Anyone up for trying this out?

Thanks for giving it a go.  The lower your dodge % is, the less often you'll dodge (obviously) - and as you NEVER know when it will work or not, you still need a decent amount of HP to make it work.  Its real tricky.  I'm going to keep working at it from time to time and see if I can come up with something. 

Testing:

• Sedna vs Erb - erb at 40% dodge vs sedna's pounce + silence. 
  
  • After Many attempts (>25), neither pounce nor silence were dodged.  Dodge does not work on these abilities. 
• TB vs Erb - erb at 40% dodge vs all of tb's abilities. 
  
  • After Many attempts (>25), no fire/ice abilities were dodged.
• DA vs erb - erb at 40% dodge vs warp strike, spine attack, and warp area.  
  • After Many attempts (>25), no da abilities were dodged.

So, after testing 3 dg's at random, looks like I'll have to agree eh.  So, the only ability that can be dodged is ub's spit (specifically, the time after each tick occurs in game, its possible to dodge that bit of damage from spit - the effect persists until the timer wears out for spit duration).  As there are other abilities that do damage over time (circle of fire) but cannot be dodge, it makes sense (at least to me) that either dodge SHOULD be applied to all abilities or that the ability to dodge spit ticks is in error. 

At any rate, the viability of dodge is reduced quite a bit if you can still be nuked by all abilities (eg you still must have a good hp stack or an easy escape to use it successfully).  This also makes dodge builds (with reasonable hp) a decent counter to dgs like ub, though (as it impacts his aa damage and his spit damage). 

Anyway, editing the op.  Going to delete some posts, but here you go - thunder and hedgie were right... now go play dow2. 

I tried different dodge builds with da. The only fun part I could have was dodging endless towers and try to steal flags. but fighting wise the dodgesystem is to weak for a character like da. Haven't thought about erebus though.

Tried a dodge erb build in a 2v2 (me laned against an oak).  Didn't end so well for me.  Sadly, I think I was strongest early (dodge + bite did pay off), then mid... but mid/late I was doomed.  DIdn't help that my teammate couldn't hold the hp flag, but the fault was my build at any rate.  Anyway, going to keep taking shots at this.  I really think pendant is a solid item for minion builds, but unfortunately, I wasn't trying to do a minion build.  And when you get assassins footguards + a small hat, there really isn't enough mana for erb. 

The reason that e.g. 15% dodge is less desirable than 15% armor is because evasion is very unpredictable.  Against another demigod, where you might only take 7-10 hits to die, the chance to block one, or maybe two of those offered by 15% dodge isn't nearly as practical as the absolute assurance that you can take one more offered by 15% armor.

Against archer towers, creeps, minions, etc-- any situation where the hit quantity is high-- dodge is equally effective and reliable.  But because of how low the cap is, and because armor is generally more affordable and is available in conjunction with significant health boosts, there's no question which is a better choice.

Then there's the relationship between health, damage reduction, and armor.  Which is stilted even further in favor of raw health and damage reduction, because both of those synergize with every other available means of boosting EHP (including armor and evasion).

Dodge confers exactly the same mitigation values as armor, as long as you have sufficient armor to never drop below 0% (Bite, Ground Spikes). This is because if you drop below 0% armor, you take massive bonus damage due to the formula demigod uses. 

In fact, point-for-point, dodge is better than armor because Dodge doesn't have DR. However, due to the mechanics of the game (the major sources of damage, abilities, is not dodgeable, and armor pieces that have dodge typically have very little HP), dodge is not something you can economically get without incurring too heavy of an opportunity cost (in skill points, or money that could have gone towards health items instead). 

Patently false. Evasion is unpredictable but there is no difference between 10% armor mitigation and 10% dodge in terms of "average" lifespan. It ultimately comes down to playstyle, if and only if armor and dodge were equally difficult to acquire. Dodge is a "feeling lucky punk?" stat, but that doesn't make it inferior, just risky. Additionally, if you have dodge are also screwing up with your opponent's mind.

I know that you said afterwards that against high rate-of-fire targets the difference is minimized but you're still misrepresenting the situation, saying that you might get unlucky and die, whereas getting unlucky and dieing is just as possible as getting lucky and winning assuming a perfectly even fight. 

If dodge and armor were equally difficult to get (again, not the case since very few dodge items provide health and are usually more expensive), since dodge doesn't have DR, everyone would stack dodge. If the items were rebalanced, and dodge was given a DR, rather than just a hard cap, then you would want to have exactly half dodge, and half armor. 

Let's say you have a 50% mitigation "budget". You could do 50% armor, or 50% dodge, or 25% armor, 25% dodge, or 10% armor, 40% dodge, or 40% armor, 10% dodge. 

1) 50% Armor gives you 50% mitigation, 200% effective hp 

2) 50% Dodge gives you 50% mitigation, 200% effective hp

3) 25% Dodge, 25% Armor gives you 1.25 * 1.25 == 56.25% mitigation, 228.57% effective hp

4) 10% Armor, 40% Dodge gives you 1.1 * 1.4 == 54% mitigation, 217.39% effective hp

5) 40% Dodge, 10% Armor gives you 1.1 * 1.4 == 54% mitigation, 217.39% effective hp

As dodge is currently implemented and itemized, in order for it to be effective, it probably would have to dodge abilities as well (probably dodge % against abilities would be halved, for balance). If dodged were itemized like armor, and found on easily accessible hp items then dodge would be awesome (see above). So awesome it would have to be nerfed and DR'd.

come play again max - been too long.  (I'll actually read your reply in little bit.. game time)

I was contesting this point. Against small samples sizes, it is not going to be consistent, as I put here

It's less desirable to YOU because the inconsistent nature of small numbers of attacks and that makes you feel uncomfortable. For some people, that is desirable. I'm saying that, mathematically, they are identical assuming that they mitigated the same things and were equally easible to attain. This is obviously not true.

All my discussion has had those assumptions in place because I'm making abstract theoretical discussions. The fact of the matter is that dodge sucks in demigod, and the only time it's really worth it is DA end-game, and Erebus' level 15 mist skill. Which is pretty much how I end all my posts.

actually, I think the best change for dodge is to make it count on abilities.  Even at a 70% dodge rate, against nuker dgs, the dodge build becomes a bit useless in normal money games.  Without huge ability nukes, its more viable.  As it stands, its what... an ok counter to a ub or perhaps an aa reg?  Anyway, I do find it strange that the dodge mechanic isn't applied across abilities.  If I can dodge an arrow... I could probably dodge a snipe once in awhile. 

You wouldn't want to evade interrupts, if we were doing it skill-by-skill. AoE attacks don't really make (logical) sense, but it's also the least inbalanced thing to dodge. 

A better way to do it, in my opinion, is to make "ability dodge" a certain % chance of your "autoattack dodge". I know that this is an obfuscation but remembering a ability:autoattack ratio may be easier than remembering what attacks can and cannot be dodged. 

I'd suggest a 3:1 ratio. That is, with 15% dodge you'd have a 5% chance to dodge abilities. I think that's pretty powerful without being insane, espiecially when you consider desperate boots (which would now have an iconic name). 

Right now, abilities are a 'sure thing' where you can count on an exact amount of damage when you use one.  This is good, because you have to spend mana to use them.  Even a small chance to dodge most of them would be extremely frustrating and would reduce confidence in ability use in general.  So unless you can argue for specific abilities that you think should be evadable, and why, I don't see it working at all, nevermind being accepted by what remains of the player base.

Re: Max

There is no way to do a 'soft cap' other than using the same system as armor, where each point of evasion gives you the same relative percent of dodge.

If this were implemented with the same formula as armor (only modifying AA evasion chance instead of AA DR), you would give items that currently have 10% evasion something like ~500 dodge points, or 16.7% evasion chance, so a net boost to begin with.  Adding another of those items for 1000 dodge points would still give you a +16.7% improvement over your current chance to dodge, but only an absolute 28.5% chance.  So each 500 point item would be just as useful as the last, just like each 500 point armor item is just as useful as the last, you're still ahead of where you would be with two flat 10% dodge items, AND you don't have to worry about having 'too many' dodge items and hitting a cap or becoming unhittable.

Let's face it - one 10% dodge item does jack shit right now.  It's never worth taking.  If that same item did 15-20%, suddenly it becomes a consideration.  But you can't just boost those percents directly without four such items making you almost unhittable by AA (60-80% dodge) or without hitting a reasonable hard cap with the third item.

For reference, here's how the point system would relate to the current absolute amounts (same as armor):

500 = 16.7%
1000 = 28.5%
1500 = 37.5%
2000 = 44.4%
2500 = 50%
3000 = 54.5%
3500 = 58.3%
4000 = 61.5%
4500 = 64.3%
5000 = 66.7%

As with armor, this system would let you push the percentages beyond the current cap without becoming unhittable.  It would also allow a potentially greater amount of relative evasion per item (say 1500 points or +37.5% relative for Cloak of Elfinkind), and would let more items with more evasion amount exist without either rendering some of them useless (already past cap) or too powerful (no cap).

The only drawbacks of this are:
1) No instant visual confirmation on percent, but if you've learned how armor works, you already know how these points work.
2) Becomes highly desirable to have at least one evasion item.  Actually less of a problem than armor, because evasion items don't also have health and are currently rather worthless.