Re: ‘Percent of max hp attacks’, because they might be too op against Heros and, otherwise, useless against everyone else. This is kind of the point; they're a counter to high hp units, just like weak area attacks are only really effective against many units.

So those bonuses for having many of the same unit in play are great.

But I'm not seeing much of a rps system between those 4 unit types. If the enemy spams tanks, it's clear I should spam rangers. What if the enemy spams melee? Are mages area attacks? if so that would help. What if they spam rangers?

I suggest mages beat melee, give melee a buff against rangers, rangers beat tanks, and tanks would be good against mages because of the mages low individual attack. Melee vs tanks would be equal (per dollar), and so would mages vs rangers. But a problem with this is where do heros fit it? they're basically tanks if they have high hp. You could make hero mages and hero rangers to some extent, but not hero melee, since those have the most strength in numbers.

I think an odd number of base unit types may be better, so there's no balanced fights.

One thing to keep in mind is there are side effects speed beyond the fact that it counters ranged units; a fast unit can get to the front quicker. this would make them worse for staring out, and better as support later. That automatic “play later” attribute is probably not very desirable, because it makes the start of play more predictable.

Since you have buildings, you should definitely add slow moving siege. Specifically battering rams and seige towers. That doesn't add to the RPS, but it does add an importance to timing; you want to send in siege precisely timed with when your other units reach the enemy walls.

Since you have ranged siege, be sure to provide at least three variants of that too: splash attacks, frequent single unit attacks, and high damage slow attacks. For this you probably want your stronger units to be completely immune to splash attacks, so that it doesn't automatically become the best counter for multiple units of any type. Unlike units, ranged siege isn't going to be destroyed easily so the balance around it is a bit different.

I think your idea to have a reinforcement learning AI has a problem: at the same time as you work on the AI, you want to work on balancing the game. So if your AI discovers a preferred strategy, instead of making the AI use that strategy, it may be better to change the game so the strategy is less effective.

merkutsam said:

Area attacks to counter numbers, then you need a unit resistant to area attacks to counter the former.

King Mir said:

A unit resistant to area attacks is simply a unit with high hp or armor. The trade off is lots of weak units, or one strong unit.

Or a building like a fortification to shelter units inside. Nevermind, probably buildings aren't a good idea.

Personally I don't like the concept of tanks with many hitpoints. I prefer different types of damage and defense.

King Mir said:
But I'm not seeing much of a rps system between those 4 unit types. If the enemy spams tanks, it's clear I should spam rangers. What if the enemy spams melee? Are mages area attacks? if so that would help. What if they spam rangers? I suggest mages beat melee, give melee a buff against rangers, rangers beat tanks, and tanks would be good against mages because of the mages low individual attack. Melee vs tanks would be equal (per dollar), and so would mages vs rangers.

To make the whole rps topic a little easier to understand and talk about, I have created diagrams (I am sorry for the low resolution; the arrows point from the effective unit type towards the weaker unit type).

I like this idea, in general, yet, there is one point, which I have difficulties with: How can we, logically, explain to the player, that Rangers are effective against Tanks but weak against Melee? Tanks should be known for wearing the heaviest and strongest armor. Arrows and other rather “light" attacks should not be effective against Tanks if they are weak against Melee, I think.

What about the following alternative? It is similar to your idea and based on your idea.

Instead of Tanks being completely immune to Ranger attacks, it could also be, that they take just very little damage.

King Mir said:
But a problem with this is where do heros fit it? they're basically tanks if they have high hp. You could make hero mages and hero rangers to some extent, but not hero melee, since those have the most strength in numbers.

Putting each Hero into one of the different unit classes is a good idea. Could you elaborate on why you think, Melee Heroes would be a bad idea? They could be simply either a Mage/Ranger(/Melee) but with extreme stats.

King Mir said:
I think an odd number of base unit types may be better, so there's no balanced fights.

Sounds reasonable, but I do not really want to remove one of my classes and adding one would not be suitable either.

What about the following rps-system?:

King Mir said:
One thing to keep in mind is there are side effects speed beyond the fact that it counters ranged units; a fast unit can get to the front quicker. this would make them worse for staring out, and better as support later. That automatic “play later” attribute is probably not very desirable, because it makes the start of play more predictable.

Right now, speed of units is handled the following way: Units of the same team form a formation anyways (e.g. take a slot in an invisible formation entity and try to stay on their proper position within the formation). This formation moves with the speed of the slowest unit. Therefore, if you would start with some slow and a few fast units, the formation would move slowly and the fast units would adapt. Only when the fast units are not in a slow formation they move fast. The disadvantage of fast units is that they take up more space, since they use a mount (e.g. horse, wolf, lion) which has a bigger width than usual units.

King Mir said:
Since you have buildings, you should definitely add slow moving siege. Specifically battering rams and seige towers. That doesn't add to the RPS, but it does add an importance to timing; you want to send in siege precisely timed with when your other units reach the enemy walls. Since you have ranged siege, be sure to provide at least three variants of that too: splash attacks, frequent single unit attacks, and high damage slow attacks. For this you probably want your stronger units to be completely immune to splash attacks, so that it doesn't automatically become the best counter for multiple units of any type. Unlike units, ranged siege isn't going to be destroyed easily so the balance around it is a bit different.

This is a good idea and could be great against enemy buildings. We will definitely play around with the concept of movable siege machines.

King Mir said:
I think your idea to have a reinforcement learning AI has a problem: at the same time as you work on the AI, you want to work on balancing the game. So if your AI discovers a preferred strategy, instead of making the AI use that strategy, it may be better to change the game so the strategy is less effective.

Yep, the AI can be used to discover weak spots and to balance the game. Fortunately, within one minute over 30 complete matches can be simulted, currently. This way, it should be rather easy to re-train the AI many times.

merkutsam said:
Personally I don't like the concept of tanks with many hitpoints. I prefer different types of damage and defense.

I guess, both have advantages and disadvantages. It can be frustrating if you have no way of protecting your precious unit (e.g. mages with low hit points). Of course, if one decides to go with tanks, those should have weak spots too (e.g. they move slowly and have a low range or are weak against magic attacks). We have made experiments with different attack types and defenses too. We tried the attack classes “Normal” (Melee, Tank), “Light” (Ranger) and “Magic” (Mage) and implemented two types of armor: “Anti Magic” and “Anti Light”. If units of any class can have that armor, this quickly makes everything more complex and, for example, weak spots of a unit class could be negated with the corresponding armor, quickly making a unit overpowered.

sbn1986 said:

@undefined is ther a vacancy for music composer?

We are already in contact with a music composer, however, if we should not agree on working together, I will come back to you ?

I prefer different types of damage and defense.

I mean instead of tanks based on hit points, I prefer them based on good defense against the normal attack but vulnerable to another attack. Ex: infantry with armor and shields is a good tank against arrows, but less effective in melee than pikes. While pikes are better in melee but vulnerable to arrows. Here you have two different melee units, each one with different strengths and weaknesses. I like to base design on reality as much as possible. If you base tanks on hitpoints it needs a counter based on percentage of health which is another unrealistic design. It can be a good game but I prefer the other way.

1. Your second RPS diagram looks fine. Make it so a tank can kill a melee unit in an integer number of hits exactly, to balance them.
2. The problem with a melee hero, is that the strength of melee is numbers. If you have a more expensive melee unit that's harder to kill, that starts to look like a tank. You might be able to have units that buff melee, but they'd need to have low hp. I guess you could still have a tank that buffs melee, balanced by a melee unit that buffs tanks, like a week support unit.
3. Speaking of support units, they can encourage specialization in one unit type is to have support units that boost only that unit type. That's probably a better mechanic over all than just “I have lots of mages so my mages are stronger”.
4. One advantage that cheap weak troops generally have is that you can ramp them up slower. So you might start with a tank, expect the enemy to counter with mages, then instead of sending a tanks cost of melee units together, you can start sending units as soon as you can afford a single one. Because of this, I do like the idea of not making melee part of the main rps loop. However, just making rangers be the only thing that's good against them encourages a strategy of using a lot of melee. What you can instead do everyone equally good against melee. That means no area attacks by your main three.
5. Making melee week against everyone, and having heroes is also nice because then you have three tiers of units, with three tiers of cost. You could even make this another RPS layer, if you make melee good against heroes.
6. Making melee good against buildings seems a bit strange to me from a realism perspective, unless you give them a special anti-building ability, like sapping underground or explosives. Plus there are more realistic anti-siege weapons.
7. The formation system sounds good, but that only makes sense for the front, not for units trying to reach the front. So fast units still have an advantage reaching the front. So you'd want to make them a tad weaker to compensate.
8. Yeah damage types are great, but you'd need to increase the total number of unit types to make it work. So if you're keeping the unit numbers small, then I agree you should keep it out.
9. I don't agree that percentage of hit points are a particularly unrealistic mechanic. It's basically a way to represent a unit that overcomes armor. To such a unit, an unarmored foot soldier or a fully clad knight look like they take the same amount of hits. It's suitable for a unit that deals blunt damage.
10. If you don't have area attacks to counter melee, you can still have attacks that target a fixed multiple of units: double attacks, and triple attacks.

merkutsam said:

I prefer different types of damage and defense.

I mean instead of tanks based on hit points, I prefer them based on good defense against the normal attack but vulnerable to another attack. Ex: infantry with armor and shields is a good tank against arrows, but less effective in melee than pikes. While pikes are better in melee but vulnerable to arrows. Here you have two different melee units, each one with different strengths and weaknesses. I like to base design on reality as much as possible. If you base tanks on hitpoints it needs a counter based on percentage of health which is another unrealistic design. It can be a good game but I prefer the other way.

I get your point and this is, for sure, something worth to consider. If we would, for example, go with Tanks that take only very little damage from attacks of Rangers, that would be an example of this design paradigm.

King Mir said:

1. Your second RPS diagram looks fine. Make it so a tank can kill a melee unit in an integer number of hits exactly, to balance them.
2. The problem with a melee hero, is that the strength of melee is numbers. If you have a more expensive melee unit that's harder to kill, that starts to look like a tank. You might be able to have units that buff melee, but they'd need to have low hp. I guess you could still have a tank that buffs melee, balanced by a melee unit that buffs tanks, like a week support unit.
3. Speaking of support units, they can encourage specialization in one unit type is to have support units that boost only that unit type. That's probably a better mechanic over all than just “I have lots of mages so my mages are stronger”.
4. One advantage that cheap weak troops generally have is that you can ramp them up slower. So you might start with a tank, expect the enemy to counter with mages, then instead of sending a tanks cost of melee units together, you can start sending units as soon as you can afford a single one. Because of this, I do like the idea of not making melee part of the main rps loop. However, just making rangers be the only thing that's good against them encourages a strategy of using a lot of melee. What you can instead do everyone equally good against melee. That means no area attacks by your main three.
5. Making melee week against everyone, and having heroes is also nice because then you have three tiers of units, with three tiers of cost. You could even make this another RPS layer, if you make melee good against heroes.
6. Making melee good against buildings seems a bit strange to me from a realism perspective, unless you give them a special anti-building ability, like sapping underground or explosives. Plus there are more realistic anti-siege weapons.
7. The formation system sounds good, but that only makes sense for the front, not for units trying to reach the front. So fast units still have an advantage reaching the front. So you'd want to make them a tad weaker to compensate.
8. Yeah damage types are great, but you'd need to increase the total number of unit types to make it work. So if you're keeping the unit numbers small, then I agree you should keep it out.
9. I don't agree that percentage of hit points are a particularly unrealistic mechanic. It's basically a way to represent a unit that overcomes armor. To such a unit, an unarmored foot soldier or a fully clad knight look like they take the same amount of hits. It's suitable for a unit that deals blunt damage.
10. If you don't have area attacks to counter melee, you can still have attacks that target a fixed multiple of units: double attacks, and triple attacks.

1. Thanks for the hint regarding balancing melee und tank units
2. I'll write a separate section about our thoughts regarding heroes in the bottom
3. We consider the following solution regarding “advantage when you have many units of the same type”: When the player has 3 units of a kind in his formation, a support unit (similar look, but carrying a banner which displays the unit class, e.g. ranger, mage or melee) is automatically spawned and will join the units. It will stay alive, even if the 3 units are killed, until enemies kill this support unit. If the player should have 6 units of a kind, a second support unit is automatically spawned. The support units boost all units of the same class of the player
4. Yeah, all unit classes would be equally good against melee, except ranger who would be naturally effective against melee due to their long range (and their cooldown is not as high as the mage cooldown)
5. Making melee strong against heroes is a good idea
6. I agree and siege machines are cool anyways (and add this additional timing aspect)
7. Already the case: only the units on the front form a formation (which adapts to the speed of the slowest unit) and other units walk on their own
8. I'll comment on this one down below.
9. That would be one way to explain it. Additionally, through an armor system (e.g. no/medium/strong armor against regular attacks or against magic attacks; the damage dealt to a unit would be attack damage minus armor value) this could be realized too: an attack that ignores armor would have a similar effect
10. This is what we are probably going to do

Now, there a few more open topics:

Heroes

Because of the formation system, units choose a position in the 1d army formation, which is suitable for their unit class. A tank will be in the front, next will be melee, ranger, mage. Most enemy attacks target the front unit of the formation. Assuming we keep it this way (an alternative would be just keeping the order in which units are built), it might be overpowered, if a Hero would choose a position in the back.

Why do I mention this?

We could put heroes into one of the unit classes (mage, ranger, possibly melee). If we do this, it would be reasonable if a mage hero would choose a slot in the back, have a huge amount of damage and rather low health (for a hero). Due to the hero having a slot in the back, if it would be protected with tanks in the front, it could deal huge amounts of damage, without taking any damage itself (position in back). Of course, this needs to be playtested, but I feel like this could be overpowered.

An alternative would be putting all heroes in the front of the formation, however that does not feel right, if we decide to go with mage or ranger Heroes, instead of just one tank-like hero class.

Unit speed and buildings

Currently, we have the “attacks destroy each other" option activated (e.g. an axe could destroy an arrow or a fireball a spear), I have talked about this in a previous post. Because of that, the player wants to have an army with as much firepower as possible. Due to that, it is actually positive for the player to have an army, which moves slowly. Walls, that block the army, are welcome, because they slow the army down, giving it more time to grow bigger and stronger. This defeats the purpose of the wall. If the health of buildings would be increased and only a few special siege machines can take buildings down quickly, it would increase this effect even more.

I have not played any game of this kind yet, which used buildings, such as walls. Maybe, because they are really difficult from a game design perspective (chances are that there are such games, which I just do not know). If the “attacks destroy each other" option is deactivated, the effect is not as big (but still exists). I like that option though, because, at least without a proper RPS unit class design, the game felt boring without this option and feels fun with the option (but is harder to balance).

Many different unit types

We have talked a lot about unit classes already (e.g. hero, tank, melee, ranger, mage). Every unit is in one of those classes. However, this does not mean, that there are only 5 unit types: Every single faction/tribe has a unique set of 5 unit types. Examples: the goblin tribe has a goblin boss, a goblin tank, etc. The dwarf tribe has a dwarf boss, a dwarf tank, etc. All those individual unit types are based on the 5 main unit classes, yet they have a few minor differences, such as

• their attack (e.g. golden spear vs mad boomerang vs god axe)
• mount / no mount
• rarity (common, rare, epic, legendary). The rare a unit type is, the higher the stats are but the more expensive it is
• optionally vampire effect (regain some health via damaging enemies)
• a few more differences (for example higher attack rate but lower damage)

My main goal is, to balance the 5 unit classes. The individual unit types should feel a little bit different too, but I am fine with them having just minor differences. I understand, that it is, basically, not managable, to create over 50 unique, well-balanced unit types.

After having put some more time into playtesting and balancing the 5 unit classes, I will think about the possibility to introduce different armor levels (as mentioned in my response to point 9).

Btw.: I really appreciate your time and your help ? Thank you!

King Mir said:
I don't agree that percentage of hit points are a particularly unrealistic mechanic. It's basically a way to represent a unit that overcomes armor. To such a unit, an unarmored foot soldier or a fully clad knight look like they take the same amount of hits. It's suitable for a unit that deals blunt damage.

Only if all units have the same amount of hit points. If a mage has less points, or a cavalry has more points, then that counter-armor unit will do it worse or better than a normal attack against those units.

@logende

I've been thinking in your game development. I'm guessing so I may be wrong.

First the game was about sequence, unit X beat unit Y, then unit Z beats unit X, then unit Y beats unit Z, and the full circle X beats Y.

This was a waste if an expensive unit beats another one only to be defeated by a cheap one. So you added a formation mechanic. Now units in the front protect weak and expensive units in the back.

But this was a constant meat grinder. Front line units wiped out, and new replacements sent non stop. So you added the missiles cancel each other mechanic.

Now both formations fight as a whole, every unit adding its attack to the struggle until one formation's firepower overcomes the other's. Once this happens, the other formation is wiped out and this is pretty much game over.

In the end what you have been saying is that you prefer a more complex gameplay, one army defeating the other combined army as a whole, not piece meal. Probably not what you want but here you have a few ideas.

• Formations may fight in an abstract way calculating the different combined attacks and defenses, and taking a percentage of their units health, and they may recover health when not in combat.
• Formations don't need to be destroyed everytime they are defeated. They may fall back to their defensive positions if they are losing.
• Formations may advance and retreat, may entrench and fortify, may besiege and assault.
• Formations may fight better in their own lands, and worse in enemy lands
• Formations may change their composition, retiring some units and adding others.
• Formations may apply different tactics for different situations.
• Support units may add different capabilities to formations. Workers, engines, cavalry, spies and counter spies
• Ranged units may have different attacks. Bowmen bombard formation with weak area attack. Crossbowmen like snipers against expensive units. Musketeers piercing attacks anti armor
• Heroes don't need to be combat units, they may be characters leading a formation with different capabilities derived from their traits.
• Mages don't need to be damage dealers, they may be support. They may add buffs and debuffs to attacks and defenses, they may increase healing rate or movement
• Economy may be based on maintenance instead of a fixed # gold per time. One player can maintain a given amount of units. If he uses more expensive units he will have less gold for other units.

1. Spawning support units automatically is an interesting variation. However, you're taking away an opportunity for the player to make a choice: the player has 3 rangers on the field, the enemy has a lot of units week to rangers. Does the player build a support unit to boost their rangers, or do they build a third ranger? In general, more player choices like this are good. Even if the choice is obvious.
2. Long range is like armor, it means you can take more strikes against the enemy before they can hit you. It's particularly effective if you can kill the unit before it reaches you. So at the very least to prevent rangers from dominating melee, you should make sure that they do little enough damage so that a single melee unit will land at least one attack vs a single ranger.
3. I want to enumerate what the choice that melee does not participate in the main RPS loop should mean. It should mean specifically, in a battle where one side is pumping out melee units as fast as it can, and the other side is pumping out one of the 3 main types as fast as they can, two things should be true: 1) the side pumping out the 3 main types should win. 2) all three main types should win by approximately the same advantage. This is a very testable design objective.
4. If you make tank armor effective against melee as it is effective against rangers, then tanks will be strong against melee too. This is good if you want to keep melee out of your RPS main loop.
5. A more traditional RPS system, without mages is archers v melee v cavalry. Archers beat melee because they can mow them down before they are reached, melee beat cavalry with long pikes, and cavalry mows down undefended archers. You're trying to have a RPS system of horse archers vs melee (tanks) vs archers (mages). Not only is this going to be more difficult to balance, but I gotta ask: where's the cavalry? Not horse archers, knights. You're missing a staple combat unit.
6. I agree that it's a good idea to have heroes fight with their matching unit type; tanks in the front, ranged power attackers in the back.
7. I don't think that having attacks cancel each other is a good mechanic. It makes units stay on the field longer. It's the same problem as healing. I think you want your units to die quickly, so the game is closer to RPS than an effort to build an army.
8. Why those 5 units? Would like to understand what led to these 5 specific unit types.
9. Totally agree that a fixed gold per minute is better than getting gold for kills. Gold for kills just ends the battle too quickly.
10. Formations falling back might work if it's automatic. If you make it a player choice, you're back the problem that it's an advantage to stay back and build up troops.
11. Fighting better in your own lands might have the consequence that your strategy would need to change depending on how deep you are into enemy lands. Not sure how desirable that is.
12. We've been giving minor suggestions here an there, but perhaps a better approach is to establish more clear design objectives. Ideally testable objectives. What this amounts in this game is, different ai strategies and how they perform against each other. Done right, some strategies will be balanced, while others will dominate. You want to balance the player experience going to be like when he himself used one of these strategies. To start off with, specifically, you have one strategy focusing on each of the units. One strategy that cycles the main rps loop evenly. Now you gotta decide, who will win: balanced or one-sided? Is it a close? What about all in on one unit type, plus the associated hero or support type?
13. If you're going to be tweaking 50 individual units eventually, then you're going to want to set up good automation to test that they are properly balanced.
14. Another design objective should be managing how much luck matters. In other words, with any given strategy combo, what are the odds that the underdog wins? Make this too high, and the game stops being strategic. Make it too low, and the game becomes too predictable.
15. I think one specific design objective should be that an ai that reacts to the units the enemy sends out should dominate one which doesn't.
16. In general, you want battles to be long enough for a player to change strategies if one is not working. But for testing design you want the opposite: you want the test to be exactly long enough to show a strategy is inferior.
17. With regards to buildings, right now I'm not sure what they add specifically. Why do you want buildings, instead of an open battle field? What feel are you trying to achieve? It would help to identify the design goal here.
18. You mentioned 50 unit types for 10 factions. What's odd about this choice is that the player can only play as one faction at a time. So there's a lot of gameplay that a player won't see in a single playthrough. I suggest keeping the variation between factions small, or have much fewer factions.
19. Will there be a campaign mode? That effects the design quite a bit because then it's important to provide a way to ramp up the difficulty, whereas otherwise you want to make all the player choices obvious and apparent from the start.