RTS Battle Of The Sands (from KoTC developer)

[Damned Registrations]

WC3 units don't deal significantly random damage. There's more variance in wasted shots from dragoons hitting dead targets in starcraft. Units dying in 6 hits instead of 7 sometimes is pretty much irrelevant, and certainly not the kind of randomness that requires one to adapt a strategy. No amount of RNG will make 10 units die to 5 of the same in WC3.

What Draq wants is units dying in a single shot. Or evading attacks entirely. At random. And when I said this would make for a shit PvP game because rolling 1d(unit_dies) over and over again is too fucking random, you tried to argue that if you just repeated it enough it'd balance out. Which is fucking retarded. And you tried to argue that adding more dice rolls to an existing roll wouldn't make it more fucking random, and then went back on it after I showed you were wrong like 5 different fucking ways and tried to pretend I was arguing against a strawman when the thing I fucking quoted was less than a page back.

 

[Average Manatee]

What Draq wants is units dying in a single shot. Or evading attacks entirely. At random. And when I said this would make for a shit PvP game because rolling 1d(unit_dies) over and over again is too fucking random, you tried to argue that if you just repeated it enough it'd balance out. Which is fucking retarded. And you tried to argue that adding more dice rolls to an existing roll wouldn't make it more fucking random, and then went back on it after I showed you were wrong like 5 different fucking ways and tried to pretend I was arguing against a strawman when the thing I fucking quoted was less than a page back.

Draq never said units evading attacks entirely, just the possibility to die in a single shot. If we assume that the rest of units die in two shots, only taking the attacks the units are specially vulnerable to, there is little statistical deviation from the mean time to kill for a decent population. Certainly nowhere near enough for 5 units to kill 10. And you haven't even asked how many units each side would have. If you get to the point of a 20v20 or so, no amount of damage randomness you can add can hope to sway the law of large numbers.

Lets pretend we set up a scenario where units have 1 HP and deal either .5 or 1 damage per shot, shooting simultaneously at the opportune target with no overkill. Just to see if you have any concept of how statistics work, give me your guess of how often a 5 unit group would beat a 7 unit group. I've already done the simulation.

 

[Damned Registrations]

Gee, I wonder why you'd assume something totally fucking weird and unrandom like perfectly efficient attack assignments that don't make any sense in an rts at all, where units won't even be in range to achieve such a thing. Let me guess: you already ran the numbers for that scenario and went 'Oh shit, he's totally right!'

 

[CappenVarra]

Uhm... Take a deep breath and relax, dude - this is nothing to get so worked up about.

Let's say Unit A has 50% chance to kill Unit B.

Given 1 attack attempt, whether Unit A kills Unit B will deviate significantly from that idealized probability: it will either have a 100% observed kill luck (it hits on that first attack) or a 0% kill luck (it misses that single attack).
Given 2 attack attempts, the observed "luck" will be more evenly spread: 25% chance of 100% kill luck, 25% chance of 0% kill luck, 50% chance of 50% kill luck.
Given 1000 attack attempts, the actual observed "kill luck" is expected to be much closer to that idealized 50% chance to kill.
Given N attack attempts, the observed "kill probability" will be closer to the ideal 50% chance to kill as N grows.
Thus, increasing the number of tests (units shooting at each other during a long game) reduces the "pure luck" (= deviation from the expected outcome).

Isn't this what DraQ was talking about?

Of course, the system designer sets the expected probability, as well as the upper and lower bounds of every random range. And of course the system designer can fuck things up if he shifts the lower and upper bounds around like a moran - but what has that got to do with DraQ's idea?

 

[AMG]

It doesn't matter that with increasing number of attempts the effect of "randomness" is diluted because games exist in context. Even if hit/miss ratio between the players is more or less even over the course of the game, lets say that in the final battle of 20 tanks against 20 tanks one side hits with every tank and all other miss. Game was won pretty much with luck even if statistics say that both side hit 50% of the time over the course of the game.

I played long ago open source game Battle for Wesnoth or something like that. It is free open source game, it is also turn-based but whatever. The point is that fighting mechanics are simillar to what you are talking about here. Every unit has evade chance depending on a terrain it is standing from 20% to even 70%. Most units also attack multiple times during one turn, and you need to hit one unit like 6 times on average to kill it. Even with low lethatlity and multiple attacks the games were often won and lost with lucky turn(which was the cause of delicious butthurt by the some people). It even has a neat statistic where it calculates expected damage you should do that turn and compares it to empirical data. And yes almost always at the end of the game there is maximum +/-5% deviation from expected outcome, usually less.
The bottom line is that high lethality plus random chance is a recipe for fail, unless the game is played at massive scale, but those games are dumb and it doesn't matter what kind of mechanics are in those.

 

[Johannes]

You really don't get bell curves, Damned Registrations. Most of these examples are simply so insane that I can't directly answer to them in a sensible fashion. Sure they have edges, but unlike in a linear distribution you're really unlikely to hit those edges. I'm not completely sure what exactly you're arguing overall, but I'm sure it's crazy.

Basically what you're saying is "more dice" = "more randomness". If you there's a constant like in d6+100 compared to 2d6+100, sure. But that's simply because the ratio of the dice rolls compared to the constant is different. d6 is still more random than 2d6, the edge values in 2d6 are much rarer than in d6.




If you roll D&D characters with 3d6, you'll get much more average guys than if you'd roll them with d16+2. Obviously.

 

[CappenVarra]

It doesn't matter that with increasing number of attempts the effect of "randomness" is diluted because games exist in context. Even if hit/miss ratio between the players is more or less even over the course of the game, lets say that in the final battle of 20 tanks against 20 tanks one side hits with every tank and all other miss. Game was won pretty much with luck even if statistics say that both side hit 50% of the time over the course of the game.

I know this is just an example, but (0.5)^20 = 0,00000095367431640625. And that's just the chance that all 20 Side A tanks hit. Since the chance that all 20 Side B tanks miss is the same (since we're sticking to the idealized 50% chance to hit), the overall chance of the scenario you're using as some kind of argument is 9,094947017729282379150390625e-13. Really?

 

[AMG]

I know this is just and example but i will do pointless calclations anyway. Exchange 20 with 12 nad 0 with 8 you happy now? Or a super important unit got destroyed by lucky hit and if it survived i would have won. Or blah blah blah... The point is obvious, that exchanges of hits doesn't occur in some simulation but in a real game where a streak of luck can change the outcome even if it is miniscule in the grand scheme of things. You play RTS with incomplete knowledge anyway, so it's not that they need more random factors.

 

[Jashiin]

Now I feel like playing dune 2000. Probably a hassle to run on w7 though

 

[Johannes]

Battle with 20 tanks is not the issue. The sample is big enough that luck won't probably affect too much - yes, one side will come on top in an even battle with some tanks left, but that's totally expected when one enters such a fight.

But if you've got a battle with say, 3 tanks on both sides early on in the game. There you're rolling so few dice that a lucky streak in favor of one side might actually decide the game.

 

[Average Manatee]

Gee, I wonder why you'd assume something totally fucking weird and unrandom like perfectly efficient attack assignments that don't make any sense in an rts at all, where units won't even be in range to achieve such a thing.

It's very, very simple to simulate this. Things like perfect attack assignment isn't out of the question (SC2 already does this), and it makes the simulation easy. It also doesn't bias the calculations either for or against one side, so it's hardly a disruptive simplification. You should understand this.

Let me guess: you already ran the numbers for that scenario and went 'Oh shit, he's totally right!'

No, you are right about absolutely nothing so far. You have done nothing but vomit irrelevant filth on the thread. If you had any pertinent knowledge it would be simply to check this, but as everyone else in the thread knows, you don't.

You want to know what the chance of 5 tanks beating 7 are? 0. Or rather .00001. As in, a statistical impossibility in any real sense. The micro a player does will be a thousand times more important than random luck.

Go play the lottery or something.

function battle(group1hp, group2hp, numbattles)
{
    basedamage = .5
    randomdamage = .5
    damagechance = .5
    wins = [0,0,0]
    for(i = 0; i < numbattles; i++)
    {
        current1hp = group1hp
        current2hp = group2hp
        while(current1hp > 0 && current2hp > 0)
        {
            shots1 = Math.ceil(current1hp)
            shots2 = Math.ceil(current2hp)
            for(j = 0; j < shots1; j++)
                current2hp = current2hp - basedamage - (Math.random() < damagechance && randomdamage)
            for(j = 0; j < shots2; j++)
                current1hp = current1hp - basedamage - (Math.random() < damagechance && randomdamage)
        }
        if(current1hp > 0)
            wins[0] += 1
        if(current2hp > 0)
            wins[1] += 1
        if(current1hp <= 0 && current2hp <= 0)
            wins[2] += 1
    }
    return wins
}

battle(7,5,10000000) returns (9999030, 154, 816)

 

[Damned Registrations]

Yes, your example in which units all target flawlessly in a way less random than SC (where units target the nearest unit by default, and everything with a projectile tends to waste shots on dead units) is less random. Shocking.

On the other hand, if every unit has it's own target, all of the sudden 5 soldiers can kill 5 enemies while 7 soldiers kill only 2 in the first salvo. Unlikely, but possible, and overkill. And if it's a scenario that gets repeated a dozen times over the course of a game, it becomes a lot more likely.

But you think units in SC2 have infinite range and smart target the lowest hp units, so this isn't even worth continuing. Just keep constructing more bullshit based on false statements and pretend I give a fuck.

 

[Johannes]

Yes, your example in which units all target flawlessly in a way less random than SC (where units target the nearest unit by default, and everything with a projectile tends to waste shots on dead units) is less random. Shocking.

On the other hand, if every unit has it's own target, all of the sudden 5 soldiers can kill 5 enemies while 7 soldiers kill only 2 in the first salvo. Unlikely, but possible, and overkill. And if it's a scenario that gets repeated a dozen times over the course of a game, it becomes a lot more likely.

But you think units in SC2 have infinite range and smart target the lowest hp units, so this isn't even worth continuing. Just keep constructing more bullshit based on false statements and pretend I give a fuck.

If you can't even see the point of his example you really are a lost cause. Sure, with good targeting and maneuvering in a real RTS you can potentially take out a bigger force which is controlled worse, but here's an example of a situation where there's just pure RNG randomness with a ton of dice thrown around. But no, in most any RTS where targeting and overkill plays a part, a situation that 5 tanks would win against 7 when they're both unmicroed, will also still happen just once in a blue moon.

The test is about the nature of randomness and bell curves, and didn't you earlier posit that the test results would support your position?


How about you go to the http://anydice.com/ site as linked earlier, and inspect what happens when you try different Xd6 values. Does the deviation go up, or down?

 

[Damned Registrations]

He's not testing the scenario that was brought up in the first place. Units don't perfectly distribute damage in an RTS. They attack other full hp units. His scenario doesn't even make sense. They fire simultaneously, but the shots somehow all land where they'd be most effective? What, the bullets swerved in midair to change targets to the guy that lost 1 hp already, even though the shot that did that was fired at the same time and therefore shouldn't even be possible to react to?

 

[Johannes]

It's not trying to be a real life scenario it's a demonstration of how RNG works, ffs

 

[Average Manatee]

Actually it is a real life scenario. Siege tanks in SC2 are specifically programmed to distribute their shots perfectly even if they fire at the same time. So you can't say that it just doesn't happen in an RTS. It's also applicable to TBS.

Otherwise yes, it's an approximation. Overkill and other factors would not change the outcome substantially. In fact, they would make it more in favor of the 7 tanks. The battle would end up more drawn out and the likelyhood of the smaller force winning would be even lower. So you are arguing for me to alter the simulation in order to make you MORE wrong.

On the other hand, if every unit has it's own target, all of the sudden 5 soldiers can kill 5 enemies while 7 soldiers kill only 2 in the first salvo. Unlikely, but possible, and overkill. And if it's a scenario that gets repeated a dozen times over the course of a game, it becomes a lot more likely.

Words like "unlikely, but possible" are the sign of a complete moron. You need to give some kind of statistic (hint: you'll be using a lot of zeros). Otherwise the chance of you winning the lottery is "unlikely, but possible".

 

[Damned Registrations]

Seige tanks in starcraft don't fucking smart target based on hp totals you lying sack of shit.

 

[Johannes]

Seige tanks in starcraft don't fucking smart target based on hp totals you lying sack of shit.

Ok whatever, I don't care either way but how is this gonna make your math any less ridiculous?

 

[Average Manatee]

Seige tanks in starcraft don't fucking smart target based on hp totals you lying sack of shit.

Yes they do. Instant hit projectile units coordinate and will target different units to prevent overkill. Very recognizable with siege tanks if you've actually played the game.

http://www.youtube.com/watch?v=cIDixqI0ycU
http://eu.battle.net/sc2/en/forum/topic/1820912701

why u mad tho?

still waiting for you to offer a single shred of statistical evidence backing up your claims.

 

[Raghar]

1d6 is random
2d3 is less random
3d2 is even less random
3 is the least random.

Linear distribution.
Triangular distribution.
Coin flipping.
Not random at all.

You are decreasing variance in this example. Try 3D6 clamp 7 to 13. BTW what would happen when you increase number of dices, but kept clamping random numbers?


So lets use a some real example. 497 tanks are fighting with 498 tanks. And these 497 managed to take two more during the first salvo. What would happen? Let's assume they take 1/10 tanks per salvo thus:
497 - 49 498 - 51
448 447
Now the expected random outcome is winning of the first group.
A simple math says 29 tanks would remain.

Well real world tanks are moving in groups of 4 to avoid clusterfucks, and issue orders like: from left, from front, from back. I guess, I don't have to explain why tank commander would like to shot an enemy tank in the back line first.

 

[DraQ]

Damned Registrations

(No the bounds don't work perfectly because dice are shitty and incapable of rolling 0).

Didn't stop DR from rolling 0 on his INT stat.

Well first you call RTS simple, then you are designing something even more simple.

No, I'm proposing simple fix for simple mechanics present in pretty much every RTS which happens to be wrong and lead to reduced tactical depth.

1. Inflated HPs are completely pointless in games where you are controlling fuckton of expendable, interchangeable units, so are HP bars, yet both are ubiquitous.
2. Ability to determine what can't damage what is needed, but usually absent so you can whittle tanks away with assault rifles and don't afraid of anything.
3. Attacking something with suitable weapon may or may not kill it IRL.

So I propose that units have fixed, HP counts, but small compared to damage ranges of weapons used against them, so that 1 hit kills are definite possibility. Unlike just using random rolls, unit that is damaged but survived is more likely to get killed by subsequent attacks.
For example tank may have 6HP, but antitank weapon may do 2d4 damage (some overkill against that tank type) this without considering any form of damage reduction and damage adjustment.

I also propose some way of adjusting for weapon effectiveness against given target, whether by simple x can damage y matrix, or DT, though DT would be trickier to adjust while maintaining the possibility of 1hit kills.

Battle with 20 tanks is not the issue. The sample is big enough that luck won't probably affect too much - yes, one side will come on top in an even battle with some tanks left, but that's totally expected when one enters such a fight.

But if you've got a battle with say, 3 tanks on both sides early on in the game. There you're rolling so few dice that a lucky streak in favor of one side might actually decide the game.

That's why low depth systems with a lot of randomness work better for massive armies in RTS games than for adventuring parties in RPGs.

 

[Average Manatee]

(No the bounds don't work perfectly because dice are shitty and incapable of rolling 0).

Didn't stop DR from rolling 0 on his INT stat.

He must have been rolling 7d6 - 14. Of course it happens less than 1% of the time, but you can see why he is now irrationally incredibly afraid of very low probability events.

 

[Humanity has risen!]

Even on the Codex no one has been able to play a game as obscure, niche and plain as this. When this happens you know you've screwed up.

 

[DraQ]

(No the bounds don't work perfectly because dice are shitty and incapable of rolling 0).

Didn't stop DR from rolling 0 on his INT stat.

He must have been rolling 7d6 - 14. Of course it happens less than 1% of the time, but you can see why he is now irrationally incredibly afraid of very low probability events.

I wonder how he can live under constant threat of suffocation due to all the air randomly collecting in upper left corner of his room.

 

[Raghar]

I wonder how he can live under constant threat of suffocation due to all the air randomly collecting in upper left corner of his room.

Because the air is replaced by a cold fresh air. It's called a natural air circulation. You'd need a high energy state to create low pressure area, it's called turbulence, but it will not create vaccum by definition. I probably should't talk about stuff like a supersonic shockwave propagation, but...

ITT DraQ has problems with physics.

Well first you call RTS simple, then you are designing something even more simple.

No, I'm proposing simple fix for simple mechanics present in pretty much every RTS which happens to be wrong and lead to reduced tactical depth.

1. Inflated HPs are completely pointless in games where you are controlling fuckton of expendable, interchangeable units, so are HP bars, yet both are ubiquitous.

Considering majority of RTS don't use number of amounitions, you wouldn't believe what one or two of these "expendable" units can do. And if it can evade swarm of enemy units, it can do it again.

For example tank may have 6HP, but antitank weapon may do 2d4 damage (some overkill against that tank type) this without considering any form of damage reduction and damage adjustment.

So 6 infantries with 1 damage would kill it in one second...

I also propose some way of adjusting for weapon effectiveness against given target, whether by simple x can damage y matrix, or DT, though DT would be trickier to adjust while maintaining the possibility of 1hit kills.

Hardcoding imunity against weapons is invitation to bug fest and unrealistic.

What's hard on "enemy needs at least 30mm DU to penetrate back armor on 90 degree angle"? Well real world tanks don't have one hit kills in majority of situations, as long as they are not Russian "it popped the tower" tanks. Why are people thinking when user would need to protect his own hide he will not do anything stupid like leaving lying ammo on floor, or stuff like that?