Elven Naked melee testing.

Having been quite frustrated with the seeming weakness of my melee units in PVP and even in recent PVE (given that I'm just now facing light-side armies for the first time)

I've done 3 PVPs for a quick test of naked elves vs. darkies
a total of:
4 HI 1v1s against orcs
2 HI 1v1s against DL
2 LI 1v1s against orcs
2 LI 2v2s against DL

and various and sundry other units which are, quite frankly, of little consequence to me, as I don't favor those units. (and/or as they were archer fights, which I knew I would win)

And my findings are:
the orc HI creamed my HI all 4 times.
levels: mine were level 6, his level 5/6

the DL HI was almost an even match, but my HI lost both matches.
levels: mine were level 6, his level 5/6

the orc LI was eventually overwhelmed, but the advantage was only slight.
levels: mine were level 7/8, his were 9

the DL LI was quickly overwhelmed, not a proper "creamed" but pretty quick.
levels: mine were level 7, his were level 7

Things to note:

my AP cost is lower; with 6 units, LI/LIA/HI, I have to have +40/+80 gear (kept on a non-included unit, for testing purposes) to match AP.

my researches are only +9 melee and +9 armor, the DL's researches are +12 each; the orc's researches are currently unknown. (Devs, do researches cost AP? we need to know that)

the third dark race, undead, is *currently* untested, and will remain so until tomorrow or later.

Results:

*these are based upon a limited sample of testing, so bear in mind that experiences will vary.*

Elven LI is pretty good, probably (from my limited experience) the best LI in the game; it's shield bonus is great. Given almost as much AP cost, my LI will fight just about on par with HI.

Elven HI cannot win fights; at least against Orcs. Or, to phrase that another way, orc HI is super-strong; how can elven HI have an advantage in the simulations? (devs- this one's for you...) did the simulations ignore the specials? or was this just some REALLY bad luck combined with him having +18 researches (which I doubt... given his fame level) the DL HI would *probably* have lost if the AP was equal, HI vs. HI.

Further evidence:

having faced elves for the first time, my PVE experience against my own race is that the HI are the squishiest I've ever put an arrow into; they're closer to the other race's LI than they are to the other race's HI in terms of dying from archer-fire.

I've consistently noticed that my LI is good; however being a person who favors HI/archer combos, I've never really liked having good LI over good HI.

Lastly, a big thank you to Bobba and Scaren, and if you have any further notes/areas of disagreements with my explanations/conclusions; feel free to comment. AND would a willing undead (kuba? but your units are TOO high... mdaborn maybe?) please stand up for the final comparison? but not until tomorrow.

Sincerely, Realf Lantow
PS if any other elves would like to re-peat any portion of this test, for their own benefit or the benefit of the community, feel free to do so; especially high-level players like faerrolon and Dakota. The more user-generated data we gather, the more we can put paid to any rumors about who or what is weaker and stronger, respectively.

One thing I meant to ask the devs as a follow-up... and let me dig up the quote from Runeslayer.

From: http://www.battleconforum.com/t587p75-new-stat-caps-poll#3530

RuneSlayer wrote:

Realf Lantow wrote:

Elven HI cannot face any other HI in PVP battle.

I have to disagree...In fact, Elves HI has around 0.5%-1% higher victory rate in 100.000 battles (simulation with both starting stats and capped stats) against ALL enemy races. Check your stats and gear up your HIs with items which will provide a better stat spread.

It is hard to argue that the stats are not balanced if they come out to be within 1% of 50/50 winrates.

Realf, I think you may just be proving that players should have a high expectation of randomness affecting them in their games. Nonetheless, this is definitely good feedback to give the devs. It shows how much variance players might get in any relatively small number of battles - which is probably about the number of battles anyone would do in an evening. Thats an important scale to examine balance at.

One slight possibility -- and I did not interpret Rune's post in this way, but it may be worth double checking for clarification. Runeslayer mentioned "against ALL enemy races" when he mentioned the +.5-1% winrate. Is there any chance he meant that the +1% higher victory rate is an average of ALL the victory rates per race?

i.e. Using hypothetical numbers: Elves V Orcs = 45%winrate, but Elves V Humans = 56% winrate. If you averaged them, it comes out to about +1%.

I got the impression that Elf Vs X race had a 100k winrate of between 50.5% and 51%. If he meant the average was between 50.5% and 51%, then the per-race winrates could vary quite a bit.

Runeslayer, any clarification on which you meant when you said this?

--
Even if that is true though, I'm kinda okay with some amount of variation in those 100k winrates for lore reasons. Elves having good light units vs good heavy units? That kinda fits my own personal sense of lore for Elves - not that that is what should dictate balance in the game - but I'm okay with some variation due to that. I'm much less okay with too much variation due to chance.

No problem Realflantow, by the way, Scaren also has both +12/+12 researches, just so you know. =)

And yeah, I'm sure randomness came into play a lot here. It's really hard to say anything is conclusive until even more controlled testing could be done with even more tests and carefully controlled stat bonuses (say 20-40 or even a lot more tests), which in my opinion is not really worth the effort. From all the PvPs I've been in, elves have been quite tough, so I don't know if I agree with the complaints people have about elves being too weak. since AP is generally lower than DL too, it should balance out ok overall. As for undead, they still may be a bit too strong compared to every other race, not just elves, but I don't know. I'd have to PvP them some times, or talk to Kuba or other undead players to ask about their recent PvP experiences since the shiver fix. And as for orc vs elf, I'd say if both sides take advantage of their options, they can have a fair match (just watch out for archers and kiting).

I think overall with DL->Elf->Orc combos it balances out pretty well. I have less experience pvping humans, dwarves, and undead, but from the couple of matches I've had with Nethack DL vs Dwarves seems pretty balanced (I was owned, but I believe it was more because of gear and strategy, not because of bad unit balancing).

If anyone disagrees with anything I've said feel free to let me know. This is all theory based off of very limited observations on a very random mechanics engine. So of course I will be wrong about some (or a lot) of this stuff.

One other thing I thought to ask Runeslayer -- when you did those 100k simulations, did you run numbers for control simulations (Same race vs Same race). This would be a great way to get error margins for the simulations. I would expect them to be exceedingly small, but its a good number to double check. It will put some additional perspective on the other numbers.

Realf,

1) Like Hegorn says, the random aspect is probably important in this game. So you'll need more fights to counter that. Your number of tests is pretty low to draw some conclusion...

2) You should also do it with your other units and post the results here (archers / cavs...). We can't really talk balance if you exclude your best troops from the test. It's a little like me testing my archers 1 vs 1 if you know what I mean and not including infantry...

That being said the idea is interesting so I hope you'll continue. But we really need way bigger numbers.

Good post Realf Lantow.

Realf Lantow wrote:

(Devs, do researches cost AP? we need to know that)

Yes, they do and they are included in the unit's AP value.

Realf Lantow wrote:

Elven HI cannot win fights; at least against Orcs. Or, to phrase that another way, orc HI is super-strong; how can elven HI have an advantage in the simulations? (devs- this one's for you...) did the simulations ignore the specials? or was this just some REALLY bad luck combined with him having +18 researches (which I doubt... given his fame level) the DL HI would *probably* have lost if the AP was equal, HI vs. HI.

Depending on who charged first, the outcome may slightly vary. This is the result after 100.000 battles with capped stats. (Victory ratio)

Elves Vs Orcs 51/49 (Elves charge first) 49/51 (Orcs charge first)

As you can see, the chances of losing/winning an engagement head on is almost 50-50.

Therefore, depending on who charged first, the tech bonuses and of course the "dice", the outcome may vary.

Hegorn wrote:

Realf, I think you may just be proving that players should have a high expectation of randomness affecting them in their games. Nonetheless, this is definitely good feedback to give the devs. It shows how much variance players might get in any relatively small number of battles - which is probably about the number of battles anyone would do in an evening. Thats an important scale to examine balance at.

Again, too many variables can determine an engagement. One HI might have "invested" in Armor while the opponent in Melee. As in an engagement you have too many stats working together to determine the outcome of each round (every tick) you cannot just judge from a small number of engagements. Who charges first? Techs? Dice? You need to conduct a lot of tests in order to exclude the luck factor from the "equation".

Hegorn wrote:

One slight possibility -- and I did not interpret Rune's post in this way, but it may be worth double checking for clarification. Runeslayer mentioned "against ALL enemy races" when he mentioned the +.5-1% winrate. Is there any chance he meant that the +1% higher victory rate is an average of ALL the victory rates per race? i.e. Elves V Orcs = 45%winrate, but Elves V Humans = 56% winrate. If you averaged them, it comes out to about +1%?

No...no...no...

Hegorn wrote:

One other thing I thought of -- Runeslayer, when you did those 100k simulations, did you run numbers for control simulations (Same race vs Same race). This would be a great way to get error margins for the simulations. I would expect them to be exceedingly small, but its a good number to double check. It'll put some additional perspective on the other numbers.

Indeed. No problem there.

One last thing gentlemen. Synergies...tactics.... This is not Mortal Kombat 1Vs1... Never forget that.

RuneSlayer wrote:

One last thing gentlemen. Synergies...tactics.... This is not Mortal Kombat 1Vs1... Never forget that.

I believe this quote will be added in my signature ahahah  

Runeslayer, just out of curiosity could you tell us approximatly how many dice rolls there are in a complete combat round (one unit hitting and the opponent hitting back) ?

Nothing fancy with flanking / rearing, just two units hitting each other "normally". It's just to have a vague idea.

Nethack wrote:

Runeslayer, just out of curiosity could you tell us approximatly how many dice rolls there are in a complete combat round (one unit hitting and the opponent hitting back) ?

Nothing fancy with flanking / rearing, just two units hitting each other "normally". It's just to have a vague idea.

1) Check to Hit
2) Check to Damage
3) Check to Save (Armor)

The lower the Reaction skill of a unit, the faster it will repeat the cycle.

Max 5 troops from each unit are attacking, unless it is being attacked by several sides where troops will rush to defend the sides as well. However, no more than 5 troops will fight per side of the unit. Obviously, if a unit has less than 5 troops it will only attack that many times as the number of troops left.

RuneSlayer wrote:

Hegorn wrote:

One slight possibility -- and I did not interpret Rune's post in this way, but it may be worth double checking for clarification. Runeslayer mentioned "against ALL enemy races" when he mentioned the +.5-1% winrate. Is there any chance he meant that the +1% higher victory rate is an average of ALL the victory rates per race? i.e. Elves V Orcs = 45%winrate, but Elves V Humans = 56% winrate. If you averaged them, it comes out to about +1%?

No...no...no...

Heh. Thanks for the clear answer on the winrate question. I figured that was the case, but I wanted to make sure.

RuneSlayer wrote:

Hegorn wrote:

Realf, I think you may just be proving that players should have a high expectation of randomness affecting them in their games. Nonetheless, this is definitely good feedback to give the devs. It shows how much variance players might get in any relatively small number of battles - which is probably about the number of battles anyone would do in an evening. Thats an important scale to examine balance at.

Again, too many variables can determine an engagement. One HI might have "invested" in Armor while the opponent in Melee. As in an engagement you have too many stats working together to determine the outcome of each round (every tick) you cannot just judge from a small number of engagements. Who charges first? Techs? Dice? You need to conduct a lot of tests in order to exclude the luck factor from the "equation".

Hegorn wrote:

One other thing I thought of -- Runeslayer, when you did those 100k simulations, did you run numbers for control simulations (Same race vs Same race). This would be a great way to get error margins for the simulations. I would expect them to be exceedingly small, but its a good number to double check. It'll put some additional perspective on the other numbers.

Indeed. No problem there.

I understand a lot is going on in the combat formulas. I'm asking about things like the sigma (std deviation) values for the 100k battles where you fought the same unit vs the same unit (same stats, same everything).

Further, if you were to do the same test with the same units, but only 30 times, what would the margin of error be? How big is the 95% confidence interval when fighting two of the same unit?

These are the numbers that indicate how much chance is baked into the combat system. Thats what will determine if players will have to fight 1000s of battles over several weeks to see a good 50% average or if we only have to fight 20 battles in a single night to see that 50% balance.

Those are the statistics that determine how skill-based this game is vs how much it is a game of chance.

--
In fact, it is possible to work backwards too. If you know the sample size (how many fights) at which you want players to "feel" that target 50% balanced winrate, you can then determine what the sigma (std dev) should be for the population (100k simulations). Then hopefully it is possible to fiddle with modifiers/multipliers in the combat formulas to achieve that level of chance in battle. At the very least, it can be used to ballpark where chance should be before playtesting it to see how it feels.

I'll throw some numbers at a spreadsheet later today and share what I come up with.

RuneSlayer wrote:

Good post Realf Lantow.

Realf Lantow wrote:

(Devs, do researches cost AP? we need to know that)

Yes, they do and they are included in the unit's AP value.

Realf Lantow wrote:

Elven HI cannot win fights; at least against Orcs. Or, to phrase that another way, orc HI is super-strong; how can elven HI have an advantage in the simulations? (devs- this one's for you...) did the simulations ignore the specials? or was this just some REALLY bad luck combined with him having +18 researches (which I doubt... given his fame level) the DL HI would *probably* have lost if the AP was equal, HI vs. HI.

Depending on who charged first, the outcome may slightly vary. This is the result after 100.000 battles with capped stats. (Victory ratio)

Elves Vs Orcs 51/49 (Elves charge first) 49/51 (Orcs charge first)

As you can see, the chances of losing/winning an engagement head on is almost 50-50.

Therefore,  depending on who charged first, the tech bonuses and of course the "dice", the outcome may vary.  

Hegorn wrote:

Realf, I think you may just be proving that players should have a high expectation of randomness affecting them in their games. Nonetheless, this is definitely good feedback to give the devs. It shows how much variance players might get in any relatively small number of battles - which is probably about the number of battles anyone would do in an evening. Thats an important scale to examine balance at.

Again, too many variables can determine an engagement. One HI might have "invested" in Armor while the opponent in Melee. As in an engagement you have too many stats working together to determine the outcome of each round (every tick) you cannot just judge from a small number of engagements. Who charges first? Techs? Dice? You need to conduct a lot of tests in order to exclude the luck factor from the "equation".

Hegorn wrote:

One slight possibility -- and I did not interpret Rune's post in this way, but it may be worth double checking for clarification. Runeslayer mentioned "against ALL enemy races" when he mentioned the +.5-1% winrate. Is there any chance he meant that the +1% higher victory rate is an average of ALL the victory rates per race? i.e. Elves V Orcs = 45%winrate, but Elves V Humans = 56% winrate. If you averaged them, it comes out to about +1%?

No...no...no...

Hegorn wrote:

One other thing I thought of -- Runeslayer, when you did those 100k simulations, did you run numbers for control simulations (Same race vs Same race). This would be a great way to get error margins for the simulations. I would expect them to be exceedingly small, but its a good number to double check. It'll put some additional perspective on the other numbers.

Indeed. No problem there.

What benefit's do units get for charging first?

Scaren wrote:

What benefit's do units get for charging first?

They get to attack first and possibly inflict casualties...and then Orcs have the special ability as well.

RuneSlayer wrote:

Scaren wrote:

What benefit's do units get for charging first?

They get to attack first and possibly inflict casualties...and then Orcs have the special ability as well.

Thank you so i'm guessing when units charge head on at each other it's reaction that decides who attack first?

Scaren wrote:

Thank you so i'm guessing when units charge head on at each other it's reaction that decides who attack first?

Well, only when 2 units go head to head. If a unit attacks a unit and the other unit is either facing a different direction or is idle, then the attacking unit gets "charge"...

RuneSlayer wrote:

Max 5 troops from each unit are attacking, unless it is being attacked by several sides where troops will rush to defend the sides as well. However, no more than 5 troops will fight per side of the unit. Obviously, if a unit has less than 5 troops it will only attack that many times as the number of troops left.

What is happening when one unit (20 soldiers) fights face to face against two units? 5 troops fight with one unit and 5 with another (i am asking about damage they can give)

Similar situation one unit (20 soldiers) fights face to face against 6 unit can you say how it looks? (how many troops are fighting, how many damage they can do)

kuba_ wrote:

What is happening when one unit (20 soldiers) fights face to face against two units? 5 troops fight with one unit and 5 with another (i am asking about damage they can give)

Troops will of course rush to meet the enemy, but only 5 to each unit per turn.

kuba_ wrote:

Similar situation one unit (20 soldiers) fights face to face against 6 unit can you say how it looks? (how many troops are fighting, how many damage they can do)

Same thing as above. However, the defending unit has only 20 troops, so it can fight against max 4 units (4x5 = 20). That means that 2 attacking units are receiving no attacks.

How about Cav?
all troops can attack?

kuba_ wrote:

How about Cav?
all troops can attack?

That is correct.

Well this went slightly off-topic.

You all are correct, this is a very inconclusive test; as was mentioned in the original post. I made sure to specifically point that out several times in the original post, and I also invited other players to get involved in the same tests, in order to gather more data.

I cannot test cav; I've yet to unlock them (and I am well over 100k iron away still... 4 building upgrades and 2 researches, to be exact.)

Archers were a very decisive victory for me, but I only tested them against orcs. Part of why I didn't bother mentioning them, is I lost connection during that fight, so I missed the majority of the archer portion of that fight. However, it was a massacre; but part of that was due to levels, my archers were quite leveled up, and it's hard to find a darkie with leveled archers. O.o

So the +melee/armor researches cost AP, that's good to know. Bobba/Scaren, that's part of why your troops cost more in AP points. a good question now (on that line) is, how much AP value does each incremental +3 melee/armor cost per unit?
(say I have +9 melee, and I then research +12 melee... how much more AP will each of my units cost after the upgrade?)

These tests were all conducted with head-on collisions, so theoretically nobody got a "charge" advantage; my elves have lower reaction across the board, and therefore should be capable of getting the first hits in every time.

LI *are* almost equal (if not entirely equal) to HI at the same AP value; but they're pretty much hitting max AP for them to even touch the same AP that an HI is putting out naked; especially once (if at all) AP caps for gear comes out. it's hard to put an LI over 200 AP with hero, it's not hard at all to put an HI at 350-400 AP geared with hero, and it's easy to put the same HI over 200 AP naked. Therefore in a true 1v1 fight (and not these contrived tests), there's really no chance of the LI actually having the same AP value.

This test was not intended to measure synergies/tactics; it was intended simply to give a guide-line as to what units seem to be stronger in 1v1 situations.

I've gotten a little more confidence from the results, that if I use melee troops (balanced between HI and LI) in PVP, instead of trying to use my archers (which still feel like a waste of AP cost in PVP) I might be able to win, GLs taken out of the calculation.

*wishful thinking ahead*
Please get rid of the forced 35% rule, some of us like HI more than LI... Also, as is pointed out, LI are capable of fighting HI AP-to-AP (at least from my experiences) AND it makes tests such as this more complex than they have to be. O.o

each +3 melee research costs 4 AP and each +3 armor research costs 2AP if I'm not mistaken. However, this AP increase is included even for archers (meaning your archers AP->missile power ratio will suffer with many researches).

they should really keep the 35% we can't just have armies of HI running around

Scaren wrote:

they should really keep the 35% we can't just have armies of HI running around

True. But sometimes I wonder if it wouldn't be a good think to allow the cavs in those 35%.

There's a dark part deep inside of me who wanna charge with full HI and cav only to cover the flanks :p Probably just because we can't, since I doubt it would be that powerful.

Guess it's best to keep it the way it is.

Maybe we should scrap the "minimum 35% LIA, LI or archers" and instead use a "max 55% of any unit EXEPT LI".
That way you could use your 50% HI and arty army or 50% Arty and 50% cav without seeing 100% HI armies.