dev1069 Phantom Roll Effect Adjustments feedback

[Byrth]

As I see it:
* Their point before was "Higher TH is better than lower TH."
* Their point this time is "Defense decreases damage taken."

I think those really might be cherrypicked single values. If they had been interested in giving us accurate game mechanics information in either case they wouldn't have had to do any experiments at all. They have the code, so they should be able to look at it and tell us the exact relationship between TH and drop rate, defense and damage taken, etc. Instead they do "experiments" with poorly defined methods.

For all we know, he is hand parsing this and has an N of 1 at each defense value. Like, pull a level 114 monster. It hits him once, he uses Defender. It hits him again, he uses BCB. It hits him again, he GM-kills it or lets it kill him.

I agree that it wouldn't be hard to test, though. You would need some way to keep your VIT/defense constant and survive something for a moderate period of time.

[Greatguardian]

If these "tests" illustrate just one thing, it's that even the JP community reps probably do not have any access to the code or the ability to gain such access from the Devs.

It wouldn't really surprise me, to be perfectly honest. Sometimes (Er, more like every time a Dev post is translated) you're left to wonder how much the Developers even understand about the mechanics of their own game. I'm still working under the impression that they fired off their math geeks after the code base was finalized and only left Designers on the team, which would explain oh so much about the past 5 years.

Imagine trying to update an IE4 compliant website to work with Chrome, Firefox, iOS, Android, and IE9 with a staff that only knows how to use Macromedia Dreamweaver 6.0 MX.

[SpankWustler]

If these "tests" illustrate just one thing, it's that even the JP community reps probably do not have any access to the code or the ability to gain such access from the Devs.

The more I think about about this sort of thing, the worse I feel for the community teams of all nations. They've gone out of their way to provide information on multiple occasions, and all we players can often do is say "Thanks but we know that already." It seems like we have access to the same testing methods, sadly, if not quite the same information and resources.

I was pretty snappy about a possible sample size of one earlier, but now that I think about it in this context, I wouldn't blame the guy if that were the case. There may be deadlines for replies; getting hit 5,000+ times takes a while.

There should be accurate spreadsheets or other plug-stuff-in game mechanic simulations available for these folks who have to discuss game mechanics with a group of people who are dedicated to complaining about game mechanics.

[Motenten]

Post with testing: here.

Conclusions:
Mobs get positive level correction of +0.05 per level above the player.

The 1.0 cRatio floor exists, but possibly only for pDif values below cRatio. This part was a bit weird.

Edit: Further testing indicates we can't reach a conclusion about a cRatio floor just yet.

Edit2: There is clearly not a cRatio floor; however the formula for the pDif range changes, and it's not the same formula as used by players.

[Tetsujin]

If defender (an increase of 25% defense) only dropped off roughly 30 points of damage, that's proof enough for why the defense stat feels so weak.

[Motenten]

More thoughts.

When considered in isolation, a defense boost is perfectly fine as a buff. It reduces damage taken at least some, and in a consistent, which is better than nothing, and the amount reduced is fairly comparable to the original proposed phalanx effect (at least for high-level mobs), which no one really seemed to have serious issues with. Much of the complaining, however, is because the players are considering things with respect to the larger picture -- what do you have to give up to get this boost?

As mentioned, with the Protect line of spells you give up nothing, so it's almost always worthwhile to use them (at least til you're fighting EP or lower mobs). With a Cor roll, however, you are intrinsically giving something up -- the option for a different roll. Plus, Cor doesn't get an analog to Pianissimo; if you're giving a buff to one person, you're probably giving it to at least half the party (can still keep the melee group separate from the mage group).

There are two things you're giving up with a defense roll: The potential for more attack of some sort (which is a fairly substantial loss if it's hitting the DDs as well), and the enmity that goes with the extra damage.

As for the attack itself? Honestly, if you triple a pld's damage output, you probably won't reduce the fight length of these tough fights by more than 5%. The pld will certainly take less damage overall due to the increased defense than due to faster kill speed if you changed rolls. The real problems are the secondary factors:

1) That damage done translates directly into enmity, and that doing damage is by far the best way to generate enmity in the game.
2) That placing the roll's effect on the DDs (due to not being able to single-target buffs) severely hampers their damage output potential.

It's possible to work around the second issue above through careful positioning (we did it often enough when levelling up pre-Abyssea). Annoying, but not crippling. The first issue, however, cannot be ignored.

Almost all the problems pld suffers from stem from deliberately inadequate damage output potential, combined with no other means of generating enmity that is anywhere near comparable with doing damage.



Camate mentioned a max of about +30% defense on an 11. I'd hope that that's without the job boost effect. In other words, I'd expect the defense bonuses on Gallant's Roll to basically mirror the attack bonuses on Chaos Roll (so up to +41% defense on an 11 with pld in the party). That would give a 17%-27% reduction in damage taken for the example scenario (though that's made somewhat less effective when combined with PDT gear).

If that were the intended buff levels, having reviewed the math for the resulting effects, I'd say that this is a fairly decent overall buff, *as long as* there are also plans to address the enmity issue for pld. Well, to be fair, it's a decent buff either way; it's just won't be used until those other issues are addressed.

[Chamaan]

Take a look at the stats listed in the update notes for today. Either we've got a bad translation fail or we may have won this fight.

=EDIT=
False alarm. Defense. Enjoy your worthless roll, boys.

[Motenten]

I'd also like to hear an explanation for the difference in the defense boost for Gallant's vs the attack boost for Chaos. A given amount of attack is -always- better than that same amount of defense, yet Chaos's percentages are about 30%-50% higher than Gallant's.

Lucky Chaos is +25% attack; lucky Gallant's is +15% defense
11 Chaos is +31% attack; 11 Gallant's is +20% defense
etc.

Both get a +10% bonus for job in party.

If we're going to get a defense boost, why is it not at least on par with the corresponding attack boost roll?

[Motenten]

Aside: Mocchi made a lengthy post detailing damage results for various attack/defense conditions vs Ig-Alima this morning (with specifics such as it being level 120 with 1059 attack). I hope to see a proper translation of it soon.

[Motenten]

~~ Review of Mocchi's post based on Google Translate


Known info:

Mob attack: 1059
Mob level: 120

Baseline case (normal attack), damage done vs.:
365 def (Berserk): 645
460 def (neutral): 645
575 def (Defender): 614
630 def (Defender+food): 563
831 def (Defender+food+Gallant's): 504


There is no difference in damage taken between 460 (neutral) and 365 (Berserk up). There's a slight reduction in damage taken between 460 (neutral) and 575 (Defender up).

From this we can deduce that it's using 1-hander mechanics, with a capped Ratio of 2.0 vs a defense of 1059/2 = 530.

Given the damage done with capped cRatio (645), and that there should be 1.05 worth of level correction for a mob 21 levels above the player, we can figure the base weapon damage for the mob.

645 / 3.05 = 211.5

This actually doesn't fit with either 211 or 212 cleanly, so it's likely there's actually a cap on level correction of +1.0.

645 / (2.0 + 1.0) = 215 (exact)

We can then see how things calculate out in the forward direction.

1059/575 + 1.0 = 2.84 * 215 = 611 [vs 614]
1059/630 + 1.0 = 2.68 * 215 = 576 [vs 563]
1059/831 + 1.0 = 2.27 * 215 = 489 [vs 504]


Note that these values do not match the results given by Mocchi. It's possible he's using the average pDif value rather than cRatio, since those are slightly different, but there's no way to really tell for sure. It's probable that there are other elements not being accounted for in the player model. At the very least, the vit from the food may have affected the value.



Additional issues:

The defense values do not add up.

Base def is 460
Def with Defender is 575 (460 * 1.25)

Def with Defender + food is 630, which means + 55 defense. That seems extremely low, especially with tacos not capping til +150 defense, and even fish mithkabobs capping at +90 def, but the wikis don't list that data. The actual percent increase for BCBs is +16% accoring to wiki, which would give +73 defense if applied only to base defense and not capped.

Def with Gallant's is 831, which is +201 over Defender+food. To get +201 defense with +30% defense (roll of 11 with pld), you need 670 base defense. That's not possible if Gallant's is part of the same term as Defender, in terms of +def%. Even if it stacks on top of Defender+food (which would increase its value), the total for those together is 630 defense, so it's still insufficient to explain the final total.


It could conceivably be that the percent increase on BCBs is listed incorrectly on the wiki, and that it's actually +12%. However no matter what juggling I do, I can't get the defense totals to add up. The closest is Base defense * 1.25 (Defender) * 1.3 (Gallant's) * 1.12 (BCB) = 836, though if you did it that way and also added in the 2 def from the 4 vit of the BCBs you'd end up with 840.

Camate -- please pass through a request that Mocchi review the numbers he used for his illustration post.