Combat Simulations: Grunt vs. CK

[DocS]

Heck, I didn't understand half of the terms you used, and I know next to nothing about statistical tests.

All I know is that in 3 out of 5 of my test combats, the Cyber-Armor bought at least one extra round of survival.

The stat tests tell you if having the armor is worth an extra 1 MD, or extra 5, or extra 10 (etc). And to be honest, looking at the variation, the armor is only slightly more useful than prayer.

[DocS]

In any case, this goes pretty well with what I've always said; the psi-sword isn't meant to be the CK's main weapon.

I'm calling the psi sword insignificant (and we may parse the differences between 'insignificant' and 'pretty much useless). And now we have the data to show it. Insignificant, a non-factor, as in, from here on, all discussions of the psi-sword should be in light of these results..

If the cyberknight it at a point where he's out of ammo, he's assuredly at a point where his armor is so beat up that he should be running away anyway, after all, by that point, he's already gotten lucky enough to have is cyber armor last to the point where he has run out of ammo.

"Weapon to be feared"... HA!

[DocS]

When we're done with this set of tests though, I think we should roll up some weaker monsters and run more battles.
The psi-sword will likely come into play more when ammo's in short supply.

The irony is here... the weaker the monsters, the less use the cyber armor will be (low MD damage rolls will mean less damage will be done before the monster beats that all important 16, killing Sir Robin). With this monster, the cyber armor already was largely insignificant.

[Killer Cyborg]

Heck, I didn't understand half of the terms you used, and I know next to nothing about statistical tests.

All I know is that in 3 out of 5 of my test combats, the Cyber-Armor bought at least one extra round of survival.

The stat tests tell you if having the armor is worth an extra 1 MD, or extra 5, or extra 10 (etc). And to be honest, looking at the variation, the armor is only slightly more useful than prayer.

Once again, 3 out of 5 times it bought an extra round.
Prayer didn't help at all the one time Robin used it.

[Killer Cyborg]

In any case, this goes pretty well with what I've always said; the psi-sword isn't meant to be the CK's main weapon.

I'm calling the psi sword insignificant (and we may parse the differences between 'insignificant' and 'pretty much useless). And now we have the data to show it. Insignificant, a non-factor, as in, from here on, all discussions of the psi-sword should be in light of these results..

For the context of these battles, sure.
But the game of Rifts overall is more than just endless battles between decently powerful monsters and fully equipped PCs.

If the cyberknight it at a point where he's out of ammo, he's assuredly at a point where his armor is so beat up that he should be running away anyway, after all, by that point, he's already gotten lucky enough to have is cyber armor last to the point where he has run out of ammo.

In my group, we have often had to decide between spending our money on armor repairs, or on ammo.
It's not a safe assumption to believe that PCs always (or even typically) enter into battle fully armed and armored.
For that matter, we've often made use of captured armor and/or weapons and ammo.
Depending on what we've been fighting, we might end up with lots of spare armor and no weapons, or vice versa.
Obviously, the Psi-Sword wouldn't be of much use if you have fully loaded guns, but if your guns are empty and your armor is full, it could come in handy.

And while we're speaking of extra armor, keep in mind that the CK starts off with 2 suits of armor.
For the purposes of this exercise we're only using the heavy armor, but in an actual adventure he might very well have time to change into his other armor between encounters, even if he's low on ammo.

"Weapon to be feared"... HA!

They're certainly fearsome to SDC creatures... whom the CK is pledged not to use the sword against.

Still, at higher levels it's a pretty decent weapon.
At 6th level a Psi-Sword can carve up somebody in Plastic Man or Huntsman armor in 3-4 hits on average.

[Killer Cyborg]

When we're done with this set of tests though, I think we should roll up some weaker monsters and run more battles.
The psi-sword will likely come into play more when ammo's in short supply.

The irony is here... the weaker the monsters, the less use the cyber armor will be (low MD damage rolls will mean less damage will be done before the monster beats that all important 16, killing Sir Robin).

I don't think that the amount of damage the cyber-armor absorbs is as important as how many hits it absorbs, because that's how many times it saves the CK's life.

With this monster, the cyber armor already was largely insignificant.

Obviously, we disagree on this.

[DocS]

I don't think that the amount of damage the cyber-armor absorbs is as important as how many hits it absorbs, because that's how many times it saves the CK's life.

Is that in comparison with anything? Right now, the armor looks like it's not even significantly better than nothing. And that's not a good endorsement. I bet if we did 20 or 30 trials, we'd find the armor to be, *slightly* better than nothing (as in, a CK with 70 MD armor and cyber armor is about as well protected as a normal person in 72 MD armor),

This makes it roughly comparable to the +1 on initiative...

Yes, slightly better, but so slightly that you have to work hard to detect it, which makes it, not really worth the effort at all. Which, fundamentally, is one of the major problems in Rifts. They give you an ability (like the psi sword) that seems cool, but when push comes to shove, it doesn't really matter.

(on the significance of cyber armor)

Obviously, we disagree on this.

However, one of us knows how to determine significance. How to analyze it, how to quantify it, and how to discuss it. This isn't a "agree to disagree" thing, significance of data in a model has a definite quantitative meaning. If you don't *know* the meaning, then please look it up, and learn it.

If you don't want to learn it, then cede to someone who does know it.... Either path is good.

However 'disagreeing' on it is a little like 'disagreeing' on the average number of monsters the CK killed. Yes, you can disagree, but saying "I disagree on the average... but I refuse to learn what an average is..."

After all, I will use the same tests of significance on the differences between the CK and Grunt performances. This is why I wanted FIVE trials (so then I can do all these tests and separate blind chance from actual difference). I will use them on my model too. Significance, in this matter is *not* a matter of opinion.

[Nekira Sudacne]

I don't think that the amount of damage the cyber-armor absorbs is as important as how many hits it absorbs, because that's how many times it saves the CK's life.

Is that in comparison with anything? Right now, the armor looks like it's not even significantly better than nothing. And that's not a good endorsement. I bet if we did 20 or 30 trials, we'd find the armor to be, *slightly* better than nothing (as in, a CK with 70 MD armor and cyber armor is about as well protected as a normal person in 72 MD armor),

This makes it roughly comparable to the +1 on initiative...

Yes, slightly better, but so slightly that you have to work hard to detect it, which makes it, not really worth the effort at all. Which, fundamentally, is one of the major problems in Rifts. They give you an ability (like the psi sword) that seems cool, but when push comes to shove, it doesn't really matter.

(on the significance of cyber armor)

Obviously, we disagree on this.

However, one of us knows how to determine significance. How to analyze it, how to quantify it, and how to discuss it. This isn't a "agree to disagree" thing, significance of data in a model has a definite quantitative meaning. If you don't *know* the meaning, then please look it up, and learn it.

If you don't want to learn it, then cede to someone who does know it.... Either path is good.

However 'disagreeing' on it is a little like 'disagreeing' on the average number of monsters the CK killed. Yes, you can disagree, but saying "I disagree on the average... but I refuse to learn what an average is..."

After all, I will use the same tests of significance on the differences between the CK and Grunt performances. This is why I wanted FIVE trials (so then I can do all these tests and separate blind chance from actual difference). I will use them on my model too. Significance, in this matter is *not* a matter of opinion.

Hey, if it saves your life even ONCE in a situation where you are not immediately assuled by another monster, that's significant.

Never get so wrapped up in numbers you loose sight of the facts.

[Talavar]

I don't think that the amount of damage the cyber-armor absorbs is as important as how many hits it absorbs, because that's how many times it saves the CK's life.

Is that in comparison with anything? Right now, the armor looks like it's not even significantly better than nothing. And that's not a good endorsement. I bet if we did 20 or 30 trials, we'd find the armor to be, *slightly* better than nothing (as in, a CK with 70 MD armor and cyber armor is about as well protected as a normal person in 72 MD armor),

This makes it roughly comparable to the +1 on initiative...

Yes, slightly better, but so slightly that you have to work hard to detect it, which makes it, not really worth the effort at all. Which, fundamentally, is one of the major problems in Rifts. They give you an ability (like the psi sword) that seems cool, but when push comes to shove, it doesn't really matter.

(on the significance of cyber armor)

Obviously, we disagree on this.

However, one of us knows how to determine significance. How to analyze it, how to quantify it, and how to discuss it. This isn't a "agree to disagree" thing, significance of data in a model has a definite quantitative meaning. If you don't *know* the meaning, then please look it up, and learn it.

If you don't want to learn it, then cede to someone who does know it.... Either path is good.

However 'disagreeing' on it is a little like 'disagreeing' on the average number of monsters the CK killed. Yes, you can disagree, but saying "I disagree on the average... but I refuse to learn what an average is..."

After all, I will use the same tests of significance on the differences between the CK and Grunt performances. This is why I wanted FIVE trials (so then I can do all these tests and separate blind chance from actual difference). I will use them on my model too. Significance, in this matter is *not* a matter of opinion.

Rather than assessing the cyber-armour's significance by the amount of MDC it added on average, look at how many times the cyber-knight's life was extended by having it, which was 3 out of 5 times. 60% of the time, apparently, cyber-armour will have some use.

As to the psi-sword being irrelevant, of course it was. You're using a level 1 character; it does 1d6. A mind melter with psi-sword would be irrelevant at level 1 because they can't even take the power that early. Any available melee weapon is going to be fairly irrelevant for a level 1 character, as the best they could have is a 2d6 vibro-sword. This hardly proves the psi-sword to be useless for all further discussion of cyber-knights.

[Killer Cyborg]

I don't think that the amount of damage the cyber-armor absorbs is as important as how many hits it absorbs, because that's how many times it saves the CK's life.

Is that in comparison with anything? Right now, the armor looks like it's not even significantly better than nothing. And that's not a good endorsement. I bet if we did 20 or 30 trials, we'd find the armor to be, *slightly* better than nothing (as in, a CK with 70 MD armor and cyber armor is about as well protected as a normal person in 72 MD armor),

This makes it roughly comparable to the +1 on initiative...

Yes, slightly better, but so slightly that you have to work hard to detect it, which makes it, not really worth the effort at all. Which, fundamentally, is one of the major problems in Rifts. They give you an ability (like the psi sword) that seems cool, but when push comes to shove, it doesn't really matter.

Buying an extra round or more of survival three out of five times is significant.
I can keep repeating that, if you like, just as you can keep repeating your claim that it's not.

But I'd rather you explain how saving the CK's life (albeit temporarily) 3 out of 5 times is insignificant.

(on the significance of cyber armor)

Obviously, we disagree on this.

However, one of us knows how to determine significance.

The other of us has a good amount of common sense.
Between the two of us, we should be able to come up with a decent picture on things.

[DocS]

The other of us has a good amount of common sense.
Between the two of us, we should be able to come up with a decent picture on things.

There is a major problem, using 'common sense' to fight facts is a form of stupidity. This is why Modern scientific thought is built on facts rather than 'common sense'. That the earth is flat is 'common sense'.

For example, you *say* that the Cyber armor saved the CK's life 3 times out of five... However, close perusal of the combat leads to come interesting questions...

Sim 1

Round 9:
Robin wins init.
R: Long Burst, 2 strike, Miss.

(every other time, he waited for the monster to go first)


M: Simo (Firebolt), 6 strike, 14 MD (armor absorbs 1 MD, rest goes to Cyber-Armor, dropping it to 37 MD)

This is a fine example of why waiting was the good choice before, why did he all of a sudden become stupid?

R: Long Burst, 18 strike, 45 MD (KILL)
R: Switches to Ion Pistol

Sim 2

Round 6:
M: Firebolt, 22 strike, 14 MD (Kills Sir Robin)
R: Simo (Long Burst), 15 strike, 45 MD

(A monster shoots at him, and he doesn't try to dodge!?)

Sim 3

Round 9:
M: Firebolt, 20 strike, 12 MD (Kills Sir Robin)
R: Simo (Aimed, 14 strike, 13 MD)

(Again where is his attempt to dodge? He can simo-and be assured death, or he could dodge and have a good chance to save his own life)
Sim 4

Round 8:
R: Short Burst, 3 strike, Miss.
M: Simo (Firebolt), 10 strike, 19 MD (takes out remaining 10 MDC of Gladiator, and inflicts 9 MD to Cyber-Armor: 41 left)
R:Reload (Old clip still has 5 shots)
R: Long Burst, 14 strike, 20 MD
R: Long Burst, 6 strike, 45 MD (KILL)

(As far as I can tell, from here, Robin is no longer dodging, he's simo-attacking in, again, some sort of acquired stupidity that he didn't have before)

Sim 5

Round 9:
R: Long Burst, 5 strike, 60 MD (KILL)
M: Simo (Firebolt), 16 strike, 14 MD (Gladiator soaks 7, 7 goes to Cyber Armor, 43 left)

Round 10:
R: Long Burst, 16 strike, 80 MD (KILL)
M: Simo (Firebolt), 23 strike, 12 MD (Kills Robin)

You claim common sense, but your common sense is a cyberknight, whose armor is in tatters, who gets shot at... and choses *not* to dodge (coincidentally, those are all the hits where the cyber armor matters). I've finished the model with a Cyberknight does try to dodge incoming blasts... Will do the grunts tomorrow. But that common sense makes a BIIIG difference on the utility of the cyber armor.

My favorit is sim 4, round eight, not only does the cyberknight change to a bizarre tactic after round eight, but he's tagged by an eminantly dodgable monster strike of 10. I'd think it funny, if it wasn't so obviously stilted to make the cyber-armor more affective than it is (he's willfully doing the stupid thing to try to make his armor look effective when, in reality, a good portion of those blasts are easily dodgable).

Please redo with a cyberknight who says "I've almost no armor left, and a mega-damage blast is coming at me from a creature with only one attack per melee.... I think that is worth a dodge"....

[DocS]

Heck, I didn't understand half of the terms you used, and I know next to nothing about statistical tests.

You should either learn them or never ever enter any sort of analysis that involves numbers that have any variability in them.

[Killer Cyborg]

There is a major problem, using 'common sense' to fight facts is a form of stupidity.

Just as using numbers to fight facts is a form of stupidity.
I guess nobody is perfect.

You claim common sense, but your common sense is a cyberknight, whose armor is in tatters, who gets shot at... and choses *not* to dodge (coincidentally, those are all the hits where the cyber armor matters). I've finished the model with a Cyberknight does try to dodge incoming blasts... Will do the grunts tomorrow. But that common sense makes a BIIIG difference on the utility of the cyber armor.

My favorit is sim 4, round eight, not only does the cyberknight change to a bizarre tactic after round eight, but he's tagged by an eminantly dodgable monster strike of 10. I'd think it funny, if it wasn't so obviously stilted to make the cyber-armor more affective than it is (he's willfully doing the stupid thing to try to make his armor look effective when, in reality, a good portion of those blasts are easily dodgable).

Please redo with a cyberknight who says "I've almost no armor left, and a mega-damage blast is coming at me from a creature with only one attack per melee.... I think that is worth a dodge"....

Nope.
You may not like the tactics, but once his main armor is gone the best bet is to start simo-attacking, trying to kill the creature before he can kill the CK.
The Cyber-Armor provides a 60% chance of keeping the incoming Fire Bolt from killing the CK.
Which means that there's 60% chance that IF the CK spends an attack dodging, that the action will be effectively wasted.
Once things get down to that point, simo-attacking seems like the better option, since it isn't as likely to be a wasted attack.

But feel free to use whatever tactics you wish in your own simulations.

[Killer Cyborg]

Heck, I didn't understand half of the terms you used, and I know next to nothing about statistical tests.

You should either learn them or never ever enter any sort of analysis that involves numbers that have any variability in them.

I don't buy it.
While I have a lot of respect for math, I've seen too many people try to use their knowledge of numbers and jargon as a bludgeon to win arguments where they're wrong.
So no, I'm not going to sit back and blindly accept everything you say simply because you know some terms I don't.
I don't trust you not to skew things, just as you wouldn't trust me if positions were reversed.
But nice try.

[DocS]

Nope.
You may not like the tactics, but once his main armor is gone the best bet is to start simo-attacking, trying to kill the creature before he can kill the CK.
The Cyber-Armor provides a 60% chance of keeping the incoming Fire Bolt from killing the CK.
Which means that there's 60% chance that IF the CK spends an attack dodging, that the action will be effectively wasted.
Once things get down to that point, simo-attacking seems like the better option, since it isn't as likely to be a wasted attack.

But feel free to use whatever tactics you wish in your own simulations.

I'm waiting for the number which is the probablility of the CK dodging the attack successfully...... Can you calculate that number? I want to see exactly how mathematically challenged you are.

Also, calculate the chances of the cyberknight surviving the blast if he dodges (and cyber armor) vs if he doesn't dodge (cyber armor alone).

Those two numbers will truly show how stupid simo-attacking is under these conditions (I've already calculated them, lets see if you can).

Also, how often does the CK end the round with attacks to spare (even if he dodges)? Cmon, stat boy, lets see them numbers. Your CK is gambling, does he know the odds?

I don't buy it.
While I have a lot of respect for math, I've seen too many people try to use their knowledge of numbers and jargon as a bludgeon to win arguments where they're wrong.
So no, I'm not going to sit back and blindly accept everything you say simply because you know some terms I don't.
I don't trust you not to skew things, just as you wouldn't trust me if positions were reversed.
But nice try.

That's the sad part, I was trusting you not to skew things, then you pulled the bit with the simo-attacks.

"A lot of respect for math"... try losing the respect and actually learning it.

This is why you should *learn* about what I am doing..... I'm not asking you to *buy* it, I'm asking you to *understand* it. This is the true difference between intelligence and stupidity. The stupid say, "you're fooling me, I don't know how, but I don't trust what you say"

The smart say, "Here's what you did, here's where it comes up well, here's where it comes up short". But it's based on understanding. I understand how you're looking at the combats, and I understand that you're not asking one simple question, "how much of what I'm seeing is due to truth, and how much is due to random variation of the dice?".

-Answer the above question.


I can answer it (that's what the math is for). If you can't, then literally any conclusion you make is suspect (I can just say 'they were lucky rolls'). We're working with dice, and you've yet to provide *Any* way to account for the fact that dice are variable. If you didn't trust me not to skew things, you'd ask how I did my calculations, you'd replicate them, and make sure I was keeping things on the level.

You've decided, a priori, "I don't know what he's doing, but he's wrong". What happens if I'm actually not skewing things? You refuse to understand, you've decided that no matter how I analyze, I'm wrong, you're not giving any other substantive analysis. What if I'm better at data analysis than you? What if I literally spend my days doing experiments and analyzing data for significant differences? That would be something for a guy to jump and scream that "You're trying to fuzzy things with math" when I've literally done this sort of thing with people all of the world and everyone who actually understood what I was doing had no problems with my technique.....

After all, you have a lot of respect for math, but not enough to learn it or to have it used or to even accept any conclusion made mathematically.. where is this respect again?

[Killer Cyborg]

Cmon, stat boy, lets see them numbers. Your CK is gambling, does he know the odds?

"A lot of respect for math"... try losing the respect and actually learning it.

This is the true difference between intelligence and stupidity.

Why is it that you can't have a conversation, even an argument, without making personal attacks?

[Killer Cyborg]

I'm waiting for the number which is the probablility of the CK dodging the attack successfully...... Can you calculate that number? I want to see exactly how mathematically challenged you are.

Nothing higher than Alegebra II in high school, and I've forgotten a lot of that.
You're the numbers guy, go ahead and show us the numbers.

Edit:
To my under-educated mind, it seems that there would be a 50% chance of a straight d20 rolling higher than another straight d20.
With a +4 to strike and a +3 to dodge, that seems like there would be a +5% advantage to the attacker, making it about a 45% chance to dodge.
But I never really got into probability and such, so this is likely wrong, and you will likely gloat and sneer about it.
But what the hell, it's worth a guess.

I don't buy it.
While I have a lot of respect for math, I've seen too many people try to use their knowledge of numbers and jargon as a bludgeon to win arguments where they're wrong.
So no, I'm not going to sit back and blindly accept everything you say simply because you know some terms I don't.
I don't trust you not to skew things, just as you wouldn't trust me if positions were reversed.
But nice try.

That's the sad part, I was trusting you not to skew things, then you pulled the bit with the simo-attacks.

You were trusting me so much that you ran your own simulations and double-checked mine.
I'm not impressed.

What I did was use the strategy that I'd use if I was playing.
That's not skewing the stats.

This is why you should *learn* about what I am doing.....

Sorry, man. I don't have the time to become an expert in every field.

I'm not asking you to *buy* it, I'm asking you to *understand* it.

Not so far you're not.
So far you've just rattled off a bunch of crap and expected me to buy into, then tossed insults at me when I didn't.
You want me to understant?
Then quit the insults, quit the jargon, and explain your conclusions and exactly how you got them.

This is the true difference between intelligence and stupidity. The stupid say, "you're fooling me, I don't know how, but I don't trust what you say"

Oh, I didn't say you're fooling me; you're not.
I didn't even say you're trying to fool me.
What I said was that I don't trust you.

If you didn't trust me not to skew things, you'd ask how I did my calculations, you'd replicate them, and make sure I was keeping things on the level.

I already asked you to explain.
You didn't.
Any time you want to, I'm listening.

Edit:
I didn't respond to these points at first because I didn't think they were worth responding to, but (again), what the hell.

The smart say, "Here's what you did, here's where it comes up well, here's where it comes up short". But it's based on understanding.

That's assuming that a smart person will automatically understand everything they encounter, which simply isn't true.
People who are smart still don't know everything, and can encounter areas of ignorance.
In this instance, I can't say "Here's what you did...", because you have not yet made it clear to me what exactly you DID do.
You just rattled off some terms that I'm unfamiliar with.

I understand how you're looking at the combats, and I understand that you're not asking one simple question, "how much of what I'm seeing is due to truth, and how much is due to random variation of the dice?"

The question indicates that you believe the random variations of the dice are lying somehow.
But since that doesn't make sense, I'll assume you meant something else.
Although I don't know quite what.

[Nekira Sudacne]

Doc S, Killer, both of you quit with the third grade bickering, PLEASE.

Doc S: a lot of people really don't understand probability. it WOULD be a lot more conductive to the discussion if you broke down and simply showed your work with breif explinations of each advanced step.

[DocS]

Nothing higher than Alegebra II in high school, and I've forgotten a lot of that.
You're the numbers guy, go ahead and show us the numbers.
Edit:
To my under-educated mind, it seems that there would be a 50% chance of a straight d20 rolling higher than another straight d20.
With a +4 to strike and a +3 to dodge, that seems like there would be a +5% advantage to the attacker, making it about a 45% chance to dodge.
But I never really got into probability and such, so this is likely wrong, and you will likely gloat and sneer about it.
But what the hell, it's worth a guess.

Your latter assertion is correct, the calculation is wrong (this is not opinion). Particularly the +5% number (by your numbers, a +10 would give you a 100% chance to dodge, but as any juicer player will tell you, it's not that good). Now the question is, will you believe it when I give the correct answer?

For starters, there are three ways it can go. Either the guy A (shooter) can roll higher, guy B (dodger) can roll higher, or it can be a tie. There is a 5% chance of a tie. And, assuming *EQUAL* bonuses, an equal chance of higher or lower. Which means, on equal dice. 5% tie, 47.5% guy A gets higher, and 47.5% guy B gets higher.

However, ties go to the defender, so the chance of a successful dodge is 47.5+5=52.5% chance of a successful dodge on straight dice.

And, every +1 difference gives a +2.5% chance shift, now we are negating critical strikes (which become more important as the bonuses go up), but at the extreme case, when in this theory you have a 0% chance of hitting someone if you have +0 and they have +19, the reality is that you have a 4.75% chance (you roll a nat 20, and he does not). So factoring in the nat 20 can be more trouble than its worth (if you want to see it get complicated, for not a lot of gain, factor in crits).

However
+4 to strike, +3 to dodge... is an even 50/50 shot. It's almost perfect, missing is 100% out of the picture, so it's simple dodge vs strike. And the natural 20's cancel each other out, except in the Crit/crit, which shifts things slightly..

(factor that in, and the absolute number is a cyberknight has 50.25% chance of dodging, toldja factoring in crits is more trouble than its worth).

Why this is important with dodging and cyber armor. The overestimating model....

Cyber armor blocks 60% of the time (at least in this case, ) a successful dodge works 50% of the time, which means, assuming no connection between cyber armor and dodge (this is incorrect as I will talk later), Then the cyber armor will absorb ONE hit, 30% of the time (60% of the 50% that aren't dodged). This assumes that, however, the cyber armor is blocking most of the undodged shots...

That assumption is wrong.

(the real consideration)

The assumption is wrong because the shots that the cyber armor absorbs (the rolls between 5-16, are *also* the ones most likely to be dodged). Dodging and armor are *NOT* independent! This gets bad in cases when the attacker has a bonus to strike (which most attackers do), and gets even *worse* in the cases of high dodge bonuses (something, players try to maximize). Add an extra melee attack (which means the cyberknight has more attacks to spend on dodge) and sixth sense (which means surprise is incredibly unlikely *and* dodge bonuses go up by an additional +3), and what you get is a whole heaping helping of cyber armor being not that good.

To put into words (I'm not going to do the stats until I have assurances that I will be believed), "The shots that the cyber armor blocks, are shots that the knight would have dodged anyway".

As for what are the statistical tests for? dice don't lie, but they do tell a different story every time you roll them. So if you're looking for differences.... you need to factor in that a lot of differences from time to time will simply come from the dice. It's why 10D6 is different fro 1d6x10 (despite the fact that both average to 35). How do you quantify the difference? Statistical tests. It's also a pet peeve from how Palladium does stats, because the folks at PB don't seem to think about this. The books are *filled* with bonuses that don't matter (+1 on init? Does that matter?), damage values that change for no good reason (what should be 10d6 becomes 1d6x10, totally changing how the weapon is), and percentiles which are silly ("Hi, I'm a Medical Doctor who can only successfully remove tonsils 65% of the time"... THESE ARE BAD DOCTORS!).

For whatever analysis we do, we need a way to tease out truth from 'random dice variation', already we have CK's who kill anywhere from 3-10 beasties.

[Nekira Sudacne]

DocS: I don't speak for KC, but you can rest assured I belive you know what your talking about

And i'd love to see more of this stuff.

[DocS]

Doc S, Killer, both of you quit with the third grade bickering, PLEASE.

Doc S: a lot of people really don't understand probability. it WOULD be a lot more conductive to the discussion if you broke down and simply showed your work with breif explinations of each advanced step.

I did

I took the values of the Cyber armor damage, 7,50,0,0,13,14 and did a Students T-test vs no armor at all (which would have values of 0,0,0,0,0) and found a P-value of 0.1, which means the cyber armor had no significant difference in damage absorption from no armor at all.

Is there another statistical test you would prefer? I've tried Standard deviations, standard error, and they're all coming in with a "that Cyber armor is... statistically useless!". My favorite is the standard deviation, where the cyber armor is

14+/- 20MD (arguably, the cyber armor could be worth -6 MD, but that's because such a large spread stymies such analyses, which is why I did the T-test)

The T-test is a measure of the randomness in the system. I used Excel to calculate it. the P-value is a measure of the likelyhood that any result you see is due to raw chance. The lower the P value, the lesser the chances of chance being a factor. P value ranges from zero to one. a P value of 0.05 or less is the standard criteria in most places for significance (most places being science labs, medical analysis, places where data is analyzed). Though you can calculate with 3 trials, the preferred method is to have 5 trials so you can start getting a representative analysis.

Standard deviation is simply a measure of the 'spread'. It is, in short "how much, on average, does the result differ from the average". I used Excel to calculate it. If, for example, the average was 6.4, but the standard deviation was 2.6, I would say, it's 6.4 +/- 2.6 (meaning that your spread is anywhere from 3.8 to 9.0). This method tends to break down if you have a spread with isolated, abberant points, (for example, 0,0,7,13, 50), when you throw the '50' in the mix, assuming a standard distribution around a mean, there would be a -20 somehwere down the line. It also means that we don't have enough points to analyse this way.

[Mack]

Recommend everyone keep their egos and their insults out of this thread.

[Killer Cyborg]

I took the values of the Cyber armor damage, 7,50,0,0,13,14 and did a Students T-test vs no armor at all (which would have values of 0,0,0,0,0) and found a P-value of 0.1, which means the cyber armor had no significant difference in damage absorption from no armor at all.

Is there another statistical test you would prefer? I've tried Standard deviations, standard error, and they're all coming in with a "that Cyber armor is... statistically useless!". My favorite is the standard deviation, where the cyber armor is

14+/- 20MD (arguably, the cyber armor could be worth -6 MD, but that's because such a large spread stymies such analyses, which is why I did the T-test)

The T-test is a measure of the randomness in the system.

Okay.
How's it work?
What exactly does "measuring the randomness in the system" entail?

I used Excel to calculate it. the P-value is a measure of the likelyhood that any result you see is due to raw chance.

How is the P-Value determined?

The lower the P value, the lesser the chances of chance being a factor.

Isn't chance always a factor?

P value ranges from zero to one. a P value of 0.05 or less is the standard criteria in most places for significance (most places being science labs, medical analysis, places where data is analyzed). Though you can calculate with 3 trials, the preferred method is to have 5 trials so you can start getting a representative analysis.

Okay.

Standard deviation is simply a measure of the 'spread'. It is, in short "how much, on average, does the result differ from the average".

Gotcha.

I used Excel to calculate it. If, for example, the average was 6.4, but the standard deviation was 2.6, I would say, it's 6.4 +/- 2.6 (meaning that your spread is anywhere from 3.8 to 9.0). This method tends to break down if you have a spread with isolated, abberant points, (for example, 0,0,7,13, 50), when you throw the '50' in the mix, assuming a standard distribution around a mean, there would be a -20 somehwere down the line. It also means that we don't have enough points to analyse this way.

So basically, we need more data?

[DocS]

I'm here because I don't know enough math for the Enworld boards..

http://www.enworld.org/showthread.php?t ... ge=1&pp=40

But one day... if I study hard...

[Nekira Sudacne]

I took the values of the Cyber armor damage, 7,50,0,0,13,14 and did a Students T-test vs no armor at all (which would have values of 0,0,0,0,0) and found a P-value of 0.1, which means the cyber armor had no significant difference in damage absorption from no armor at all.

Is there another statistical test you would prefer? I've tried Standard deviations, standard error, and they're all coming in with a "that Cyber armor is... statistically useless!". My favorite is the standard deviation, where the cyber armor is

14+/- 20MD (arguably, the cyber armor could be worth -6 MD, but that's because such a large spread stymies such analyses, which is why I did the T-test)

The T-test is a measure of the randomness in the system.

Okay.
How's it work?
What exactly does "measuring the randomness in the system" entail?

I used Excel to calculate it. the P-value is a measure of the likelyhood that any result you see is due to raw chance.

How is the P-Value determined?

The lower the P value, the lesser the chances of chance being a factor.

Isn't chance always a factor?

P value ranges from zero to one. a P value of 0.05 or less is the standard criteria in most places for significance (most places being science labs, medical analysis, places where data is analyzed). Though you can calculate with 3 trials, the preferred method is to have 5 trials so you can start getting a representative analysis.

Okay.

Standard deviation is simply a measure of the 'spread'. It is, in short "how much, on average, does the result differ from the average".

Gotcha.

I used Excel to calculate it. If, for example, the average was 6.4, but the standard deviation was 2.6, I would say, it's 6.4 +/- 2.6 (meaning that your spread is anywhere from 3.8 to 9.0). This method tends to break down if you have a spread with isolated, abberant points, (for example, 0,0,7,13, 50), when you throw the '50' in the mix, assuming a standard distribution around a mean, there would be a -20 somehwere down the line. It also means that we don't have enough points to analyse this way.

So basically, we need more data?

Try looking at his post above that one, it explains it in detail

You may have to go back a page

[DocS]

Okay.
How's it work?
What exactly does "measuring the randomness in the system" entail?

How is the P-Value determined?

Isn't chance always a factor?

Chance is always a factor, but to different levels in different systems. I'll use a marksmanship analogy... If you see a single shot hit the bullseye, the question is, 'did the shooter get lucky'? So you have him shoot 5 shots, and if they all hit the bullseye, then it means that there wasn't much 'chance' shifting things about.

If however, he hits all over the target, then you can point to the bullseye and say "he got lucky", you've essentially found that chance plays a big roll, and that the positive result was just lucky.

So, now the question is, you have two marksmen, who both have some degree of spread.... how do you quantify how good they are? What if they're just as good as each other, just one got a little lucky and one got lucky in different ways? You can just average all how far all the shots are from the bullseye, but then all you have is an average. But since there's random variation, what are the odds that the two averages are *really* different, or just the result of luck (if you already have a spread, that means one or two lucky shots could really shift the mean one way or another).

You need to somehow put a 'number' on 'spread' so that you can compare spreads to determine differences.

And no, sports do not do this with scores, and yes, they should if they were *really* interested in truly determining who's the best marksman/golfer/runner etc. But it's a lot of work, and in the end, if you win The Masters by getting lucky, it's just as photogenic as winning with skill.... so no one truly cares. But if you're interested in the *truth* of the matter, you need to tease out the 'chance' by quantifying the 'spread' somehow.

I normally hate to Wiki (hate is such a 'small' word, despise is more accurate, loathe, etc), but this breaks down "how" the calculation is done (I've done it by hand, it's good for the soul, but I've been using Excell for six years on it now, so I'd need to go over the Eq's again to calculate it). And I've checked the Wiki entry with other sources, and the Wiki appears right.

look under
Student's distribution probability function and p-value
http://en.wikipedia.org/wiki/Student%27 ... stribution

Now the p-value is just a number that relates to a probablility, it's a measure of likelihood, so then you need to arbitrarily put your flag down on what 'p'value means a significant difference (yes, p of 0.05 is arbitrary). However, in most places, the flag has already been put down, if your p-Value is above 0.05, it's considered to not be significant. p<0.05 is... dare I say it... universally arbitrary. So I'm not budging here.

So basically, we need more data?

And here's the sticky wicket. The more data we get, the tighter things should get, the better the test, and the greater likelihood of significance being observed between differences (yes, folks do take bad tests, do them 1000 times in order to find significance). Which leads to a dreadful question..

If we need to do the test 20 or 30 times to see a difference.... is the difference large enough that we really should be worrying about it?

[Nekira Sudacne]

Okay.
How's it work?
What exactly does "measuring the randomness in the system" entail?

How is the P-Value determined?

Isn't chance always a factor?

Chance is always a factor, but to different levels in different systems. I'll use a marksmanship analogy... If you see a single shot hit the bullseye, the question is, 'did the shooter get lucky'? So you have him shoot 5 shots, and if they all hit the bullseye, then it means that there wasn't much 'chance' shifting things about.

If however, he hits all over the target, then you can point to the bullseye and say "he got lucky", you've essentially found that chance plays a big roll, and that the positive result was just lucky.

So, now the question is, you have two marksmen, who both have some degree of spread.... how do you quantify how good they are? What if they're just as good as each other, just one got a little lucky and one got lucky in different ways? You can just average all how far all the shots are from the bullseye, but then all you have is an average. But since there's random variation, what are the odds that the two averages are *really* different, or just the result of luck (if you already have a spread, that means one or two lucky shots could really shift the mean one way or another).

You need to somehow put a 'number' on 'spread' so that you can compare spreads to determine differences.

And no, sports do not do this with scores, and yes, they should if they were *really* interested in truly determining who's the best marksman/golfer/runner etc. But it's a lot of work, and in the end, if you win The Masters by getting lucky, it's just as photogenic as winning with skill.... so no one truly cares. But if you're interested in the *truth* of the matter, you need to tease out the 'chance' by quantifying the 'spread' somehow.

I normally hate to Wiki (hate is such a 'small' word, despise is more accurate, loathe, etc), but this breaks down "how" the calculation is done (I've done it by hand, it's good for the soul, but I've been using Excell for six years on it now, so I'd need to go over the Eq's again to calculate it). And I've checked the Wiki entry with other sources, and the Wiki appears right.

look under
Student's distribution probability function and p-value
http://en.wikipedia.org/wiki/Student%27 ... stribution

Now the p-value is just a number that relates to a probablility, it's a measure of likelihood, so then you need to arbitrarily put your flag down on what 'p'value means a significant difference (yes, p of 0.05 is arbitrary). However, in most places, the flag has already been put down, if your p-Value is above 0.05, it's considered to not be significant. p<0.05 is... dare I say it... universally arbitrary. So I'm not budging here.

So basically, we need more data?

And here's the sticky wicket. The more data we get, the tighter things should get, the better the test, and the greater likelihood of significance being observed between differences (yes, folks do take bad tests, do them 1000 times in order to find significance). Which leads to a dreadful question..

If we need to do the test 20 or 30 times to see a difference.... is the difference large enough that we really should be worrying about it?

Weellll....

In THIS case, i'd have to actually say yes.

After all, an average cyber knight will have 20-30 fights before getting to 5th level.

[Dead Boy]

Let's see if I can help here.

I took the values of the Cyber armor damage, 7,50,0,0,13,14 and did a Students T-test vs no armor at all (which would have values of 0,0,0,0,0) and found a P-value of 0.1, which means the cyber armor had no significant difference in damage absorption from no armor at all.

So you set the Alpha at the typical 5% and made the null hypothesis "the armor made no significant contributions". That's nice and all, but because of the limited number of samples to draw on, it's probable that the randomness has not been worked out yet. For a better test I'd suggest politely asking KC to run 15 more Simulations so you can have a more acceptable population base to draw your conclusions from and ensure there isn't a sampling error in play.

[DocS]

Let's see if I can help here.

I took the values of the Cyber armor damage, 7,50,0,0,13,14 and did a Students T-test vs no armor at all (which would have values of 0,0,0,0,0) and found a P-value of 0.1, which means the cyber armor had no significant difference in damage absorption from no armor at all.

So you set the Alpha at the typical 5% and made the null hypothesis "the armor made no significant contributions". That's nice and all, but because of the limited number of samples to draw on, it's probable that the randomness has not been worked out yet. For a better test I'd suggest politely asking KC to run 15 more Simulations so you can have a more acceptable population base to draw your conclusions from and ensure there isn't a sampling error in play.

20 sims would be better... but that's a LOT of sims! If he's willing to do it, I'm fine, but the CK will need to be trying to dodge the incoming blasts (doing simo-attacks when the armor is tatters doesn't seem to reflect the sort of conditions expected, and it's also so easy to calculate the statistical likelihood at that point that trials seem a labor-intensive way to handle it).

but I've run 5 CK sims myself, I personally don't wanna do 20. I'm curious as to how much MDC that cyber armor acts like in this case... but I'm not 'willing to do 20 sims' curious.

I still gotta run my Grunt sims (an HPLC waylaid me this weekend). And then the question comes about whether it's worthwhile to do 20 realtime sims or to just do the stats on "If the bad guy has a +4 to strike, and you have a +3 to dodge... what % of the undodged shots will be absorbed?

[Shadyslug]

So...rather than reading all that extraneous stuff...

What's the verdict?

[Nekira Sudacne]

So...rather than reading all that extraneous stuff...

What's the verdict?

hasn't been reached yet.

[Shadyslug]

So...rather than reading all that extraneous stuff...

What's the verdict?

hasn't been reached yet.

Will it ever?

[Nekira Sudacne]

So...rather than reading all that extraneous stuff...

What's the verdict?

hasn't been reached yet.

Will it ever?

As soon as DocS posts the simulations he claims to have already done...

[DocS]

As soon as DocS posts the simulations he claims to have already done...

Haven't done the grunts yet, and haven't posted the CK, been busy at werk. Sorry about the delay. TIME TIME! I NEED TIME!!! AARRGGHH!

[Shadyslug]

OK, I'll buy this
to dodge is better than to be hit (makes sense)
and AR hobbles cyber armor significantly (also makes sense)
I think that the situational combat isn't necessarily the whole tale of the tape here
to wit: lock both PCs in a jail cell naked - who has a clear advantage?

also, we're looking at very short exchanges of straight combat. What happens when said grunt runs out of ammo and Mr. Cyber-Knight takes a quick nap (to recharge ISP for his friend Mr. psi-sword)?

Also the coalition's true strength lies in overwhelming numbers. A lone grunt in coalition colors is another name for "target practice" in a lot of parts of NA. They should always be rolling in a gang. With air support.

I think that situational modifiers should be added for realism here.

DING DING DING...give the panda a cigar...

[Killer Cyborg]

As soon as DocS posts the simulations he claims to have already done...

Haven't done the grunts yet, and haven't posted the CK, been busy at werk. Sorry about the delay. TIME TIME! I NEED TIME!!! AARRGGHH!

Yeah; I'm having the same problems.

At least there's no deadline with this project.

[Dead Boy]

20 sims would be better... but that's a LOT of sims!

Want to know what of one my hobbies is? I make dice. Experimental die shapes that few have ever thought of on behalf of grand die maker Lou Zocchi, owner of Game Science. I made a non-top-spinner D14 that is so beautiful, it would bring a tear to your eye. It also violates the laws of the Platonic Solids, but I still have faith in it's ability to hold an even probability of coming on any given side. In testing any new die design, the new shape has to undergo a marathon 1,000-roll test to see if it truly is an even roller, because anything less still has too much randomness in it to be statistically significant.

A few years ago I did such a 1000-roll results test on an experimental non-top-spinner D18, and the results were fantastic! Shame it didn't go into production, (you can find the shape in what is called the "Bouncing D18", unfortunately it's made of rubber and isn't numbered).

The point is, if I can find the time for such a time consuming test, 15 more sims on this should be no big deal.

[DocS]

20 sims would be better... but that's a LOT of sims!

The point is, if I can find the time for such a time consuming test, 15 more sims on this should be no big deal.

Ah, but you underestimate the power of my ADD.

Oh look, a pigeon!

[Dead Boy]

So, statistical debates aside, when do we get to see the other side of this with the brave CS Grunt standing against the monster hoards attacking in single file down a narrow hallway?

[rat_bastard]

also, we're looking at very short exchanges of straight combat. What happens when said grunt runs out of ammo and Mr. Cyber-Knight takes a quick nap (to recharge ISP for his friend Mr. psi-sword)?.

as the cyber knight's psi-sword is free and instant, that does not happen.

[Killer Cyborg]

So, statistical debates aside, when do we get to see the other side of this with the brave CS Grunt standing against the monster hoards attacking in single file down a narrow hallway?

1. Once we've finished with the CK.
Since my initial sims are not entirely acceptable to Doc, I'm going to run some more (as well as read over his more technical posts when I find the time, and respond to them).
But when this happens, be prepared for more arguing about details.
2. It's not a narrow hallway; it's an infinite and featureless plane.
(And presumably the monsters are teleported in, and the corpses are teleported out again, which is why everybody's instantly in range and why nobody uses the MDC corpses for cover against the other monsters)

[Dead Boy]

... be prepared for more arguing about details.

Like you expected anything less. There's ALWAYS someone who nitpicks about something.

[Noon]

You guys aren't writing a program to run the simulation?

Besides, neither of you are going to concede if you can keep saying 'Oh, but factor X wasn't included'

Every time a results almost determined, one of you will bring in factor X, and if that doesn't change anything, factor Y, then factor Z

The result wont be uncertain for game world reasons, it'll be uncertain for entirely meta game reasons - that each party knows they can keep introducing extra factors.

Anyway, on the program, what's the fight involve so far - grunt keeps shooting, CK keeps hacking/dodging?

[Killer Cyborg]

You guys aren't writing a program to run the simulation?

I can't program my way out of a paper bag.

[Ziggurat the Eternal]

remember that ck's have major bonuses to battling tech opponents, or are we talking about pre RUe cyber knights? cuz they sucked....hard

[Killer Cyborg]

remember that ck's have major bonuses to battling tech opponents, or are we talking about pre RUe cyber knights? cuz they sucked....hard

I always wonder why people bother posting when they obviously haven't actually read the thread.

[Nekira Sudacne]

Your latter assertion is correct, the calculation is wrong (this is not opinion). Particularly the +5% number (by your numbers, a +10 would give you a 100% chance to dodge, but as any juicer player will tell you, it's not that good). Now the question is, will you believe it when I give the correct answer?

Would actually take a +20, and if you eliminated critical strikes it would give him a +100% chance to dodge, but with it it would give him a 95.25% dodge, unless you have some rule change where a defensive roll fails on a 1....which would change everything in a complex way but still result in a 90.25% dodge. That is assuming the +20 is in relation to the strike bonus of the attacker, not just a flat +20....

The problem here (and in later calculations, but i'm cropping for space), is that your calculations do not take into account attackers bonus to strike.

Ex. if you had (somehow) a Cyber knight with +20 to dodge, well, he would dodge a lot, but always sinse opponents will likely have bonuses to strike.

Going back to the origional exsample in this thread, the monsters in question had +4 to strike with their fire bolt attacks. ergo, +20% eliminating critical strikes and automatic misses would not be 100%. the monster could roll a 19 and the CK a 2 and would be hit, also such a hit would also automatically bypass cyber armor.

actually, such a bonus _would_ render cyber armor completely ineffective sinse any attack that could hit in the first place, would bypass the Cyber-armors AR as well.

[Noon]

You guys aren't writing a program to run the simulation?

There doesn't need to be, the end result: The battle turns out diffrently every single time you run it.

Given the math your into in the post above, I'm surprised your saying there will be no trend toward one side winning.

[Noon]

Where are the stats for the monster, in case I feel like programing a simulation?

[Killer Cyborg]

Where are the stats for the monster, in case I feel like programing a simulation?

http://www.palladium-megaverse.com/foru ... 14#1564714

Anyway, here are the stats:

The Monsters:
+3 initiative
+1 strike
1 attack per melee
49 MDC each total
IQ: 5 (pretty stupid)
PS: 23
PP: 19
PE: 25
Spd: 5
Horror Factor: 9
PPE: 10
22' tall (extra MDC added to total)

Weapons:
-Large retractable claws- 3d6 MD
-Tail- 2d4 MD (not an extra attack)

Abilities:
-Nightvision
-Swim 90%

Powers:
-Calling
-Fire Bolt once per melee Damage: 4d6 MD, +4 strike

Feeds On: Flesh
Weakness: Water