Balancing for Beta 2

[KGB]

SnotlingG,

I agree with Max Warlorder that capitols should not start at L8. The player should definitely need to spend the 400 gold to get that last level of protection.

I'm not big on waiting 1 turn per upgrade time. It will make it one more thing for players to micromanage. In other words if my walls are at L2 and I want to go to L5, I will have to for 3 consecutive turns remember to go to that city and upgrade the walls. If you can set multiple turns worth, then you can't charge players for gold until the upgrade happens because it's not fair to charge if the city is lost mid upgrade. Plus this feature makes the game more Age-of-Wonders like where you had to wait many turns for everything to complete. Warlords is mostly a 'do it now' game other than unit build times. I prefer it that way so that if I have 800 gold to spend, I can spend it right now and upgrade a city on my front lines based on my needs.

The 2 hp to 3 hp change + healing seems like too big a change for Beta 2. The ABSOLUTE first thing I will want is the ability to change the order in my stack on EVERY unit. In other words if my stack contains just 2 Griffons and I attack another stack and 1 Griffon gets wounded so it's down to 2 points then if I attack another stack I want to change to my healthy Griffon up front, not my wounded one (same at the end of my turn in case I get attacked on someone else's turn). So all that happens is players will put wounded units in the back and healthy ones up front.

And you are totally right that good stacks will become even better. It won't be a slight bit better either, it will be massively better. I'd suggest you model this in a simulation for quite a while before implementing it in the game and posting the simulation results. For example take a 4 strength unit vs a 2 strength unit. The numbers work out to:

4 Strength unit chance of hit 4/20*18/20 = .18
2 Strength unit chance of hit 2/20*16/20 = .08

.18/.08 =- 2.25

So the 4 strength unit is 2.25X as likely to get a hit, not the intuitive 2x.

The chance of winning for the 2 strength unit is:

.08/.26 * .08/.26= .094 (2 hits in a row. Bottom denominator .26 is .08+.18)
.18/.26 * .08/.26 * .08/.26 = .065 (2 hits out of 3, with 4 strength unit hitting first)
.08/.26 * .18/.26 * .08/.26 = .065 (2 hits out of 3, with 4 strength unit hitting second)

for a total of .224 of 22.4% chance to beat the 4 strength unit.

When you go from 2 hits to 3 hits the chance of a 2 strength unit winning drops by almost a third

.08/.26 * .08/.26 * .08/.26 = .029 (3 hits in a row)
3 hits out of 4 case: (4 chose 3)
.18/.26 * .08/.26 * .08/.26 * .08/.26 = .02 (3 hits out of 4, with 4 strength unit hitting first)
.08/.26 * .18/.26 * . 08/.26 * .08/.26 = .02 (3 hits out of 4, with 4 strength unit hitting second)
.08/.26 * .08/.26 *.18/.26 * .08/.26 = .02 (3 hits out of 4, with 4 strength unit hitting third)
3 hits out of 5 case (5 chose 3). Here I will not show all the cases, just the final number.
.18/.26 * .18/.26 * .08/.26 * .08/.26 * .08/.26 = .013 (3 hits out of 5, with 4 strength unit hitting first and second)
.013 (3 hits out of 5, with 4 strength unit hitting first and third)
.013 (3 hits out of 5, with 4 strength unit hitting first and fourth)
.013 (3 hits out of 5, with 4 strength unit hitting second and third)
.013 (3 hits out of 5, with 4 strength unit hitting second and fourth)
.013 (3 hits out of 5, with 4 strength unit hitting third and fourth)

for a total of .167 or 16.7% chance to beat the 4 strength unit.

The numbers get even more dramatically skewed against the lower strength units as the other unit gets better (ie strength 6 instead of 4 or 8 instead of 4)

If you are *really* convinced that the random number generator you have isn't doing a good job the other possible thing you can do is run each complete stack combat 100 (or 1000) times. Then use the average results for those 100 or 1000 complete battle rolls to determine the 'average' result for that combat. From there you can calculate a standard deviation (easy math calc) and only allow 1 (or 2) standard deviations from the norm. Then run an actual combat and take the first result that falls within 1 (or 2) standard deviations from the norm. That way you can't get those massive fluke battles where 8 light infantry fight 8 light infantry and one side somehow kills 8 with only 1 or 2 losses instead of the average of 6-7 losses.

As far as healing goes, what do you mean 'in the same turn'. As in your turn or an entire game turn? For example, imagine I am player 1 in an 8 player game. Lets say player 2 damages one of my units on his turn. Will it heal at the end of player 2's turn, at the end of the game turn (after player 8) or at the start of my next turn? If it heals at the end of player 2's turn everything works out fine. If it heals at the end of the game turn or at the start of my next turn then you get the situation where after player 2 damages my unit, players 3-8 benefit from that damage. But if player 8 damages me instead then none of the other players benefits because I go immediately after player 8. So you get uneven benefits from damage based on where you are in the turn order and where the player you damaged is. This issue ended up in Warlords IV which has damage/healing/large numbers of hit points and has never made anyone happy. So overall I'm not a fan of permanent damage lasting beyond a players turn (ie once player 2 finishes attacking, all damage should be restored to all units so no one benefits/loses out by where they are in the turn order)

KGB

[piranha]

I think you will be able to set the wall level when creating a map so some maps could start with lvl 0 while others with lvl 8.
For the random map generator there might need to be a standard setting but I like the idea that you get to do some buildup and decide where to spend you money rather than start with L8.

I agree with KGB that it should not take time to build lvls. Besides the fact that is another thing to micromanage and not forget to do its hard to predict enemy movement 3 turns away, and when FOW is introduced it will be even harder.

You will be able to organize order of unit types, but I don't know if you will be able to decide which unit of the same type to be first or last, perhaps snotling can comment on this?

I will talk the dice over with Snotling but it might change the game too much with 3 hits. But making a test function and let it run 1000 times and get a better idea of the results sounds like a good idea.

The healing thing will work in such way that it will heal all units for all players at the end of every players turn. Sure you may put you full health unit first when you attack. The idea is to make it more useful to attack with many stacks vs another army. In particular to wear down a very powerful army that is walking around with STR 13-14.

[KGB]

Piranha,

I think you will be able to set the wall level when creating a map so some maps could start with lvl 0 while others with lvl 8.
For the random map generator there might need to be a standard setting but I like the idea that you get to do some buildup and decide where to spend you money rather than start with L8.

That seems very reasonable to me. If you create a custom map you can set it as you please while random maps max out at 7 for capitols.

You will be able to organize order of unit types, but I don't know if you will be able to decide which unit of the same type to be first or last, perhaps snotling can comment on this?

Sure you may put you full health unit first when you attack.

Your 1st statement sort of contradicts the 2nd. If you are going to carry damage for an entire turn it's very important that you can re-order each unit individually so that you can put a damaged unit behind a full strength one of the same type (ie 2 Griffons).

The idea is to make it more useful to attack with many stacks vs another army. In particular to wear down a very powerful army that is walking around with STR 13-14.

In Warlord3 to wear down very powerful armies a skill called assassination was added to the game. It is just a straight % chance to kill a unit outright. It only works against the first unit faced and happens before the combat rolls take place. So for example if a unit has a 30% assassin skill it makes a 30% roll against the 1st unit it faces in combat. If it rolls successfully the enemy is slain instantly. If not, normal combat proceeds. If the assassin skill unit survives it does NOT make another assassin roll against the next unit it faces. This ability meant that units with this skill could be used against super stacks.

The other way to wear down super powerful stacks is to increase the dice range from 20 to say 25 or 30. Warlords 2 did have this feature. It was called 'enhanced combat'. I believe it changed the max dice roll to 30. This means weaker units tend to get more kills against stronger ones and is a way to wear down those super hero stacks.

The healing thing will work in such way that it will heal all units for all players at the end of every players turn.

Perfect. That's the only fair way to do it.

I will talk the dice over with Snotling but it might change the game too much with 3 hits. But making a test function and let it run 1000 times and get a better idea of the results sounds like a good idea.

Essentially this works as a smoothing function to prevent extremes. Here's how I'd do it.
A) Run the combat 1000 times and get the average result.
B) Round up/down the number of survivors to the closest whole number
C) Allow a 1 (or 2) unit standard deviation from the average
D) Run the real combat. If it falls within the standard deviation range, use those results. Otherwise re-run the combat.

Example 1: Side A: 8 light infantry vs Side B: 8 light infantry (no bonus's)
A) Running 1000 results will show one side or the other winning about 53% of the time with 1/10 man left (assumes the random number generator isn't perfectly going to get 50/50). The reason you'll get about 1/10 of a man left is that because sometime one side will win with 5 men left while others it will be 2 or 1 or 4. When you add up all the wins for a side + all the men left and average those it will not be an integer.
B) Round to the nearest whole number = 0 survivors
C) A 1 unit deviation means either side A or side B can win with 1 man left. 2 unit deviation (probably better) means either side A or side B can win with 1 or 2 men left.
D) Run the real combat. Assume the first result is side A with 3 men left. This is outside the range of 1-2 men. So run again. This time it's side B with 1 man left. This is the result that is used for the combat.

Example 2: Side A: A hero stack with 8 units vs Side B: 8 average units
A) Lets assume running 1000 results shows Side A wins on average with 4.8 survivors.
B) Round to the nearest whole number is 5 survivors
C) A 1 unit deviation means side A wins with 4-6 men left. A 2 unit deviation means side A wins with 3-7 men left (again I like 2 unit deviation better)
D) Run the real combat. Assume the first result is Side A wins with 6 men left. This is the result that is used for the combat.

Example 3: Side A: 1 light infantry vs Side B: 4 light infantry (no bonus's)
A) Run the combat 1000 times to see that Side B wins on average with 2.9 men left
B) Round to the nearest whole number is 3 survivors
C) A 1 unit deviation means Side B wins with 2-4 men. A 2 unit deviation means Side B wins with 1-4 men (again I prefer 2 unit which I think overall works better to allow a wider range of results)
D) Run the real combat. Assume the first result is side A wins outright. This is outside the range so it is not used. So run again. This time Side B wins with 1 man left. This is in the range (for 2 unit deviation) so it is used for the combat result. If a 1 unit deviation is used, the combat must be run again until side B wins with 2-4 men left.

KGB

[Itangast]

I dont think that introducing 3 hp and "healing hp" each turn would be for the better.

I very much appriciate the strategic challange in putting together good stacks trying to forsee the outcome of different battles. Just to get a slight strategic edge.

Though the more I think of it I realize how important randomness really is - to keep the fun factor. I have encountered fights during beta now that were upsetting and even unbelievable ones. But I do not think a radical change is required. Just a small tuning of the calculations.
In short: I dislike the idea of fighting bad odds with no foreseeable chance of luck.

Also keeping track of wounded units during a turn introduces a new uncomfortable factor to consider. Stack order will be a thing to consider as mentioned all ready. Simplicity is some cases is more fun and therefore the better option, keep it simple.

Also the introduction of Ally units changes a lot of things and probably would require a new balance change.
(ie better to wait with radical changes)

My suggestion is to change only minor things battle-wise. The unit changes you have made will also need to be considered and perhaps its better to try them first as well.

If you want to evaluate 3 hp + healing hp, then perhaps make this a configurable option when starting a game. Then the creator of the game will decide. This will result in easier way to compare a 2 hp game vs a 3 hp game... comparison is good.

[piranha]

The assassin skill sounds like a interesting idea to deal with super stacks. Is it only the attacker that get to use its assassin skill or both sides?
What happens if both roll a assassin kill? will they both sneak up on each others back and put a knife in the back simultaneously ?

We might arrange units so that those with most move left fight last, and those with most HP first.

If we add more sides to the dice, that should increase the randomness to the result. But perhaps adding sides to the dice and also 1 more hit would balance it.

About the idea to run the battle 1000 times. Battles are calculated on the server and I have a feeling that it would be wasting resources which is something we try to minimize. I'd be more interested in changing the dice, changing hits or adding a new features like assassin.

We wont have ruins/temples/items in the next beta and those might be needed in the game to know for sure if something new works, but I want to try some of the ideas we talked about.

[KGB]

Piranha,

The assassin skill sounds like a interesting idea to deal with super stacks. Is it only the attacker that get to use its assassin skill or both sides?
What happens if both roll a assassin kill? will they both sneak up on each others back and put a knife in the back simultaneously ?

Both sides (attacker/defender) get to use the assassin skill.
If 2 units that both have an assassin skill meet then the skill levels cancel each other out. So if Unit A has assassin skill of 30% and it faces Unit B with an assassin skill of 20% then Unit A gets a 10% (30-20) chance to assassinate unit B. If the numbers are equal (30% vs 30%) then neither gets an assassin strike.

Remember that the assassin skill is a 1 shot deal against the first enemy unit faced. So for example:
Side A: 1 unit - 30% assassin skill
Side B: 2 units - each with 20% assassin skill

When the first 2 units meet, Side A will get a 30-20=10% chance to assassinate Side B's first unit. Assuming Side A's unit wins (either by assassination or by normal combat rolls) then when the 2nd side B unit enters battle it gets a 20% chance to assassinate Side A's unit because Side A's unit already used up its assassin skill.

So even if a super stack has an assassin unit, it only helps against the first enemy it faces, after that it can be assassinated by the remaining units.

If we add more sides to the dice, that should increase the randomness to the result. But perhaps adding sides to the dice and also 1 more hit would balance it.

Maybe. It's impossible to predict without running a LOT of calculations and tests. That's why I'd recommend leaving it out of the next Beta. The other thing is that because the 3 hp affects the chances of winning it means that it also should affect the production times + costs of units. In the example I showed a few posts above with the 4 vs 2 strength unit the winning % changed from 22% to 16% or roughly by 1/3. So to keep the balance, the 4 strength unit should now cost 1/3 more to buy and take 1/3 longer to build. So you'd end up needing to work out new production times/costs for units. That doesn't seem easy to do.

About the idea to run the battle 1000 times. Battles are calculated on the server and I have a feeling that it would be wasting resources which is something we try to minimize. I'd be more interested in changing the dice, changing hits or adding a new features like assassin.

Does 1000 battles really use that much CPU? I wrote a complete combat calculator (C language) for Warlords 3 back in 1999 and doing 1000 battles took less than 1 second on the computer I had then and the Warlords 3 combat rules were a lot more complex.

The absolute easiest way to weaken super stacks is to change the dice (make it a game option as it was in Warlords 2) from 20 to 30. That will tend to favor quantity (masses) over quality (a few super units/stack).

The next easiest would be to add an assassin skill to some unit. In reality it might make sense to have a couple of units have the assassin skill at differing levels. In Warlords 3, there was a 1 turn unit (equivalent move/strength stats to light infantry) that had a 10% chance, a 2 turn unit (equivalent move/strength stats to heavy infantry) that had a 20% chance, and a 3 turn unit (equivalent move/strength stats to heavy cavalry) with a 40% chance. Later another 3 turn unit with a 50% chance was added that had light infantry movement but only 1 strength (basically 1 strength light infantry).

KGB

[piranha]

I've spent the day on working with battle functionality. I'm posting the latest balancing ideas.

I've entered sight radius on units but FOW won't be implmented when we start beta2 because there are so many things that we need to changed and tested first.
Assassin, or critical strike which we call it is a chance to kill the enemy unit right away. Each unit have 1 chance to make a critical strike just before i starts to fight. When it missed its critical strike the battle continues as normal. If both units in a battle makes a critical strike the dices for critical strike will roll again. When facing a new opponent it wont get to make another critical strike.

The critical strike skill should work to counter super stacks pretty well. The Ghost unit is changed to a unit specialized in critical strike. A army of only ghosts would be quite dangerous to a super stack because each ghosts bonus is added to the total chance to make a critical strike. Its weak in normal battle so should avoid low level and medium level armies.

We have a option when hosting a game to decide if allies can be bought in the city or if they can only be found when searching or buying a hero. Default will be that you can buy allied units in cities.

Archers are now called elf and does +2 vs flying units.
Archers and wolfs are quite good at critical strike and can be used to hunt super stacks.

Dwarf can group move in hills at the cost 1.

Ram is introduced and does -1 vs city while catapult is more expensive now and does -2 so a city with lvl 8 will still have def 1.

Bat 1 turn, cost 100, STR 1, move 22, sight 5, assassin 0%, Flyer
Scout 1 turn, cost 100, STR 1, move 18, sight 4, assassin 0%, move in difficult terrain at 3
L. Inf 1 turn, cost 100, STR 2, move 14, sight 4, assassin 2%,
Orc 1 turn, cost 125, STR 2, move 14, sight 3, assassin 2%, move in swamp, +2 in swamp
H. Inf 1 turn, cost 200, STR 3, move 12, sight 3, assassin 3%, +1 in open
L. Cav 1 turn, cost 300, STR 3, move 22, sight 4, assassin 2%, +1 in open
Elf 1 turn, cost 350, STR 3, move 16, sight 4, assassin 5%, +2 vs flyers, move in forest
Dwarf 1 turn, cost 400, STR 4, move 8, sight 2, assassin 2%, +2 in hills, group move at 1 in hills
Pikeman 1 turn, cost 500, STR 4, move 8, sight 3, assassin 3%, +2 in open
Wolf 2 turn, cost 500, STR 4, move 16, sight 5, assassin 10%, +1 group in hills, move in hills and forest
Giant 2 turn, cost 600, STR 5, move 18, sight 5, assassin 1%, +1 hill, move in hill
H. Cav 2 turn, cost 750, STR 5, move 20, sight 4, assassin 1%, +2 open
Minotaur 2 turn, cost 800, STR 5, move 14, sight 4, assassin 3%, +1 city
Ram 2 turn, cost 850, STR 4, move 16, sight 3, assassin 0%, -1 city
Spider 2 turn, cost 1000, STR 5, move 12, sight 3, assassin 2%, +2 city
Pegasi 3 turn, cost 1050, STR 5, move 20, sight 5, assassin 2%, +1 group, Flyer
Griphon 3 turn, cost 1200, STR 6, move 18, sight 5, assassin 2%, +2 city, Flyer
Catapult 3 turn, cost 1250, STR 4, move 10, sight 3, assassin 0%, -2 city
(A) Elemental 3 turn, cost 1300, STR 6, move 18, sight 7, assassin 10%, (possible +2 sight to group), Flyer
Medusa 3 turn, cost 1400, STR 7, move 12, sight 3, assassin 6%, -1 group
(A) Demon 3 turn, cost 1400, STR 7, move 18, sight 5, assassin 6%, +1 group, Flyer
(A) Wizard 2 turn, cost 1400, STR 5, move 50, sight 4, assassin 3%,
Elephant 3 turn, cost 1500, STR 6, move 14, sight 5, assassin 0%, +2 group in open
Unicorn 4 turn, cost 1500, STR 7, move 14, sight 4, assassin 2%, cancel terrain bonus
(A) Ghost 3 turn, cost 1600, STR 4, move 18, sight 0, assassin 8%, +5% assassin to group, cumulative
(A) Devil 5 turn, cost 2000, STR 7, move 18, sight 4, assassin 4%, +1 group, cancel non hero bonus
(A) Archon 5 turn, cost 2000, STR 7, move 16, sight 5, assassin 3%, cancel hero bonus, Flyer
(A) Dragon 5 turn, cost 2200, STR 8, move 24, sight 6, assassin 2%, +2 group, Flyer

[Itangast]

Elves being a 1 turn unit with assassin bonus of 5%, combined good movment and bonus vs flyers sounds as if they will be a bit too good i'm afraid.

Without having tested the assassin bonuses I think max 3 % would be OK for 1 turn units (as starting values).
Elves 3% assassin, is my suggestion.

Wolves having 10% is way too good. We will all be running around with wolf stacks lead by a pegasi.
5% for wolves would still make them desirable

Increase the assassin bonus for Minotaurs to 4 or 5% to make them more desirable.

[KGB]

Itangast,

I don't think you have to worry about Elves assassin bonus. At 5% it means it takes 20 Elves to get 1 kill from assassination on average. 32 Elves in a city (completely full) only gives 1.5 kills on average from assassination.

Most of the numbers are so low that at best you'll get 1 lucky kill from assassination. Even Wolves at 10% only kill 1 unit per 10 Wolves. At 20 turns to make 10 Wolves I doubt it's going to be very effective and I very much doubt anyone is going to have stacks of Wolves lead by a Pegasi.

In Warlords 3, players rarely worry about any assassin values <30% because it takes such huge numbers of units to give a real chance to do damage.

The Ghost with the +5% cumulative group chance is the only real effective assassination unit as a couple of Ghosts in a city would add 10% to every unit. If Ghosts can be built in cities then they become really dangerous.

KGB

P.S. Piranha, in your development blog I see you mentioned finishing Naval combat and wondering about bonus's. I agree with Itengast that all fliers (including a flying hero) should retain their bonus's when grouped with actual naval units. Players are used to a Pegasi giving a +1 stack bonus to naval units and it makes sense. One thing you *might* consider adding is a Water combat bonus similar to the Open/Woods/Hills combat bonus. The Serpent Ally unit (sea serpent) from Warlords 2 would make an ideal unit to give a +2 combat in Water.

[piranha]

I checked and you are right that the bonus is applied to naval units. I will change this for the next beta.

Water bonus could be interesting to add for some units.

How do you think the ghost should be treated? Should there be a max Critical Strike percent. If you would have 20 ghosts in a city it would be quite expensive to capture it since you would more or less lose a couple of units for sure until you killed of some ghosts.

On the other hand, sitting with 20 or more ghosts in a city is like sitting there with 15 dragons in a city. I can't see it being a winning strategy in the game.

About the assassin / Critical strike bonus. Like KGB says, it tried it quite a lot while creating it and the effect is small. If you think about how the dice can let a weaker unit kill a stronger quite often and compare that with the assassin kill chance I doesn't seem like it will be overpowered.

The assassin concept is to make some low level units quite dangerous to the highest level units. So even the best stacks have to worry about facing some units.

What do you think about the following changes:
Wizard to move 36 and +3 vs flying units
Elf assassin 7%
Wolf assassin 16%
Minotaur: assassin 12%

Is giant underpowered?
Elemental?
Any other unit that comes to mind that noone want to buy?