Evasion Penalties

[droit]

So "evasion" has two components: dodge DS and the chance to completely evade an attack. Armor and shields have built-in dodge penalties associated with them that increase with the asg/size of the defensive item. These penalties have a quantifiable effect on the dodge DS equation. My question is, do the heavier armors/shields actually decrease your chance to outright dodge an attack?

From KP:

Dodge skill also aids a character in avoiding an attack entirely. The higher the character trains in dodge, the higher his chances of dodging the attack. A character singled in Dodge has approximately a 5% chance of dodging an attack outright in stance offensive. A character double trained in Dodge has a 10% chance, and tripled has a 15%. The exact values also depend on stance, Agility bonuses, Intuition bonuses, encumbrance, and spells such as Mass Blur. There is no available formula to determine the exact chance that an attack will be evaded, but character level plays into this greatly. Also, ranged weapons and bolts are harder to dodge than a melee attack, even though there's a DS bonus against such attacks (See the Defensive Strength section above).

It doesn't mention armor or shield choice, but we all know there can be inaccuracies in documentation.

Any insights?

[Fallen]

Race may also factor into the dodge formula beyond simple stat bias. Environmental room conditions may also be a factor. Creature vrs attacker size may also be a factor.

Just shooting in the dark here.

[droit]

That may be, but I'm more interested in whether I'm losing out on more than just a few points of DS by using a medium shield instead of a buckler, at least in terms of dodging.

[thefarmer]

I would think that the outright chance to DODGE an attack is lessened (as you go up in shield size) is balanced out with a better chance to outright BLOCK with your shield.

[Celephais]

That may be, but I'm more interested in whether I'm losing out on more than just a few points of DS by using a medium shield instead of a buckler, at least in terms of dodging.
AFK in rats for a few hours with a buckler. Then AFK in rats for a few hours with a large shield. Then settle the debate for all of us. :)

(If you need help parsing the logs, attach them in this thread and I'll parse out the percent dodged).

My only fear is that the extreme level difference will result in you reaching a capped % chance to dodge, despite any penalties...

[Celephais]

So I decided to do the legwork and had a good character to test it with.

Level 10 Half-Elf Rogue.

Strength (STR): 93 (21) ... 93 (21)
Constitution (CON): 63 (6) ... 63 (6)
Dexterity (DEX): 93 (26) ... 93 (26)
Agility (AGI): 95 (32) ... 95 (32)
Discipline (DIS): 73 (6) ... 73 (6)
Aura (AUR): 72 (11) ... 72 (11)
Logic (LOG): 55 (2) ... 55 (2)
Intuition (INT): 55 (2) ... 55 (2)
Wisdom (WIS): 61 (5) ... 61 (5)
Influence (INF): 37 (-1) ... 37 (-1)

Skill Name | Current Current
| Bonus Ranks
Armor Use..........................| 102 24
Shield Use.........................| 82 18
Combat Maneuvers...................| 102 24
Edged Weapons......................| 102 24
Ambush.............................| 102 24
Physical Fitness...................| 58 12
Dodging............................| 102 24
Stalking and Hiding................| 99 23
Perception.........................| 66 14
Climbing...........................| 25 5
That's 1.5x Shield, 2x Edged, 2x Dodge.

Wearing double leathers, using a small shield, no encumberance, stance advanced, being attacked by rats:

Evade: 234/ 524 (44.7%) ( 44.7%)
Block: 35/ 524 ( 6.7%) ( 12.1%)
Parry: 70/ 524 (13.4%) ( 27.5%)
AS/DS: 185/ 524 (35.3%) (100.0%)


Wearing double leathers, using a great shield, no encumberance, stance advanced, being attacked by rats:

Evade: 96/ 524 (18.3%) ( 18.3%)
Block: 79/ 524 (15.1%) ( 18.5%)
Parry: 100/ 524 (19.1%) ( 28.7%)
AS/DS: 249/ 524 (47.5%) (100.0%)

Wearing studded leather, using a small shield, no encumberance, stance advanced, being attacked by rats:

Evade: 224/ 524 (42.7%) ( 42.7%)
Block: 35/ 524 ( 6.7%) ( 11.7%)
Parry: 66/ 524 (12.6%) ( 24.9%)
AS/DS: 199/ 524 (38.0%) (100.0%)

Wearing studded leather, using a medium shield, no encumberance, stance advanced, being attacked by rats:

Evade: 297/ 684 (43.4%) ( 43.4%)
Block: 68/ 684 ( 9.9%) ( 17.6%)
Parry: 84/ 684 (12.3%) ( 26.3%)
AS/DS: 235/ 684 (34.4%) (100.0%)

Wearing studded leather, using a large shield, no encumberance, stance advanced, being attacked by rats:

Evade: 451/1462 (30.8%) ( 30.8%)
Block: 163/1462 (11.1%) ( 16.1%)
Parry: 226/1462 (15.5%) ( 26.7%)
AS/DS: 622/1462 (42.5%) (100.0%)

Wearing studded leather, using a great shield, no encumberance, stance advanced, being attacked by rats:

Evade: 120/ 652 (18.4%) ( 18.4%)
Block: 89/ 652 (13.7%) ( 16.7%)
Parry: 121/ 652 (18.6%) ( 27.3%)
AS/DS: 322/ 652 (49.4%) (100.0%)

Wearing Metal Breastplate, using a small shield, no encumberance, stance advanced, being attacked by rats:

Evade: 591/2579 (22.9%) ( 22.9%)
Block: 228/2579 ( 8.8%) ( 11.5%)
Parry: 488/2579 (18.9%) ( 27.7%)
AS/DS: 1272/2579 (49.3%) (100.0%)

The second percent is the percent of attacks left that were effected (and should be the value considered), essentially once you have evaded something, it doesn't bother checking to see if you can block or parry it.

The real value of concern for the character in question would be the AS/DS attacks that got through, and it's quite clear that a greatshield not only hurt the DS of this character, but hurt the EVP percent as well.

Anyone who is curious, I used the parser posted here: http://forum.gsplayers.com/showthread.php?t=33162

[Drew]

Good research Celephais.

[thefarmer]

Wearing double leathers, using a small shield, no encumberance, stance advanced, being attacked by rats:

Evade: 234/524 (44.7%) ( 44.7%)
Block: 35/524 ( 6.7%) ( 12.1%)
Parry: 70/524 (13.4%) ( 27.5%)
AS/DS: 185/524 (35.3%) (100.0%)


Wearing double leathers, using a great shield, no encumberance, stance advanced, being attacked by rats:

Evade: 96/524 (18.3%) ( 18.3%)
Block: 79/524 (15.1%) ( 18.5%)
Parry: 100/524 (19.1%) ( 28.7%)
AS/DS: 249/524 (47.5%) (100.0%)

The second percent is the percent of attacks left that were effected (and should be the value considered), essentially once you have evaded something, it doesn't bother checking to see if you can block or parry it.

The real value of concern for the character in question would be the AS/DS attacks that got through, and it's quite clear that a greatshield not only hurt the DS of this character, but hurt the EVP percent as well.


The block percentage went up (almost equal % to the evade drop)with the larger shield size, like I suspected it would to counterbalance the lower evading. Blocking is just as good as evading something, really.

So really I see it as only a DS issue.

[Celephais]

The block percentage went up (almost equal % to the evade drop)with the larger shield size, like I suspected it would to counterbalance the lower evading. Blocking is just as good as evading something, really.

So really I see it as only a DS issue.
You are terrible at math... the block percent went up 6% ... Evade went down 26%. Most importantly total EVP went down about 12%, which is very significant.

[thefarmer]

You are terrible at math... the block percent went up 6% ... Evade went down 26%. Most importantly total EVP went down about 12%, which is very significant.

I'll admit I'm not a math person.

However I used the first number blocked and the second, the same with dodge and it seemed fairly close the same up and down.

At 2x dodge, 1x shield, which I think droit is, he'd see less of a drop than 12%

[Fallen]

Dodge
Dodge Defense Strength (DS) bonus:
(Dodge Ranks + (AGiLity bonus) + (INTuition Bonus/4) + Spell Mods) = Base Value
Base Value * Stance Modifier * Shield Penalty * Armor Hindrance Penalty = DS bonus
Stance (http://www.play.net/gs4/info/combat_guide/factors.asp#stance) Modifier: 75% + Stance/4.
Example: In Stance Guarded (Stance = 80), the Dodge DS stance modifier would be 75% + (80/4) = 95%
Shield Penalty: Using a shield hinders attempts to Dodge. The penalty is 20% for a small shield, 30% for a medium shield, 40% for a large shield, and 50% for a tower shield.
Armor Hindrance penalty (http://www.play.net/gs4/info/armory/armor.asp#ActionPenalty): Wearing heavy armor hinders attempts to Dodge. The penalty is 1/2 of the armor's Maneuver Hindrance value.
The odds for an outright dodge are based on the defender's Dodge ranks compared to the attacker's level. It is affected by stats, certain spells, and stance. A character with 1x Dodge training can evade a like-level foe roughly 5% of the time in stance offensive. A character with 3x Dodge training and all other factors the same would have roughly a 15% chance to evade that same attack.

Evade DS bonus is increased by 50% against any ranged attacks (arrows, hurled, bolts). It is possible to evade ranged attacks completely, though the chance of a successful evasion is less than the chance against melee attacks. Spells that improve attempts to Dodge include Mobility (http://www.play.net/gs4/info/spells/spelllist.asp?circle=9#618), Mass Blur (http://www.play.net/gs4/info/spells/spelllist.asp?circle=5#911), Elemental Refraction, and Song of Mirrors (http://www.play.net/gs4/info/spells/spelllist.asp?circle=10#1019).

[Celephais]

I have added two more sets of data points by changing armor. I should be fully trained to remove all the hinderance from studded leather, and I am not fully trained to be wearing metal breastplate. Any statiticians tell me if 2% is a statistically significant number out of 524 samples? Do I need to get more data?

I'll admit I'm not a math person.

However I used the first number blocked and the second, the same with dodge and it seemed fairly close the same up and down.

At 2x dodge, 1x shield, which I think droit is, he'd see less of a drop than 12%
By dodging less the first number is artificially inflated... if you dodged zero attacks, you would have more opportunities to block and the number would be higher, this does not necessarily mean you got better at blocking. So you have to use the 2nd percentage to see your actual percent chance to block (same with parry, which should be constant throughout). I do agree that with droit he will see less than the 12% total drop (and really all you care about is the total). This would also be heavily influenced by the level of things attacking him, since I was fighting something 9 levels lower I was much better at EVPing, but the swing of 12% was more than a 33% increase in number of AS/DS results, which is massive IMO.

[droit]

Alright, I'm a bit drunk at the moment (where's fallen's "PUI" tag?), but let me see what I can pull out of the above research. The character Celephais tested this with is 2x dodge, 1.5x shield, correct? Just for reference, Droit is 1x shield, 1x dodge, but he receives 73 phantom dodge ranks from mobility plus the blurs bonus the majority of the time, so almost 2x in effective dodge ranks. So the difference between dodging and blocking is even more pronounced for him. Obviously, the small shield trumped the greatshield in terms of outright epb because the drop in evasion was more than the gain in block. That makes sense, but Celephais, would you mind doing one more test, this time with a small shield versus a medium shield? I'd like to see if the same inequity exists when only jumping one shield size.

[Celephais]

Alright, I'm a bit drunk at the moment (where's fallen's "PUI" tag?), but let me see what I can pull out of the above research. The character Celephais tested this with is 2x dodge, 1.5x shield, correct?
Woops, I wrote that he was 0.75 shield, and you drunkenly got it right at 1.5x shield...


Just for reference, Droit is 1x shield, 1x dodge, but he receives 73 phantom dodge ranks from mobility plus the blurs bonus the majority of the time, so almost 2x in effective dodge ranks. So the difference between dodging and blocking is even more pronounced for him.
Correct, it would only be worse.


Obviously, the small shield trumped the greatshield in terms of outright epb because the drop in evasion was more than the gain in block. That makes sense, but Celephais, would you mind doing one more test, this time with a small shield versus a medium shield? I'd like to see if the same inequity exists when only jumping one shield size.
No I wouldn't mind, but you're going to have to wait until tonight when I can run the test. Of course with my character being more shield oriented than yours, I wouldn't be surprised if my character benefited from a medium shield for total evp, while your character was detremented from it.

[DaCapn]

I have added two more sets of data points by changing armor. I should be fully trained to remove all the hinderance from studded leather, and I am not fully trained to be wearing metal breastplate. Any statiticians tell me if 2% is a statistically significant number out of 524 samples? Do I need to get more data?

Can you clarify where the 2% is exactly? It's been a little while since I've done this, but I think the statistical uncertainty in collection of N points (say, N evasion incidents for example) is sqrt(N). So to quote N with 2% accuracy, you would need sqrt(N)/N = 1/sqrt(N) = 0.02 or 2500 counts. So that would be for determining whether or not you could detect a 2% change in evasion events between the two armor trials. Anyway, it really depends where that 2% lies.

It would be quite an interesting piece of information to have accurately crunched out. All that is known about the action penalty for undertraining is that it is scaled down from the maximal value to the base value of the armor when fully trained. There's no notion of how it's scaled through undertraining, nor the plateau point. For the complete picture, it would have to be done for every AsG for every number of armor use ranks up until fully trained (which turns out to be roughly 800 trials).

Not to stray off topic, but after looking at the armor table a bit more closely now, is it just me or is it the case that there are NO ranks of armor use required for full leathers? It has a 0 action penalty, but an RT adder of 1. So doesn't that mean you could get away with 0 ranks of armor use and a high AGIDEX to get over the RT penalty? It would take 2 ranks to overcome RT penalties, but 0 are required for action penalty right? I'm just so used to seeing the "4 ranks for maneuver penalties" which is handed down as lore and I never actually looked up the action penalty for full leathers until just now.

[Celephais]

Can you clarify where the 2% is exactly?
The 2% swing came from small shield & double leathers vs small shield & studded leather evasive percent. 234 evasions out of 524 with double leather, and 224 evasions out of 524 with studded leather.

Using your formula it sounds like I'm accurate to 4.4% with 524 swings? Ugh... takes long enough to get those, for 1% accuracy I need 10,000 swings? Even if I had the right characters to test it with (haste swinging wizard around the same level as the dodger) it would take almost 3 hours per test.

I added Studded Leather/Medium shield to the mix... I'll retest tower shield with studded leather shortly (I decided this character should be wearing studded leather anyway, so I'd rather test with what he's going to use).

[Celephais]

:rofl: I'm pissing off the noobs in the catacombs.

Lestholas says, "I daresay you are becoming a neusance."

[Celephais]

Large shield studded leather added...

So far it looks like this character has the best DS and the best EVP using a medium shield. It also looks like the amount lost to being in studded leather isn't worth not being in it.

How that will pan out for someone who is 1x Shield and 2x dodge I don't know... but my gut is that if you're losing DS you're losing EVP.

[DaCapn]

Using your formula it sounds like I'm accurate to 4.4% with 524 swings? Ugh... takes long enough to get those, for 1% accuracy I need 10,000 swings? Even if I had the right characters to test it with (haste swinging wizard around the same level as the dodger) it would take almost 3 hours per test.

Well, if you're trying to determine your block percentage with uncertainty with this data set as an example:


Wearing studded leather, using a large shield, no encumberance, stance advanced, being attacked by rats:

Evade: 451/1462 (30.8%) ( 30.8%)
Block: 163/1462 (11.1%) ( 16.1%)
Parry: 226/1462 (15.5%) ( 26.7%)
AS/DS: 622/1462 (42.5%) (100.0%)

You have the following:
raw delta = sqrt(163)=12.767
percent delta = 12.767/(1462-451)=0.0126=1.26%
percent=163/(1462-451)=0.16122=16.1%
Block : 16.1 (12.6) % OR 16.1% +/- 1.26 OR 14.8-17.36%

I'm not sure if you thought you had more, less, or exactly the same precision. But anyway, I'm pretty sure that's how the error propagation goes.

Also, is that how EBP works? Evade takes precedence over block takes precedence over parry?

[Celephais]

Wait... you're taking the sqrt of 163? to determine the 1.26% +/-?

If I were to say I have a 16.1% chance to block, I could also say I have a 83.9% chance to fail my block roll. Why would my confidence be decreased and result in a 2.9% +/- if I did the math for chance to fail instead of chance to succeed?

The way I understood what you were saying was that if there are 1011 total rolls to block, you sqrt it (31.8) and take the inverse for a 3.1%. Which does seem rather high but... makes more sense to me at least.

[DaCapn]

If I were to say I have a 16.1% chance to block, I could also say I have a 83.9% chance to fail my block roll. Why would my confidence be decreased and result in a 2.9% +/- if I did the math for chance to fail instead of chance to succeed?

2.9% is a smaller fraction of 83.9% than 1.26% is of 16.1%. That's the delta, not the percent uncertainty. A delta has no true meaning unless it's paired with the base value. (Consider the phrase "I was $20 short" with regards to buying a car new off the lot, and when paying for this week's groceries). Your delta will always increase with a larger data sampling but the delta/base_value fraction will continue to decrease as the base value gets larger.


The way I understood what you were saying was that if there are 1011 total rolls to block, you sqrt it (31.8) and take the inverse for a 3.1%. Which does seem rather high but... makes more sense to me at least.

That means your 31.8 has an error margin of 3.1%, giving you a range of 30.8142-32.7858.

[Celephais]

2.9% is a smaller fraction of 83.9% than 1.26% is of 16.1%. That's the delta, not the percent uncertainty. A delta has no true meaning unless it's paired with the base value. (Consider the phrase "I was $20 short" with regards to buying a car new off the lot, and when paying for this week's groceries). Your delta will always increase with a larger data sampling but the delta/base_value fraction will continue to decrease as the base value gets larger.
I understand the proportional value, but the thing is you're saying that with the values I gave, I would have a 14.8-17.36% chance to succeed in blocking, but an 81.0-86.8% chance to fail in blocking.

[thefarmer]

Also, is that how EBP works? Evade takes precedence over block takes precedence over parry?

Dodge takes precedence over block and parry. I'm pretty sure Block comes before Parry, but I dunno.

[Celephais]

Dodge takes precedence over block and parry. I'm pretty sure Block comes before Parry, but I dunno.
Yeah EBP is in the order stated... and the numbers seem to support it. It's roll to evade, if fail, roll to block, if fail, roll to parry.

[DaCapn]

I understand the proportional value, but the thing is you're saying that with the values I gave, I would have a 14.8-17.36% chance to succeed in blocking, but an 81.0-86.8% chance to fail in blocking.

Oh, yeah, I see what you're getting at. Yeah, this model doesn't account for the fact that you know the full sample space of trials with absolute certainty (so there's a margin of error if you add up the max/min or min/max ranges of the success/failure percentages because it assumes that the total has some associated error). I don't know if I'm skilled enough to work that into the model. It's not the "back of the envelope" job that I thought it might be, at any rate. Well, if I can't work it out or if some stats pro doesn't come save the day, at least you can get some kind of rough estimate of your error here. If it's any consolation, the sum of those percentages (which lie above and below 100%) generate values that are about equal to the the error ranges predicted for the total sample size of 1011.

[droit]

Was the small vs. medium test done?

[Stunseed]

I'm interested in this as well. I went from 1x in shield use to near 2x, and I'm wondering what shield would help me out the best. I've got a 9x heater now.

[Celephais]

Was the small vs. medium test done?


I'm interested in this as well. I went from 1x in shield use to near 2x, and I'm wondering what shield would help me out the best. I've got a 9x heater now.


Wearing studded leather, using a small shield, no encumberance, stance advanced, being attacked by rats:

Evade: 224/ 524 (42.7%) ( 42.7%)
Block: 35/ 524 ( 6.7%) ( 11.7%)
Parry: 66/ 524 (12.6%) ( 24.9%)
AS/DS: 199/ 524 (38.0%) (100.0%)

Wearing studded leather, using a medium shield, no encumberance, stance advanced, being attacked by rats:

Evade: 297/ 684 (43.4%) ( 43.4%)
Block: 68/ 684 ( 9.9%) ( 17.6%)
Parry: 84/ 684 (12.3%) ( 26.3%)
AS/DS: 235/ 684 (34.4%) (100.0%)

1.5x shield, 2x dodge and medium shield is better EVP and DS than small shield. Then again, that's within the statistical +/- I'm sure (because I don't think evade would go up from switching from small to medium). If I get some more swings in, I'll bump w/ more accurate numbers.