Removing randomness from combat in a strategy wargame

The boys and I have a strategy wargame idea we've been kicking around for years. Our original idea for the combat system is very simple: given an attack strength A, versus a defender strength D, then the probability of winning is A/(A+D). If a random.value is less than that, the defender is destroyed; otherwise the attacker is destroyed.

But the randomness of this is bothersome. If my attack strength is 9 and your defense is 1, I expect to clobber you, and have probably built my entire turn plan around the assumption that I will... but 10% of the time, the attacker will lose in this case. This feels unfair, even when I know the math and understand probability very well intellectually. It leads to the temptation to save-scum, which in fact I have been known to do fairly often in Civ Rev, though I'm not proud of it.

So! Now we're wondering about eliminating the randomness. Can we make combat that's interesting, and takes into account attack and defense strength, without rolling any dice?

I know that in Advance Wars and derivatives, units start with 10 hit points, and lose those based on attack/defense strength. But, while I'm told there's no randomness to it, the calculations going on are so complex that I can't predict them — the result feels random. And I end up with a bunch of middling-damaged units which, since they do less damage than fully healed units, are more or less useless. I strongly prefer Civ's approach: three hit points, and it only takes one turn of healing to gain a hit point back.

Moreover, to keep things moving along, I'd like each encounter to resolve with one of the units being destroyed, and the other unit taking some (possibly zero) damage. Having to spend multiple turns with two units whaling on each other is just too slow-paced for us. There could be multiple rounds of combat, but we're going to resolve them all at once.

Perhaps we still compute A/(A+D), but then look up the result in a table:

• up to 0.25: no damage
• from 0.25 up to 0.5: 1 point of damage
• from 0.5 up to 0.75: 2 points of damage
• 0.75 or more: 3 points of damage

So on the first round of attack, if evenly matched, the attacker does 2 points of damage. If the defender had 2 HP or less, it is destroyed; otherwise it immediately fights back, but this time using D/(D+A), and looking up the result in the same table. And we continue until one unit is destroyed (or we realize that neither unit can hurt the other!).

In the case of two perfectly matched units (my attack equals your defense), both in perfect condition, I would take your unit out but suffer 2 HP damage. That seems reasonable. If my attack is twice as strong as your defense, then A/(A+D) = 2/3, so I'd destroy your unit and take 1 point of damage... though if my unit had only 1 HP left and yours was full, I would deal 2 HP to you but be destroyed on your first counter. If it's the attack 9 vs strength 1 case, I would destroy your unit in one hit, and take no damage (regardless of my condition).

Computing all this might still be a bit much for the player to do in their head, especially when the attack/defense strength numbers get larger. But perhaps we could show the outcome ahead of time, before the player confirms the attack. So it's never a matter of luck or guessing what's going to happen.

But this is all just brainstorming. Please let me know what you think, or if you have better ideas!

 

That sounds terrible. Play some Fire Emblem and see if you still like the idea of your units instantly dying in a turn. The worst case scenario is you see half your forces annihilated in a single turn because of the enemy steamrolling through.

In most of these games, battles of attrition aren't that common. If that becomes common it's because you aren't leveraging type advantages or terrain advantages properly.

It's been a long time since I've played an Advance Wars game. I could have swore it was random, but I also knew what my probabilities were.

 

Although I completely get why it would feel unfair when looking only at the stats that is not something you should do, right? If the results of the encounter were based entirely on stats then sure you could look at it the way you are.

When the random factor comes into play it is representing something. Some factor not known from the stats. Perhaps as simple as slipped on a banana peel / tripped over a rock / distracted by a sound or any of countless other possibilities (okay the banana peel would be pretty far fetched).

Maybe if you want to do this kind of thing with the randomness you can simply define a number of possible random events that can happen. Obviously, you will be able to come up with a better list but if you had something like this...

• tripped
• misjudged strike (missed)
• accidentally moved into opponent's strike
• weapon broke
• lucky strike
  
• etc

Then if your random factor came into play (I know in your case everything is actually random and compared to stats) you could display a message "unfortunately little dude tripped" or whatever so now there is a concrete reason as to what happened. It wouldn't feel like what in heck happened... I should have won! because you know what happened. Of course, I suspect it would feel like oh come on! The dude tripped over a dang rock and lost the battle?! which may be frustrating but at least you now know what happened and there is some satisfaction in that. At least for me.

Basically, just providing more information, more feedback on what happened. Because when you are using random in this manner there has to be some reason for it, right? I mean the random should represent something that occurred beyond the stats that influenced the outcome of the battle. Currently, you just don't know what that was. Did a unit get lucky scoring a critical hit? Did a unit falter (tripping, being distracted, etc)?

A bit unconventional I imagine and I am just throwing it out as something to think about.

 

Have you tried playing out battle rounds on a game board? Change rules on the fly etc.

If battle results are known based on simple non random battle results then players can play strategically & plan ahead like in chess. If you introduce a small random element then players need to start playing tactically & have support units around just in case they make a bad roll.

If you wanted to add an element of 'controlled' variability you could add a risk/reward into the game by using things like ammo/overheating/risk of jamming etc. if these things are predictable (the longer you shoot without cleaning the weapon the higher chance the gun will jam etc), & take a player turn to counter, then the player has to decide whether to push their luck or spend time countering the risk to increase the chance of success later. Of course the opponent could exploit that break if they have planned ahead for those things themselves.

Things like ammo do this to a small degree but there is so much ammo lying around most maps, or guns that can be safely picked up, & reloading is so simple & quick, that there is little risk to the player that they need to even consider lack of ammo as a risk.

 

That's a fun idea. I don't know if it would really solve the "oh come on!" reaction, but it might help.

Particularly if you had animated battles, and a good (and prolific) animator, one could have a lot of fun with this — different scenarios that "explain" the outcome, and the more rare an outcome it is, the more outrageous the animation could be. In Civ, I've occasionally lost a wing of bombers against a spearman. If the battle commenced and I saw an animation of the spearman sneaking onto the airfield and poisoning the gas tank on all the planes, at least I would get a chuckle (and some believable story) from it.

I'd like to explore the non-random combat idea more, but this idea has real merit and I'll tuck it away for future use!

Thanks,
- Joe

 

No, but it's a good idea. This would be easy to prototype with little paper units on a big grid.

Right, in fact the system proposed reduces to Chess if every unit has an attack of at least 4 and a defense of 1 (i.e., the attacker always wins without taking damage). It just allows for a bit more nuance — there will be many cases where one unit can't take out another unit without taking some damage, or even where several weak units have to gang up (sacrificing all but the last) in order to defeat a stronger one.

Yeah... I'm just curious to see how far we can go without any randomness at all, beyond the board configuration and what other players might be doing (it's not a perfect-knowledge game). Adding luck to it certainly reduces the strategy and makes it more approachable for beginners. There are lots of good reasons to do that, for sure. But games with no randomness have a long tradition, too, so it seems worth exploring.

 

This could be difficult if it's a war strategy with both sides being equal & having ranged attacks. Without randomness they will always be able to hit each other without ever missing so there is a distinct first player advantage. If players get an attack or defence bonus from positioning then players will camp those spots, if they get support bonuses when grouped then players will just move their people around in a group.
You could possibly look at simultaneous turn based play. Maybe players have a pool of action points that they can spend on actions for their troops however they want, with different actions costing different totals. Players plan out their actions, trying to anticipate what their opponent will do, submit, & when all players have submitted the moves play out simultaneously. Frozen Synapse is the closest I have seen to using this outside of board games. With this you could have 100% success on the actions, misses etc are caused by mists img when to shoot, or not predicting the correct square that your opponent will be in when you shoot etc.

 

All true, but again, look at Chess. Both sides have the same pieces, and nearly the same starting position; any piece can take out any other piece in one hit. In a battle, the first player to move wins. Yet the game has not suffered for these design flaws.

That's kind of another issue, but it is also something we've talked about a lot, because we want to support up to 12 players... and if it was strict turn-taking, it would take much too long! So for a while we planned exactly the sort of orders-now, execute-all-at-once solution you describe. But I was never fully comfortable with it, because there's something much more visceral and engaging about moving your units around as compared to just telling them where to move, and then watching them actually do it sometime later.

So, we're now planning a sort of hybrid: your units move when you move them, and you can mostly do this in any order you want, except where one of your units is within range of an enemy unit. In that case, whichever unit has the highest priority (defined by unit type for example) has to go first; any enemy units within range have to wait. I'm thinking that when cycling through your units, we would present any units that are making other units wait first. The net effect ought to be, most of the time, you can move most of your units without even noticing the priority blocks. But in a few cases, where it matters, then you and the other player would be essentially turn-taking at the unit level.

In talking with some friends who play more games than me, I learned that this solution is not new either; there are a couple of big games that do this (though the names escape me at the moment).

 

I kind of like the compute-and-lookup scenario you described at the end of your first post. While it may be difficult to calculate, it still seems simple enough that I think it would "feel" right. In my opinion strategy implies a scale at which minor random events are unlikely to matter much. But I guess that's why the sub-genre "grand strategy" was developed.

Simultaneous order resolution gets pretty tricky if the players have the option of moving units more than one step in a given turn. On a previous iteration of my fasciation with Diplomacy, we developed a knockoff game with more unit types and more complex order options (not yet understanding the game design consideration that "more" is not always "better") and multi-step got complicated fast. We ended up using a system where long-distance units (like bomber wings and rail transport) had their route pre-planned but still moved single-step.

I once wrote a Risk-like game that sought to remove the dice-roll-style randomness, but then I added in a bunch of dramatic, rare, randomized "special events" that were very popular with my players. (Now that I think about it, I probably got the idea from Sim City, it was the hot new game around the same time.)

 

Check out the combat system from Kemet (board game). It's totally deterministic, yet adds enough variety to be interesting.

Cry Havoc is another board game example of a deterministic but interesting combat system.

The trick from these two systems is to give the player more then one objective they care about. In both systems players can dedicate combat resources to area control or reducing their opponents forces. This provides an interesting decision point.

 

Sounds like Starcraft II. All units have set attack and Defense, no randomness. it can work

 

I'm late to the party and haven't read the full thread, but it seems magic the gathering hasn't been mentioned yet. I suggest to take a look at that system.

Also I could imagine it to be interesting to have pseudo randomness with foresight, meaning a unit has a sequence of RNG rolls lined up and you can see x rolls into the future for that unit and can plan ahead on how it will fight in the next combat. That could be an interesting variable to tweak too. Imagine a unit that is rather weak on average, but when roling a 20 on a D20 it gets super strong, and you can see 10 rolls into the future of that unit. You'd want to keep it in low-resistance battles till the 20/20 roll is lined up and then it could backstab a really powerful enemy that normally would be way out of its league. Or a fortune teller character, that can reveal x future random rolls of enemy units. "Dumber" brute characters like Orcs and Trolls might be strong foes 50% of the time, but lack the ability to see ahead on the RNG sequence unless you use another unit's ability to reveal it. I could imagine plenty of interesting synergies to come from this, like sacrificing a cheap unit in combat, because you know that will advance the RNG sequence of the enemy by 1, and can turn the tide for the next unit you attack it with.

 

Your No Man's Sky moment: where do I preorder this game?!?

 

Ask @JoeStrout, he needs to make it first ^^. I hear it's gonna have multiplayer and a fully procedurally generated universe, where everything is pre-seeded and every player starts on the same seed, and all the differences in how the universes of different players evolve are determined butterfly-effect-style by the choices of the player. Empires shall rise and fall by the actions of one brave soul!

 

It was calculated on the fly. You could see the amount of damage you would do to a unit before attacking:



Which I argue removes the "feels random" problem. In every case, you know exactly how much damage you will do to any enemy you can attack on your turn, even if you don't know how it's calculated. You don't know the potency of the counterattack, though. There are no critical hits.


I wonder if this is too simple, but what about just subtracting A/(A-D) * 100 percent of the max health when attacking? Then you could calculate a counterattack the same way, and repeat until something dies. This would mean that no unit could win a fight against any enemy with full health without taking damage, though, unless you added some more rules.

 

I'd love to jump in on this discussion. Going right back to the start, the problem at hand is that the randomness is creating situations that feel "unreal" rather than the randomness being a problem factor?

So if two evenly match opponents face off, the 50/50 random feels ok, but if you have a giant army fight 1 guy, the randomness allowing that 1 guy to win feels wrong?

If that's the case, should you add in some sort of non-linearity into the random? If the difference between each side is negligible then you are facing probabilities around 50% but as the gap widens the probability of the weaker side winning grows exponentially less likely?

 

If you want to simulate more human friendly randomness, then make most of the factors known to players. So 2 out of 10 wouldn't normally beat 9/10, however on this day 9/10 had seen his side losing battles earlier/hadn't fought for a while/underestimated 2/10, etc, etc. Notice those factors are not entirely random either, the player has some control if they have been observant/know how the game works, and realise the battle will be closer than it looks. Add external conditions too. It's raining, what does that mean? Low light? Perhaps some types match up well against other types, so 9 vs 2 is much more like 7 vs 4, etc.

 

I need to be brief, but Combat in 'World in Conflict' felt very non-random.

Each unit basically had a 100% chance against its targeted type.

Support killed Air.
Air killed Tanks.
Tanks killed Support & held positions.

Infantry would die so easily by EVERYTHING 99% of the time, except in certain terrain types, where they would dominate everything. I'd focus heavily on trees, but buildings could hold for a brief time.

So this was more rock-paper-scissors strategy, but where terrain would transform infantry from paper to rock to scissors units. You mixed luring arrogant players by stepping out to open terrain, then rushing into different terrain. Dominate a player enough as infantry, and they may lay waste to that building regardless of whether or not you're still there. The end of a map was often a barren wasteland of craters & destroyed buildings. Fog of War was a huge deal.


So the 'randomness' was tactical, based on ever-changing terrain, terrain destruction, and resource management (one napalm would destroy all infantry in a field along with all that terrain. One airstrike lay waste to a building. All players earned resource points for these limited attacks by beating opponents).

So combat felt very random, not knowing when they'd use their major resources, but non-random too with absolute rock-paper-scissors. entirely strategic, reactionary, tactical, and focused more on resource decisions than anything else. Use your napalm on that infantry cluster, and you have to handle those tanks or aircraft yourself. Make a great decision to carpet bomb an entire army, and you could earn more resources than you spent.

Whatever your unit (role), you had ONE job. Take out ONE type of Unit. That's it.

Just an idea. By splitting combat into two parts - unit management (infantry, tanks, artillery, and aircraft) & limited but extremely powerful support strokes (from napalm to nuclear bombs). It can be as predictably rock-paper-scissors as you want, but giving options that make it feel better than that. So more like Rock-Paper-Scissors-Nuke.

To this day, I don't even know if combat was random in any way. Probably was. But didn't seem to matter. Enemy Player's decisions were the ultimate random factor in combat, because you'd battle in fog of war to make players waste their major resources (major strikes) while hitting hard with your own resource use. The actual unit combat was important, but very very simple & you'd know who won the moment you encountered one another (unless a virtual tie, such as two groups of the same tanks. then you're guaranteed mutual destruction.) But a solid resource use (precision bomb strike) could instantly turn the tides of battle. if it hit. Convincing a player to bomb a forest where you had only one unit, but they thought you had many, was a major victory because they wasted those resources.

 

I believe World in Conflict's rock-beats-paper method is more realistic than "Sniper has 1% chance to beat Tank."

Humans invent weapons to beat other weapons/armor. And when that happens, there is only victory for the weapon when targeting what it was designed to beat. Otherwise it wasn't designed well.

Tanks don't shoot down aircraft. Anti-Air missiles don't target tanks. Helicopters dont out manuever anti air missiles. But carpet bombs, chemical attacks, incredibly fast aircraft strikes, drones, nukes: these destroy their targets without mercy, at insane speeds that no human can counter even with advanced technology. Except when that technology was made specifically to counter that weapon.

 

I'm going to have to check out that World in Conflict... it sounds like my kind of game!

Thanks for the thoughts, everyone!

 

I loved that game so much, and was always surprised no one ever made a clone of it.
Unfortunately, it was (basically) a multiplayer-only game, so once the community died...well, I'm sure it's long since dead. Even when I tried to find out if it was still alive, I had lost all my keys (I owned 4-5 copies of the game, since their horrible DRM forced keys upon install & I kept losing mine).

It's also very action-y. I left out a few components I didn't think were relevant to the design. Just letting you know, the game may be very different than I described due to the parts I didn't mention. Like the real-time aspect or how players only control a single type of unit (except in singleplayer? If that is even a thing. I don't remember, but if there was singleplayer I think I might try to play through it.)

 

Sadly it was apparently Windows-only, and no, I'm not going to fiddle around with Wine trying to get it to run on my Mac. Not when I have my own games I should be working on.

But if it's that good, maybe you should consider making a clone of it? And then you could make a Mac build, and I could play that.

 

Virtual machines are your friend.

 

This says it has single player... might have to check this one out!

https://www.gog.com/game/world_in_conflict_complete_edition

 

Damn this thread is long. But since Windows decided to boot my machine for me, I'll respond on my phone. I also haven't read every post (yet?). But the Advance Wars attack formula is available somewhere on the Internet. And my game now is currently using it unmodified right now, while I get the rest of my game architecture set up. While AW combat damage values are precalculated, this is only using base damage and tile defensive values. At the time of attack, there is a small random value added to the attack value. Since my damn machine is taking its sweet time? I'll try to find it.

Edit:
Heh
http://awbw.wikia.com/wiki/Damage_Formula

 

Just stumbled upon an automated solution (uses wineskin).

http://portingkit.com/game/594

Still wine-based though so it might be hit or miss to how well it runs.

 

I know there was a singleplayer campaign and I finished it back in the days. I don't remember many details, but I liked the game. Never played the multiplayer though.

 

Still a very interesting thread.

Right now I am about to prototype my turn based polyhedral dice rpg-like system for Traveler's Tale.

Polyhedral dice are extremely random. 1d4 could beat a 1d12 even though the 1d4 represents a newbie while the 1d12 represents an expert. Although I also like the idea of grandmasters being 3d4 (more consistent averages and minimums to reflect wisdom and extreme experience) while more wildly strong or lucky characters (ex. Reckless Barbarian strength) being 1d12's.

The original idea being that players roll their dice and try to beat a target number, or dice vs dice conflicts.

Since polyhedral dice are so cool and "PnP"-like, I really didnt want to change that part. So I am having to work backwards in design by adding in features which UNDO that randomness.

Literally removing the randomness in some cases, rather than designing a system with less/no randomness.

So far I have a few design changes to do this

• On all but hardcore difficulty, show the player the Target Number (TN) to beat when applicable. Let them know what to expect before rolling. On easy difficulty at least, show the enemy dice too for dice vs dice Conflicts (You vs Goblin's 3d4)
• Items, Abilities, and Risk Management to increase a roll before and after the fact. Miss the TN by 3? Use a rope, potion, or break a weapon to raise your number by 3+. Miss it by too much? Accept the loss and try again or use that lucky item for a reroll or added dice.
• Risk Management as a huge factor in decisionmaking. Players need more control than just rolling random dice. You control a large party of characters. Each committed character gets to add their dice to the roll, but also risks harm. Non-contributors can't be harmed by another character failing.
• Upon losing a round in a conflict like combat, the lowest dice rolled (character) is chosen to be harmed. So 1d4 non-combat types placed in a combat conflict get you extra 1d4 dice BUT are most likely to get themselves killed. Which fits combat, where alot of training is defensive and how to NOT get killed. However even an expert with 1d12 can get hurt by rolling lower than everyone else - it is just less likely.
• Fumbles for any character who rolls a 1, no matter their dice. A fumble in combat means the character is harmed even if you succeed.
  
• In this way, the player can roll as many dice as they want (or have) but the more they risk (and the lower the dice) the higher the chance of harm. This is to encourage more experts and less 1d4 unskilled persons participating in conflicts.
• Levels of Success & Failure. In combat? Barely winning by 1 just hurts them. Good sized win pf +5 is a severe wound. Large win pf +10 kills them instantly (3 wounds = death). This way a player could discard a dynamite item to add +5 to their roll, turning a wound into a kill after-the-fact or the explosion of an item could add +1 wound level to all participants, scratching the entire party but killing the enemies.
• Possibly a feature for borrowing extra points above the success TN to boost or prevent harm/fumbles. (ex. TN is 10, roll 16, can trade multiple points to cancel a fumble or to rescue a character who failed their roll.)
• Possibly Special traits per classes. (ex. A warrior can reverse a combat fumble or protect a character from fumbles before rolling)
• Multiple conflicts simultaneously. For example 3 combat conflicts resulting in 3 duels, rather than one huge pile of dice vs pile of dice. Conflicts which fit the narrative. (A big war battle conflict would be All dice vs All dice. But another more skirmish-y battle could have 3 individual conflicts like defending a gate, defending the tower, and protecting the children - each having narrative success/failure after the battle. Player picks which characters goes to (and rolls) each conflict before any are rolled.)

The idea being the more player control & decisionmaking the less randomness from the uber random dice rolling. Risk management. Item usage. Special abilities. Crit & Fumbles. Narrative based conflicts / rounds. Multiple conflicts simultaneously splitting characters up. Etc.

Anything I can do to give more player control and less randomness in a polyhedral dice based game.

Also something I am going to try really hard to do, since the odds are players will constantly fail conflicts throughout my game even with great dice due to chance - I want to make failure a big part of the adventure.

Spoiler

"Losing is Fun" IMO isnt true for permadeath games. Starting over is fun in a roguelike, but losing still sucks. This is ESPECIALLY true in randomized games or text driven games. (In Neo Scavenger, the designer decided it was a good idea to do the horrific bad decision to allow text based choices to instantly kill your character. In a permadeath game. It is so stupid...)

Instead I want to really own that mantra "Losing is Fun" in all my games. In all my games I want to turn losing into additional adventures or "side missions" so the player has just as much (or more) fun losing a conflict as they do winning.

For example instead of reloading a save, have the player wake up in a jail cell and have to escape.

Ex2. Instead of character death, have them become a raised zombie who completes a side mission to break free of their necromancer to continue their original quest as the undead. This is just plain rewarding if a zombie gets a bunch of positive traits and stat bonuses.

Ex3. Upon defeat by bandits have the character just lose their gold and epic weapon but then have a sidequest for revenge with backup gear or the option to just continue onwards with a revealed luck bonus - their choice. Torture and stalk those jerks for revenge one by one, just knock them out peacefully given their desperation/starvation, or kill em all for their loot as well as yours! Alot of ways to go narratively.

Losing is so common in games, especially on high difficulties, that IMO a good portion of the content should go into making losing part of the game itself.
With enough variety and always giving the players options (like skipping already played content) I think it would be a good decision. Alot of good story is birthed from failures (ex. Dwarves caught by spiders, freed by Bilbo wielding Sting. What if you were normally the dwarves but only if defeated in combat had to play as Bilbo? If then lose again as Bilbo, play as a super OP Gandalf or Radaghast?

 

*Didn't read though all the replies so this may have already been suggested

How about making the values of A and D squared(or even cubed) to make the strength curve non-linear?

The difference between 10 and 5 is 5( A/(A+D) = .666.... ), but if squared you get ( A/(A+D) = .8 ), cubed you get ( A/(A+D) = .888.... ) and so on.