Talk:Epic dodge

Prerequisites
You don't need Tumble 30 for this feat. Maybe you need 1 point there, though. -- Olivenmann 15:26, 8 September 2005

Epic Dodge and Attacks of Opportunity
I'm not sure but doesn't epic dodge come into play when the character is subject to an attack of opportunity? I remember using up epic dodge one round when using a ranged weapon in melee combat even though the attack roll wasn't high enough to breach my AC. Harleyquin 14:34, 12 May 2006 (PDT)


 * I have also seen Epic dodge randomly stop working when multiple attacker are in differnt modes. (ie ranged + Melee) Breifly breaking contact usually resets it. Also seen the same problem, again not consistantly, with different attack modes using Power attack, Expertise/Improved Expertise, Flurry of Blows and Dwarven Defender Stance. DaBear 08:50, 15 May 2006 (PDT)

Should description match manual/game description?
As pointed out, the manual states that one needs 30 ranks in tumble. I've seen a few comment here and there on the wiki that descriptions should be left alone, even if they are contrary to the actual game rules, and exceptions be mentioned in the notes. Is that correct? And if so, should the description on this page be changed back to say "30 ranks in Tumble"? Kato 21:59, 2 June 2006 (PDT)


 * Yes. I think it's clearest if we present the actual ingame text (from the talk table) and then note any errors. -- Alec Usticke 10:01, 3 June 2006 (PDT)


 * Done. Description now matches Hordes of the Underdark manual (dated 10/14/2003 which accompanies Diamond Edition) as well as talk table entry (dialog.tlk strref 8639) as of 1.67. -- Kato 11:46, 6 June 2006 (PDT)


 * Good stuff. I did remove those 2 stars in the original description you added however as they are not in the manual and in general in other wiki articles, incorrect info in the original descript is not denoted in the original descript itself either. GhostNWN 12:00, 6 June 2006 (PDT)


 * Yea, my bad, I meant to take those back out before I hit post and didn't. -- Kato 12:54, 6 June 2006 (PDT)

Versus non-attacks
Since your Epic Dodge use for the round can be fired off at an enemy making an AoO even if they missed, I'm curious: Does it fire off if a Tumble check succeeds? What about concealment? I would assume that either way, Spring Attack prevents those AoO's from even being initiated, and would save your Epic Dodge uses until it's actually useful, correct? --67.2.18.186 00:58, 8 January 2012


 * If a tumble check succeeds, there is no attack of opportunity. How could something that does not exist be the first attack in a round? --The Krit 20:15, January 14, 2012 (UTC)

"indirect +5 AC"
What do other people think about revision 64217? I think the part about AC is demonstrably false if the opponent is hasted. Or gets an AoO. Or if "opponent" is really "opponents". That leaves the part about reducing the number of incoming attacks per round by one, which I thought was already mentioned as the entire purpose of this feat. --The Krit (talk) 02:46, November 5, 2013 (UTC)


 * Hmm... even in a "normal" case the math does not support the "indirect +5AC" claim. For this case, assume you are fighting a single opponent who gets three attacks per round, hitting on 5, 10, and 15 (80%, 55%, and 30% of the time). If your AC increases by 5, the chance for each swing to hit would be 55%, 30%, and 5%, so the expected hits per round is 55% + 30% + 5% = 0.9. If instead you pick up epic dodge: the first swing will never hit (0%), the second swing will only hit if the first and second rolls were otherwise good enough to hit (80% &times; 55%), and the third will only hit if at least one of the first two swings was good enough to hit (80% + 55% &minus; 80% &times; 55%) and the third swing was good enough (30%). That works out to 0% + 44% + (135% &minus; 44%) &times; 30% = 0.713. Significantly lower (better) than an AC boost of 5. I could try putting this info in the article, but there are so many variables I don't know of the stats could be made understandable and accurate. (For example, if you give the above attacker haste, the expected hits per round become 1.45 for increasing AC by 5, and 1.4626 for epic dodge. Suddenly epic dodge is the lesser option.) --The Krit (talk) 03:02, November 27, 2013 (UTC)


 * I should also correct myself: reducing the number of incoming attacks per round by one is only what the description of the feat claims. The feat is actually more effective than that, so one could argue that the description is less than the entire purpose of this feat. --The Krit (talk) 03:12, November 27, 2013 (UTC)


 * The reason I added this description (which as you say is not entirely correct) is to illustrate how useful the feat is beyond reducing the number of attacks by one. So if the opponent is not hasted, we still have an indirect AC boost of (better) than +5, which, when compared to other epic feats like Armor Skin (+2), is hugely useful in my mind. Maybe a compromise is in order, we could say something like "In addition to reducing the number of rounds blah blah Epic dodge can be thought of as giving an approximate indirect AC increase of +5 or possibly even better against the character's current opponent." --130.237.223.173 09:41, 27 November 2013‎ (UTC)


 * Support your claim. I've already demonstrated why the attack progression ("+5") should not be considered an accurate assessment. What evidence do you have that "+5" has any special meaning in this context that warrants being mentioned in the article? I had to add haste to my above calculations to get something close to "+5". (The non-haste case was much closer to the 0.7 expected hits per round that would come from a +7 AC increase.) I did not do comprehensive calculations&mdash;I left out variations such as changing the number of attacks per round, changing the chance of the top attack hitting, factoring in off-hand attacks, etc.&mdash;but I did do a more in-depth calculation than you did. Since you still think "+5" is worth mentioning, it is up to you to back that up (with calculations, probably). Why should there be any claim made beyond something like "an indirect AC increase somewhere between +0 and +&infin;"? (My support for mentioning infinity here is the case where the opponent has a single attack per round, in which case the opponent never hits.) --The Krit (talk) 20:03, December 1, 2013 (UTC)


 * Like I said, I'm not saying the indirect AC bonus is (if thought of this way) +5, (because as you've shown, it is not). All I was saying is that adding a line about how the reduction of an earlier attack causes you to become much harder to hit (approximately equivalent to +5 in most common scenarios) is worth doing in the article, because it may make people understand this feat in terms of approximate power in a unit they recognize. Popular science magazines do this all the time - for example, they might say the oxidation potential of some oxidant is 500 times more powerful than coastside rusting, even though that happens through completely different mechanisms.83.183.17.141 15:20, December 10, 2013 (UTC)


 * The problem with that is that 'most common scenarios' would include complicating factors such as more than one enemy, monk unarmed BAB, concealment, extra attacks, etc. If everything was factored together then the average AC boost that would cause an equivalent number of misses could easily come out to +4 or even +3. +5 is hardly a common average. WhiZard (talk) 02:40, December 11, 2013 (UTC)


 * I agree (and count dual-wielding in), but I still think it's useful to put an approximate number on how powerful this feat is, through an approximate lower bound. --83.183.17.141 02:45, 11 December 2013 (UTC)


 * I'm saying it's not even approximately equivalent to +5. Any attempt to pin down a number is going to be grossly inaccurate because of the great variability here. Trying to specify an equivalence to AC is going to be roughly as tough as trying to specify an AC equivalence to concealment. It's an "apples and oranges" situation as far as I see. Now if you can present some evidence to the contrary, rather than just reiterating the same argument over and over again, maybe you could convince someone. --The Krit (talk) 05:18, December 11, 2013 (UTC)


 * If we want a black-on-white claim in a very strict sense, it is this: disregarding AoO's, haste and other bonus attacks made at the full base attack bonus, Epic Dodge is at least as good as if the defender had +5 AC (often much better). It is never worse. Proof: Suppose ED blocks the first attack. Then the opponent's remaining effective attack progression is as if it was -5, giving the defender (at least) +5 AC. Suppose ED blocks a non-first attack. In this case, all previous attacks had missed (the probability of this happening is uninteresting, because we are speaking of this very case), making the remaining attacks act under -5 (or at least +5 AC). 83.183.17.141 15:20, December 10, 2013 (UTC)


 * I already understood this argument based on what you had written earlier. This is written out more explicitly, and I suppose the "at least as good as" is new (although only applicable in a rather specific case), but it is the same argument. Got anything different? --The Krit (talk) 05:18, December 11, 2013 (UTC)

So I did an exhaustive research on this subject. Since you wanted to be convinced by calculations, I wrote a Java program that iterates through approximately 1,000,000 cases of different number of enemies, concealment, blind-fight, epic dodge, attack rolls, AC, attack progressions, attacks per round, extra attacks (from haste, AoO, circle kick and dual-wield). The program calculates the expected number of hits per round for the cases with Epic Dodge, and then searches for the AC needed to get the same expected number of hits. (This is what we're searching for - an AC that makes you hit just as often as with Epic Dodge). Here are the results (see chart):

Totally 1,075,200 cases

Median: 1000 (infinite AC) Smallest effective AC bonus: 1 Mean (including 1000s) 669.0547535342262 Mean (excluding 1000s) 4.338961282433692 Percent cases with AC = 1000: 66.7612537202381% Percent cases where AC = 1 or more: 100.0% Percent cases where AC = 2 or more: 96.88206845238095% Percent cases where AC = 3 or more: 89.30208333333333% Percent cases where AC = 4 or more: 83.26683407738095% Percent cases where AC = 5 or more: 78.77455357142857% Percent cases where AC = 6 or more: 75.56045386904762% Percent cases where AC = 7 or more: 72.95433407738095% Percent cases where AC = 8 or more: 71.12509300595238% Percent cases where AC = 9 or more: 69.86662946428571%

You can find the Java program here: http://pastebin.com/bH0iVCCa

Long story short: 80% of reasonable combat situations will result in an effective AC increase of +5 or more.  Using the data above, I want to prove two points. First, I really dislike the attitude that because something is complicated, you cannot make generalized conclusions about it. I think that this data really proves that the attitudes of the editors below is much too backward and reserved about results on a macroscopic "on average" level. It's not apples vs. oranges - ask any economist how anything can have an effective price!

Secondly, this data shows many things: in the cases an Epic Dodge defender can actually be hit by an opponent, the mean equivalent AC is +4-5 (the mean). The histogram above, from its shape, seems to follow a Poisson distribution (which is expected because this type of combat can be modeled as a Poisson process). Notice also the effective AC increase is never below 1. The test cases were, specifically, 1-4 identical opponents with and without blind-fight with AB -40 to 30 versus AC 20, -5 and -3 attack progressions (and 1-4 and 1-6 attacks per round respectively), with extra attacks due to haste, attacks of opportunity, circle kick and dual-wield, where the defendant has 0-50% concealment. (Feel free to edit the program to do an equivalent experiment for concealment only)

I'm going through with an appropriate edit in the next few days unless anyone has anything to add? Bombax42 (talk) 22:49, December 11, 2013 (UTC)


 * An initial reaction from me: I really dislike people putting words in other peoples' mouths. Nobody here expressed an attitude that one cannot make generalized conclusions about something complicated. The closest I see is me saying that *I* am not going to do the work required to make such a conclusion. There is also the "apples and oranges" comment, but that was directed at the types of calculations being performed. In fact, "apples and oranges" was supposed to encourage someone to jump to a higher level of calculations. So lay off the insults, okay? --The Krit (talk) 01:00, December 12, 2013 (UTC)