Lots of statistics from the Monster Manual

[Jhaelen]

I decided to determine a couple of statistics as well, since I'm more interested in the distribution of certain characteristics among monsters of a given role or origin.

For example here's the distribution of attacks per role:

Attacks     vs. AC | vs. Fort | vs. Refl | vs. Will
----------------------------------------------------
Artillery:  43.0%      18.3%      28.5%      10.2%
Brute:      67.3%      10.6%      18.2%       4.5%
Controller: 34.8%      20.6%      17.6%      27.0%
Lurker:     57.0%      13.0%      18.0%      12.0%
Skirmisher: 72.1%       9.1%      13.6%       5.2%
Soldier:    60.5%      16.3%      16.3%       6.9%
Leader:     51.7%      18.3%      14.2%      15.8%

Some observations:
- So, the majority of monsters attacking Reflex Defense have the Artillery role.
- Similarly, the majority of monsters attacking Will Defense have the Controller role.
- Controllers have the most even distribution of attacks against the diffferent defense types.
- A majority of Skirmishers only have attacks against AC.

Anyone else determined something like this?

[cooperflood]

This thread is great, I have already referenced it several times for various reasons. I hate to ask, but I was wondering if it would be possible to add some information on the type of attack not just what defense it targets. For example what percentage of monsters have melee, ranged, area, close, burst, or blast attacks. A variety of paragon tier feats, magic items, and various other effects give specific bonuses against only a certain attack. The reason why this question came up origionally was in a discussion on the value of Combat Anticipation. If it were further divided up by which defense those attack types attacked would be even better. For example I think (with no real evidence) that 90% of all Melee attacks will target AC.

I have been debating cracking open the MM my self, but if someone else wants to do it...

[Alaxk Knight of Galt]

Thanks for putting this together Kerbarian. All sorts of interesting information.

[Magus Coeruleus]

Since various people are taking these data and reporting different analyses from them, perhaps there could be a Wiki entry on this so that instead of sifting through the thread one could check out the entry. Someone asked about a PDF/DOC before but if people continue to add interesting stuff to this then an evolving wiki entry might be better.

[Mistwell]

So it seems to me that radiant damage that attacks will is very useful.

Having a high will defense of your own, however, isn't as useful as having other high defenses.

[Voadam]

Originally Posted by Aria Silverhands

You don't need to grok the math to make an interesting story. The math is balanced behind the scenes already and you just have to follow the xp budgets and guidelines. The only math required is simple addition. Everything else is worthless and takes away from the game.

Analyzing the math can show whether it is balanced behind the scenes and works how you want it to.

Do creatures in the MM fit within the DMG guidelines?

Do statistics/attacks/defenses vary by monster role as you'd expect them to?

If you try and match attacks to creature weaknesses described in flavor is that effective?

I see three ways to determine that the numbers are balanced.

1 Faith: You can take WotC's word for it that things are balanced.

2 Experience: You can take an MM creature and see how it works through actual play.

3 Analysis: You can analyze the numbers.

Option 3 seems a valid method for determining how well the numbers work and do their job.

[TikkchikFenTikktikk]

First, as a DM this is very useful. Thanks for the hard work, kerbarian.

The guidelines in the DMG are generic to the system. But this kind of info will help me design adventures tuned to the characters my players are actually running. It also helps me advise players (especially new players) on how to develop their characters be effective in my game.

Anyone who replies with "a good DM could do without" gets a big, premptive you.

Second, this a great peer review of the 4E designers' work. This is the kind of analysis that keeps WotC honest (more honest, at least).

[Zeroth]

Hey, I was wondering if you could make the raw data available? The analysis, while useful, is still flawed. Means are susceptible to outliers, of which I know there are several in the MM. I'm interested in seeing the median, the quartiles, and the standard distribution of the data.

This, I feel, is more useful for the DM, to learn what really is a challenge, what is exceptionally weak, etc. In addition, I have a gut feeling the data isn't centered around a single median, but has a few clumps.

[Mustrum_Ridcully]

Originally Posted by Zeroth

Hey, I was wondering if you could make the raw data available? The analysis, while useful, is still flawed. Means are susceptible to outliers, of which I know there are several in the MM. I'm interested in seeing the median, the quartiles, and the standard distribution of the data.

This, I feel, is more useful for the DM, to learn what really is a challenge, what is exceptionally weak, etc. In addition, I have a gut feeling the data isn't centered around a single median, but has a few clumps.

I suppose the raw material is called "Monster Manual".

[Zeroth]

Originally Posted by Mustrum_Ridcully

I suppose the raw material is called "Monster Manual".

I'm inherently lazy. OP has already done the hardwork, and has achieved the kudos for it, so why not share the raw data if he has it?

Besides, I want to use my Stats knowledge on actual data sets instead of ridiculous and silly data sets. :/

[kerbarian]

Originally Posted by Zeroth

Hey, I was wondering if you could make the raw data available? The analysis, while useful, is still flawed. Means are susceptible to outliers, of which I know there are several in the MM. I'm interested in seeing the median, the quartiles, and the standard distribution of the data.

This, I feel, is more useful for the DM, to learn what really is a challenge, what is exceptionally weak, etc. In addition, I have a gut feeling the data isn't centered around a single median, but has a few clumps.

The raw data I used is just the full text of the MM from the pdf (available here or here). I selected the entire pdf, then copied and pasted it into a text file.

What I can provide is the perl script I used for parsing. My first script, which I used for most of the statistics, operated on a text file that had a lot of somewhat manual pre-processing done to it, so that script wouldn't be very useful. The latest analysis script, though (for monster attacks), is much cleaner and operates directly on the pdf text, so I've included that script.

I'm hoping to spend a little more time on this soon and do some of the extra analysis that has been requested, and in the process I'm planning to convert the rest of the analysis over to the new style and add comments. If/when that happens, I'll have a much more thorough perl script I can post. Until then, this script at least contains the basics for separating out monster entries and such.

parse_MM_attacks.pl:

Spoiler:

#!/usr/bin/perl

open MMFILE, $ARGV[0] or die("usage: $0 <filename>\n");
read MMFILE, $mm_text, 1e7 or die;

@entries = ($mm_text =~ /(^[^\n]*Level \d+ (?:Elite |Solo )?(?:Artillery|Brute|Controller|Lurker|Minion|Skirmisher|Soldier).*?Cha \d+ \(.?\d+\))/gsm);
for $entry (@entries)
{
	$entry =~ s/\s+/ /g;
	$entry =~ s/Refl ex/Reflex/g;
	
	$entry =~ /Level (\d+)/;
	$level = $1;
	if($level <= 10) {
		$tier = 0;
	} elsif($level <= 20) {
		$tier = 1;
	} else {
		$tier = 2;
	}
	
	$count[$tier]++;
	$count[3]++;
	
	@attacks = ($entry =~ /(\+\d+ vs\.? (?:AC|Fortitude|Reflex|Will))/g);
	@def_seen = ();
	for $attack (@attacks) {
		$attack =~ /\+(\d+) vs\.? (AC|Fortitude|Reflex|Will)/;
		if($2 eq "AC") {
			$def = 0;
		} elsif($2 eq "Fortitude") {
			$def = 1;
		} elsif($2 eq "Reflex") {
			$def = 2;
		} else {
			$def = 3;
		}
		
		if(!$def_seen[$def]) {
			$def_seen[$def] = 1;
			$def_count[$tier][$def]++;
			$def_count[3][$def]++;
		}
		
		$attack_minus_level[$tier][$def] += $1 - $level;
		$attack_minus_level[3][$def] += $1 - $level;
		$attack_count[$tier][$def]++;
		$attack_count[3][$def]++;
	}
	
	if($def_seen[0] == 1 && $def_seen[1] == 0 && $def_seen[2] == 0 && $def_seen[3] == 0) {
		$def_count[$tier][4]++;
		$def_count[3][4]++;
	}
}

for $def (0 .. 4) {
	for $tier (0 .. 3) {
		printf "%10.1f", 100 * $def_count[$tier][$def] / $count[$tier];
	}
	print "\n";
}

print "\n";

for $def (0 .. 3) {
	for $tier (0 .. 3) {
		printf "%10.2f", $attack_minus_level[$tier][$def] / $attack_count[$tier][$def];
	}
	print "\n";
}

[Zeroth]

Okay, guess I will be compiling the raw dataset from the from the MM then. :P Like I said, there are plenty of outliers, unusual cases, etc.

[Helepolis]

Well, I can't help with much of it, but if you're interested, I went through the MM and counted up the number of resistant/vulnerable/immune monsters of each energy type.

On another note, I also counted what energy keywords are used by every player power in the PHB, divided by class.

Yeah, I was bored. But never would have been able to pull out something like what the OP did here. This is amazing.

[MatthewJHanson]

Originally Posted by Rechan

What had resistance to poison beside Undead and the green dragon? I find that a shame, because I like poison damage.

Off the top of my head, constructs are also immune, and Yuan-ti resist.

[Helepolis]

Worth noting is that there's about 4, literally, powers in the PHB that deal poison damage.

[Mal Malenkirk]

Originally Posted by Mistwell

So it seems to me that radiant damage that attacks will is very useful.

Having a high will defense of your own, however, isn't as useful as having other high defenses.

It may seem so but attacks that target will may have worse secondary effect on average than those target reflex. Getting dominated by a vampire is rather annoying, for example...

It's hard to really determine which defend is the most useful just from these numbers (though they are great stuff!).

Maybe we could do even more math and attach a value to the secondary effect and see, on average, which defenses protect the PCs against the worst secondary effects. Weeee!

[Vaeron]

Wow... Nice work

[Perenon]

Hi.
Are there any statistics about which type of damage monsters are able to deal?
I.e. like 10% of all monsters (respectively all monsters of heroic tier etc.) can deal fire damage, 2% deal psychic damage...

I didn't find a table about this information here, so does anyone know another site to look for it? Or did i maybe just fail to see it here?

Thx!

[Perenon]

No one has seen anything of that kind?

[Rel]

Originally Posted by Perenon

No one has seen anything of that kind?

Might be worth doing some Advanced Searches using the Compendium to see if you could compile that kind of info.