a superstitious belief that the hard-earned chevrons make them indispensible sniper units.

The test: England vs Mongols in a castle siege battle.

England are attacking with 4 units of ballistae - 2 triple-gold and the other 2 raw recruits; and a unit of town millitia as a general's unit.

Mongolia is defending with a unit of elephants. Because they are defending with one unit the AI will not move from the castle square unless physically provoked, leaving my arty to their dirty business.

Each arty unit is set the same distance away from a tower and given the order to fire. As ballistae only cause a couple of percent damage per hit to buildings it is easy to see the hit rate for each unit.

The result: each unit destroyed its target tower at very much the same rate as the others. Sometimes one unit would be in the lead by a couple of percent but there really was no difference in hit rate between experienced and non-experienced units.

The corollary: arty units don't gain accuracy with experience and can be disbanded when not needed without worrying. After all, they are expensive to maintain.


Postscript: The above battle ended when I had the ballistae fire their remaining bolts at the elephants, and when the elephants stampeded I sent the ballista crews chasing them with their butter knives in a bizarre, Benny-Hill style chase. Needless to say, the greenhorns would rout after one stomp, but the triple-gold guys were very persistent indeed.

Someone suggested that accuracy vs. troops is different for arty units, but I never managed to contrive a test battle where the enemy would stand still and get skewered under controlled conditions. Your strategy of being the defensive/skewered faction might just work though.

[ataribaby]

I just had a little mess around with a unit of ballistae vs. a unit of my own armoured sergeants; unit sizes on large; weather clear; morning; VH difficulty.

On Grassy Plain the enemy would always advance to point-blank after they'd reduced my unit to around 72 men. Those AI fellas love a bit of point-blank, gangland execution style huh?

So I thought by having a forded river in between sides, the ballistae wouldn't be able to advance. I used the River Jordan map and that worked well; the enemy stayed put throughout and allowed my lot to advance into range exhausting all of their ammo.

There is the question of how many ranks deep you deploy the defending unit, as projectiles always hit the central files and sometimes take out a whole file of men at a time. This means that early shots have more chance of greater damage as later shots are only fired at a single row of men. I suppose this effect would average itself out though across many tests.

Secondly, you have an effect where the defending unit creeps backwards gradually over time so you have to watch that you start off well within the ballistae range to avoid intervening by moving your men forward mid-test.

[Old Geezer]

Good suggestion for trials, Ataribaby!

Several days ago I tried running trials on huge city gates and on level 0 ballista it took between about 54 to 74 on three to knock down the gates. I then tried a level 0 and it took 94. I gave up.

With the human player controlling infantry and the AI the artillery, I think it would be best to set the men in 2 ranks and stop counting after somewhere around 10 volleys. (If nothing else you can run trials a lot faster.)