Charm, Root and Lull Resist Mechanics
 

Charm works like this: (although minor unknown unknowns may exist)

On spell lands, the target's resist is modified by (CHA - 75) / 8, max 25. So if the enchanter has 275+ CHA then they'll get a -25 MR bonus to casts, making it considerably easier to land. If CHA is under 75, there is no penalty to having lower CHA.

Charm tick saves are handled this way:

• Tick saves in EQ do more than one roll. The first roll is a chance to roll against the MR. This roll is 50% for charm. If this roll fails, charm will hold without checking MR.
• The caster gets a +4 level bonus when rolling against MR. This makes charm hold better on targets near the caster's level.
• Resist adjust modifiers on the spell DO apply to tick save throws. (Boltran's has a -10 resist mod)
• The target's effective MR resist is floored at 5 points, and that is what causes a break chance on an otherwise 0 MR target.
• Live's resist roll is 1-200. Al'Kabor's was either 0-200 or 0-199. Both use a >, not a >=. This means the charm break resist check is 2.5% for Live and 3% for AK. (since there is a 50% prior roll, break chance per tick is half these %s before TD)
• Charisma DOES NOT affect the target's chance of breaking charm at all.
• Total Domination is merely a 15%/25%/35% chance (for ranks 1-3) to keep holding if the caster would otherwise lose the tick save throw.

Root tick saves work like this:

• Like charm, root has a preliminary roll. This roll is 75%. If this roll fails, root will hold without checking MR. I.e. a 1 in 4 chance to NOT check MR.
• The caster gets a +4 level bonus when rolling against MR like charm does.
• Resist adjust modifiers on the spell DO apply to tick save throws.
• The MR resist is floored at 5 points, and that is what causes a break chance on an otherwise 0 MR target.
• Live's resist roll is 1-200. Al'Kabor's was either 0-200 or 0-199. Both use a >, not a >=.

Lull fails work like this:

• When a lull spell lands on a NPC, the NPC's MR is ignored and instead 15 MR is used.
• The 15 MR is still modified by level difference, meaning mobs 6 levels and under below the caster will never aggro.
• If the spell resists, then a second roll is done using the caster's charisma. This roll is: 90 - CHA / 4. Failing this roll results in a critical fail and the NPC aggroing on the caster; otherwise the NPC will ignore the resist.
• If there is a cap to the charisma critical failure chance, then it's not seemingly hit at 305 CHA. I did not parse beyond 305.
• Lulls have a chance to fade early. The precise logic is unknown, however parses resulted in this:
  • 7% per tick on a +5 a red con
  • 2% per tick on white cons
  • 1% per tick on a -1 blue.
  • 0% per tick on a -5 blue

Evidence For These Claims

I have two primary sources for these claims. I recently automated casts on NPCs for weeks and I was also given information from two highly respected members of the Al'Kabor community who got information years ago from quote 'somebody who ought to know'. They are essentially intermediaries to a real source, but for simplicity I'll refer to these two individuals as 'my sources' for the rest of this thread. Once I was given the information from them I set out to try and verify it with Live server tests, which wasn't simple because they misinterpreted some things, but I was able to figure out what they mistook. I also have some secondary sources, including public dev comments. (which are remarkably inaccurate, but did help in some ways)

I was told by my sources that charm's save throw preliminary roll is 40%, the minimum 'can't go below 5%', the resist check has a +4 level mod, the CHA bonus is CHA - 75 / 8 (max 25), and TD is 15/25/35%. At first I was excited to see this and plugging that into my simulator produced results close enough that I thought it was accurate. Furthermore, the 5% minimum had been mentioned in a 2015 dev post and TD %s were added to the client AA description some years back, so that matched and gave it credibility. I knew that a preliminary roll had to exist simply because 5% by itself didn't work. It looked really legit.

I decided to run tests on Live and also squeeze more AK data from old logs to verify the claims using log parsing scripts I had written. It didn't take long to realize that it didn't work. It was kind of close but too far away to be right. After weeks of data collection on Live and playing with with simulations to try and match it, I found out that the best fit to my Live data is 47% with a MR floor of 5. This got my sources talking more and they mentioned a 'Rand0()' function to me. They got the preliminary roll wrong because they assumed that 'Rand0(4)' meant roll a number from 0 to 4, so comparing that with a 2 (they mentioned a '< 2' here) would result in a 40% per value instead of 50%. EQ's random functions are known to use 0 to n - 1 as client decompiles show. (thanks to Mackal for the assist here) Root data also proved the n - 1 to be correct, so charm's preliminary roll is 50%.

The '5% minimum' could also mean one of several different things. The two most obvious would be either a MR floor of 10 (since the resist roll is 1-200) or a separate 1 in 20 roll. Neither of these worked in simulations. Trial and error to get the simulations to fit the data resulted in an MR floor of 5, but that's not 5%. Further correspondence with my sources later mentioned quote: "I got the strong impression that the person we talked to at least thought that the 5% minimum was not a separate roll but tacked onto the end of the resist function itself." So that strongly implies a MR floor instead of a separate roll. I tested this on a 5 MR NPC and the result was an average duration within the margin of error of a 0 MR NPC, which confirmed it since a separate roll would result in a much lower average.

There is a 2015 dev post here: https://forums.daybreakgames.com/eq/.../#post-3678593
That says: (I added what the SPAs are)
Prathun's pseudocode says 'resist chance' in place of the variable that is in actually the target's effective resist VALUE. Since the roll is 200 and not 100, it's essentially the resist chance multiplied by 2. Sony's resist code likely names their effective resist value variable 'resist_chance' or similar. In fact my sources told me: "We can confirm that the general resist check contains something along the lines of "if resist_chance < 5, then resist_chance = 5"" Niente and my sources' contact seem to both have failed to take that into account, resulting in erroneous statements. Even had they said 2.5%, (AK is 3%) they'd still be misleading because charm breaks are two rolls (3 with TD) which need careful explanation and the preliminary roll was omitted entirely in Niente's post. Live's minimum break chance per tick is actually 1.25%. (without TD) There are 10 ticks a minute so at 5% it clearly doesn't work.

Prathun's pseudocode link for reference: https://everquest.allakhazam.com/for...01109310546959

Niente also states that charm tick saves get a CHA bonus. This is incorrect. I ran tests at various CHA levels and they resulted in the same duration charms with large data sets ruling out margin of error. (well over ten thousand casts) In fact due to margin of error, the 15 CHA log resulted in the longest charms in my unicorn data. Niente also implies due to her wording that the CHA bonus to charm caps at 200, which is also untrue for charm lands; although perhaps she just meant a 200 cap for blurs or meant CHA above 75. These are good examples why developer comments should not be taken as gospel and that hard data should be the final arbiter.

Geoffrey Zatkin answered a question in 1999 where he claimed that charisma had a role in charm durations. He was also either wrong or charm changed before PoP went Live. He also mentioned a softcap at 200 CHA in the year 2000 which is not observed in either charm or lull, but I couldn't say how CHA works for mez and mem blur. As a developer myself I can attest that it's easy to get stuff wrong when taking a cursory look at code, so disputing developer claims isn't as presumptuous as it seems.

https://www.tapatalk.com/groups/othe...sma-t1046.html
https://web.archive.org/web/20030611...60&Action=View


Charm Data

Now on to the raw data. I'll start with a link to 2003 data from the enchanter forums on Wayback: https://web.archive.org/web/20050224...pic.php?t=1148

I was very fortunate to find that link while in the middle of my data collecting, because not only is it in and of itself convincing evidence for CHA doing nothing to modify charm durations, but also dates this to 2003, which is the era I'm most interested in; and since this agrees with my Live data, it suggests that charm has not been modified much since then. This guy ran a careful test and lists individual charm durations with known de/buffs and charisma, so this is good stuff. (note that the character's name is Yandie in the table below)

My data comes from two primary sources: Al'Kabor logs, and Live tests I conducted using automated casts. I wrote a script to parse logs to extract charm data, which I will include below. I will present the data in table form because there is a lot of it. The script output looks like this however:

https://i.imgur.com/d0oRjiy.png

Note the following:

• I am able to determine an NPC's resists by parsing spell casts on it. Generally I use a wizard's temperate flux staff for MR. For all-or-nothing spells, the effective resist value is merely resist rate * 2. E.g. 50% resist rate = 100 MR.
• The resist modifier from level difference is: diff^2 / 2, capped at -40. Charm ticks add +4 to the caster's level here. For example, the Crystalline golem's effective MR vs a level 65 would be: 50 - INT((65-62)^2 / 2) = 46. But for charm ticks: 50 - INT((69-62)^2 / 2) = 26.
• The '1-200' and '0-200' columns are the results of simulations (the average of 150,000 charms) using those rolls and with a MR floor of 5. ERV = effective resist value.
• Check the golem and unicorn data-- which were done at various CHA levels-- to see that CHA is modifying the target's MR on lands and not ticks. There are many thousands of casts proving this. I wore a significant amount of -CHA gear to get down to as low as 15.
• The 5 MR parse of the Crystalline Arachnae more or less proves that they are using a MR floor and not a separate roll.
• The AK data is clearly lower than the Live data. Since a 5 MR floor is very sensitive to how the 200 roll is done, I endeavored to try and determine precisely how Sony does this, which I outline here: http://www.eqemulator.org/forums/sho...6&postcount=19 The evidence is strongly in favor of AK having had a 0-200 roll and Live using a 1-200 roll and would explain why AK charms were somewhat shorter. Rashere's 2006 resist tinkering could have been the point it changed.
• Boltran's resulted in longer charms on a 50 MR green NPC, proving that the spell's resist adjust works on tick saves. Level advantage caps at -40, so the spell's resist adjust is lowering it down to the 5 MR floor.
• The simulations don't fit the unicorn data very well. It's possible that some minor factor is not being accounted for that only shows up when parsing very short duration charms, which may be from how the data is collected or simulated or Sony is doing something I'm unaware of.
• The PoValor and HoH NPCs had higher MR in 2003 and on AK. I estimate it was 65. Today it's 50. At some point Sony reduced resists on almost all PoP NPCs (perhaps in 2006 by Rashere)
• Some of the AK data is better than others, as it was from logs of characters leveling up and not clean automated tests like the Live data. The Herv and Torvie data is pretty solid however, as is the 150 cast druid sample, which was on a green mob and farming in a low level zone. The other druid logs should be considered less reliable.

The logs I created from automated casting on Live servers can be downloaded here: https://drive.google.com/drive/folde...Gg&usp=sharing


Root Data

Root and charm work in a similar manner. My sources claimed at first that root's preliminary roll was 60%, but once the Rand0() mistake was corrected, the roll is now known to be 75%. i.e. Rand0(4) < 3. Simulations using a preliminary roll of 75% and a resist floor of 5 matched my Live data very well. Since this fit so well, this bolsters the claims made for charm.

Here is my Live data with simulations under each data set for comparison. Note that 'Test' NPCs on the Test server have a base MR of 50 and that root also gets a +4 level modifier.

Shout-out to Kayen who almost nailed the root logic years ago from parsing it. He had come up with 70% + a 5 MR floor, so the emus had it pretty close already. His charm logic worked pretty well too.

I did some root early break logs awhile back (breaks from spell damage) but I can't seem to find them. I believe the chance was 50% or around there on white cons with some scaling to level difference.

The root logs which produced the above data can be downloaded here: https://drive.google.com/drive/folde...4Q&usp=sharing


Lull Data

Prathun's resist pseudocode had the 15 MR substitution for lull spells in it, so that part was easy. Note that this lull behavior was the result of a patch in the late Luclin era and is different for most of Luclin and prior eras. They might have just did the 15 MR substitution then and the rest of this applied to prior eras but I have no idea.

My sources gave me the lull formula of 90 - CHA / 4. I had already done lull parses in 2015 for TAKP so I had data on-hand to compare with and it matched. The formula I came up with for TAKP was only slightly different. This matching also bolsters the charm and root claims.

Here are the notes I wrote 6 years ago:

Those results suggest that a charisma cap for lulls is not met at 305. Since Al'Kabor's max CHA was 305, they had no reason to cap it then if developers wanted some low minimum fail chance, so my source would not have mentioned a cap but they might have added a cap in later years. I have no idea if lull becomes unaggroable at 360 CHA or not on Live. I'm sure players of later content could answer this.

Lulls can fade early and this is easily seen with 'worn off messages' on Live servers and easily tested for on Test server Arena mobs. My 2015 notes on this are:

"Chance is not affected by MR or charisma.
On Live, fade chance per tick was about 2% per tick on white cons, 7% on a +5 a red con, 0% on a -5 blue, and 1% on a -1 blue."

This data could be fleshed out more, but for TAKP I'm currently using: fadeChance = GetLevel() - caster_level + 2; in DoBuffTic()

My lull logs can be downloaded here: https://drive.google.com/drive/folde...dA&usp=sharing


Additional Claims

My sources claim that blind and fear also use a 75% preliminary roll, and that blind also has that 5 MR floor; however they're unsure if fear uses a 5 MR floor. Awhile back I did raise blind's preliminary roll on TAKP to the same level as root was (70%; but I'll raise them to 75% very shortly) because I did some crude tests on Live and it seemed to need to be raised, so that agrees with the claim. Beyond that I have no data to share.

I use lua scripts to parse logs or run simulations. I will paste them here.

This is my charm simulator:

This is my charm log parser:


This is my root simulator:


This is my root log parser:

Is it possible to get a Memory Blur/Mez addendum to this?

I have an update to this thread as I've done extensive research on the original resist system over the last year, and some of the findings apply to modern EQ. See my new thread about it, which has charm root and lull information in it.

A summary of discoveries that are relevant to this thread:

• Blind is confirmed to use a 5 MR floor with a 75% preliminary roll.
• Fear is confirmed to use a 75% preliminary roll but the 5 MR floor on it was removed at some point. (perhaps Sept 4 2002)
• Lulls also use a 75% preliminary roll. The fade early chance is a tick save resistance check.
• EQ's original spell resist system (pre-Sept 2002) had several qualities that made all of these spells much less effective.
• Charisma did reduce charm tick save MR in classic.

I did some parsing recently of fear and blind spells. The results were conclusive:

Since these spells all work the same, the root simulator script can be used for all of them. I also made a log parser for fears and blinds. The results match +4 caster level on tick saves with a 5 MR floor, except in the case of Fear the floor is gone. Note that, on Live, the spell durations will end up 0-7 seconds longer than the spell is set to be, because the amount of remaining time left in the server's current tick when the spell lands doesn't count as one of the spell's duration ticks. For my simulator I just added a 50% chance to add a tick on casts to account for this.

It was recently discovered that the oldest EQ clients (in classic and Kunark) had most of the server-side resist logic in them from the period. We've decompiled these and made discoveries from it. I found the lull tick save roll in them, but the other tick save rolls were not found. This is how I know the lull tick saves are 75% +4 caster level like the others. For example, a white con had a 2% chance to fade early in my tests, and a white con would have a lull MR of 15 in modern EQ when the spell lands. But for tick saves we add 4 to caster level, so the MR becomes 15-8=7. That's a 3.5% resist chance, now multiply by the preliminary roll: 0.75 = 2.625% chance to fade every tick.

The original resist system had two big things about it that made these spells worse. One is that the resist roll used to be 0-99 instead of 1-200, meaning the tick save floor of 5 resulted in twice as many resists. (actually a bit more than 2x)

The original resist system also had a level difference floor that could not be debuffed under, and this floor was often 10 if the NPC was anywhere near the player's level. This floor applied to tick saves. The consequence of this floor combined with the 0-99 roll is that tick saves broke the spells four times as frequently as they do in modern EQ for NPCs near the player's level prior to September 2002.

Lulls in classic EQ did not have that 15 MR override. Instead they had a floor that went up the higher level the target was. This tiered floor was very unforgiving and lulls in classic EQ were pretty much unusable. NPCs level 50 and higher were outright immune to lull before September 2002.

The classic resist logic clearly shows charisma modifying charm tick save MR. This was seemingly removed on Sept 4 2002.

See my other thread for details about the classic resist system. It also has some mem blur details that apply to modern EQ.