Mob difficulty isnt scaling with character
 

Mobs in general don't behave like they should. They act as if they are the same level to you no matter what your level is. So say you are lvl 5 and the mob is lvl5. It should hit you like the chart I posted representing live and eqmac. Also that mob if its not undead it should flee at 12% or w/e it is. When the mob is GREEN to you, say you are lvl30 and the mob is lvl5. It should hit you, if it doesnt miss, for 1ish. very rarely does a green mob of that level difference actually hit you for more than that. And it should flee at 20ish% hp not 5 or 6 like they are currently.

My best example would be to take an averaged geared toon at lvl 30 into kurns tower on the first floor. At lvl 30 you should not get aggro from any of the skeletons, but you get aggro right away. Second, they should almost always hit you for 1 at lvl30. most of the time they are hitting you for the same as if you were lvl14.
Hits on eqmac
http://i35.servimg.com/u/f35/14/25/38/02/hits_o10.jpg
Hits on live
http://i35.servimg.com/u/f35/14/25/38/02/hits_o11.jpg
Hits on a 0.8.0 server
http://i35.servimg.com/u/f35/14/25/38/02/hits_o12.jpg

Does anyone know how to fix this?

All of these were done with an iksar monk no gear all same starting stats. had max defense on the 0.8.0 peq server because i could set the skills. The other servers i didnt have quite max defense for lvl 5. but you can see that its still way worse here.

Here is the tanking data from 0.8.0 peq server 100 spiderlings.

/GU Tanking summary for: Testmitigation --- Total damage: 4510 --- Avg hit: 5 --- Swings: 949 --- Defended: 23 (2.4%) --- Hit: 752 (79.2%) --- Missed: 174 (18.3%) --- Accuracy: 81.2% --- Dodged: 23 (2.4%) --- Parried: 0 (0%) --- Blocked: 0 (0%) --- Riposted: 0 (0%) --- Absorbed: 0 (0%)

Here is the tanking data from live 100 spiderlings.

/GU Tanking summary for: Valkezisa --- Total damage: 2993 --- Avg hit: 3 --- Swings: 1322 --- Defended: 23 (1.7%) --- Hit: 754 (57%) --- Missed: 545 (41.2%) --- Accuracy: 58% --- Dodged: 23 (1.7%) --- Parried: 0 (0%) --- Blocked: 0 (0%) --- Riposted: 0 (0%) --- Absorbed: 0 (0%)

Here is the tanking data from eqmac 60 spiderlings.

/GU Tanking summary for: Osirous --- Total damage: 1033 --- Avg hit: 2 --- Swings: 724 --- Defended: 6 (0.8%) --- Hit: 381 (52.6%) --- Missed: 337 (46.5%) --- Accuracy: 53.1% --- Dodged: 6 (0.8%) --- Parried: 0 (0%) --- Blocked: 0 (0%) --- Riposted: 0 (0%) --- Absorbed: 0 (0%)

EQMac data. Was only able to get 60 spiders before leveling to 6 and you dont lose exp til 10 on eqmac. I might make another monk and level it to 5 to get the rest but the %'s are fairly close to live /shrug.

Meant to put this in support:general can someone move this please.

Unfortunately an enormous amount of data would be needed to remodel the formulae, and if adopted it's a significant impact to established servers.

Fleeing is an optional setting for NPCs that can be set in the npc_types table. You can enable or disable it depending on if you want the NPC to flee or not. There are a few flee settings in your rules table that can be adjusted to however you want them. Here are the defaults from the source:

Database settings will override the source defaults, so check your rule_values table settings and adjust as you wish.

I believe skeletons (maybe all undead) will attack you no matter what level you are. This mimics how it works on Live, unless that has been changed. On EQEmu, I think you just set the intelligence below 60 (or something) if you want a KoS NPC to attack players even if they are green/grey cons.

As far as adjusting actual combat damage/hits, that is a bit more complicated. You can adjust multiple stats in the npc_types table to get your parse results closer to Live. Most likely, it would be hard or impossible to make them match Live exactly unless you do source code edits to the combat code itself. Since we do not have the actual formulas for every factor into combat, it is nearly impossible to match Live exactly. Much of the combat code for EQEmu was created by guessing formulas that got as close to the desired results as possible within a reasonable amount of time investment.

If you can find or figure out all of the exact formulas to mimic Live content, then I am sure they can be added to the source. But, without adjustments that are huge improvements toward mimicking Live, it is generally not worth the headache it will cause all server admins as noted by jsr in the above post. The core combat system has had major adjustments to it over the years and each time it requires manually adjusting nearly all content on all servers to balance content out again. This is something that is not well received by server admins due to the huge amount of time it takes to make the adjustments.

The problem with all of those settings I believe is a result of the mobs not taking into consideration the players level. Mobs flee at different hp percentages based on the players level so just changing the data in the npc table would just make it worse. Not only is it flee percentage and maybe kurns tower is a bad example due to them being undead but the fact that they hit you as if you are the same level or lower level than the mob when in fact you are much higher level means that the problem is not with fleeing or mob damage but actually that the mobs just do not take into consideration what your player level is. Also I have noticed that p99's servers do not have this problem. So obviously nilbog or rogean knows how to fix this.

There's a lot of stuff P99 and SoD admins know how to do that they won't share. Get used to it.

Fleeing on EQMac is not an exact percentage, and is certainly a range. I spent a ton of time working on another project using EQMac, and found that NPCs will flee at random percentages, even at the same level. The highest percentage I saw was 32%, the lowest was 3%. These are not estimates, I have access to the exact health percentage of the NPC. It is true that health updates play a part in it, but there are several times when I was fighting a NPC with 9% or 10% health and I felt it should have started fleeing. I hit it again, the update took place, and it fled at 3-5%. Then other times, the NPC fled at a high percentage - same level or at least level range. My project only involved NPCs level 10 and under, and my character was in the same level range. One major problem with EQEmu is that NPCs don't seem to flee as often as they do on Live/EQMac. Though, the rules may be able to help that some.

As for combat, my own parses has shown there is a considerable difference between PC Live and EQMac. Basically, on EQMac the NPCs hit you more often and harder at all levels using toons with comparable stats/equipment. Live is what we are emulating, so that is going to be our model.

I will agree that combat doesn't feel exactly right on EQEmu at the lower levels, although it certainly is closer to Live than EQMac which is intended per our mission statement. At the higher levels however, EQEmu is very accurate I feel at least compared to Live. Parses confirm this. So, the problem isn't that NPCs don't take player level into account (they most certainty do) but rather, our formulas do not scale very well at lower levels.

Yes, undead and a few others on Live are still dumb and will attack you regardless of your level if you are KOS. The only problem with EQEmu in that regard is on Live it seems to be there is a chance it can attack, on EQEmu it always does. This problem also effects Threateningly mobs (they should roll to see if they attack you, not just attack like on EQEmu.) Scowling NPCs that are within your level range always attack on EQEmac and Live (although on EQMac there sometimes is a delay for the update to occur) EQEmu is correct in that regard.

They do. I can't remember if they did before I implemented heroic charisma or not but it is in:

aggro.cpp (320)
edit: Although I guess it depends on what the value of THREATENLY_ARRGO_CHANCE is(too lazy to look it up). But the framework is in for randomness.

Looks like it is 32:

http://code.google.com/p/projecteqem...G=Search+Trunk

WTF is Arrgo? :P

For the record, it wasnt me lol. The whole thing is messed up... THREATENLY != threateningly

Cavedude when you say higher levels do you mean 90 or 60?

As you can see from the charts I posted that mobs hit you for max more often than all others. So is there a way to add something to hit like random int 0-99, when 0-40 hit min, 41-85 hit avg, 86-99 hit max?

With this in place you could put in rules values which percentage ranges you want to hit for min avg and max or just a true false for people that dont want it on and people that do. Just a thought.

I was thinking more about it and in addition to what I posted above you would need different percentages for different level ranges. Something like

11+ levels would be 75%(to 90%) of the time hit for max
5+ levels 50% of the time hit for max
3+ levels 35% of the time hit for max
EVEN con would be what I posted above
dark blue 10% of the time hit for max
light blue 5% of the time hit for max
green/grey 0% of the time hit for max

The existing formular is far more complicated than that. What you've suggested is to throw away AC, AA's, and other things and instead just use level.

These aren't simple formulas and as Trevius already said, nothing is going to change unless you're able to propose something which makes significant advances towards being live-like. You'd also have to prove this, and that it scales properly across all levels.

If you're prepared to parse and analyse a million or so fights at various levels with various AC values.. That's probably the level of effort you're looking at. (I don't think I'm exaggerating too much :))

no you will still have ac and aa's. AC determines how much you get hit for and not what the mob is going to hit you for. If that was the case then there wouldnt be min and max hit tables in the npc_tables. it goes like this. mob rolls rand int to determine if you get hit for min avg or max. then it checks ac and aas and what not.

Also if you would actually look at the graphs i posted you can see that mobs are hitting for the exact opposite of live. being that they hit for max the majority of the time on eqemulator as opposed to min on live. These were done with a naked character so no extra ac was involved. All had the same starting stats and same class and all were the same level on the same mobs.

Say the max hit in the npcs table is 200. So if the rand int rolled 99 it would look at the table and see that the most it could hit for is 200. Then it would check the character being attacked AC defensive AAs and what not to determine how much dmg is actually done to the character. This of course is only on a successful hit towards the player character.

Dude you have 3 parses from a level 5 toon and all you are suggestions are hugely generalized. Go look at the code, find out where its inaccurate, parse way more data and then came back with update formulas. Nobody is going to do all that work just because you spent 5 minutes parsing a low level toon, so if its important to you... have at it.

Hmm.. actually I did look at the graphs and they are interesting as far as level 5 goes. Unfortunately all they do is demonstrate a problem with the current formula, at those levels, without any variables except defense skill and level. You have succeeded in pointing out an issue.

I'm not really sure what you mean by "AC determines how much you get hit for and not what the mob is going to hit you for" .. For what it's worth I have worked on the combat code and have a thorough understanding of it.

At the end of the day no-one really cares about level 1-5 if the existing formulae work at the high end. Whether you think people are listening or not, the bottom line remains that you either come up with a workable formula and prove that it is significantly more accurate than the current, or if you want someone else to come up with a formula I suggest parsing ~10,000 lines of combat for each level, and finding the common min hit AC and common max hit AC thresholds at each level. That sort of data would be useful in trying to reverse engineer the mitigation formula. The graphs you posted highlight an issue at level 5, but they don't offer much in the way of solutions. The only reason I continue to respond is for the off chance you would actually do this :P I'm happy to build models if there is sufficient data.

Heres the question I have. When you say high end or end game what do you mean? I am looking at this for server that will have a max level of 60. Do you mean 85 for UF or 90 for HoT? Do you mean 65 for GoD? I have lvl90 toons on FV so I could get the data for that. They have a mix of hot uf and VoA with max aas. Gear should not be a problem to attain the data from live, but if you are looking to balance lvl65 and I am looking for lvl60. From what you have said so far about my data only showing lvl1-5 and it wont work for balance high end then that means that I would want to collect data for max lvl60 and you would want me to collect data for whatever your high end is. I dont mind taking the time to collect the data because I can already see that at lvl30 its still a problem. I have lvl30ish toons on TGC and I can see that its a problem. Also I would like to note that my C++ experience is limited and I am still learning. I dont know all of the stuff I would need to make it work correctly but I know that there are people here that do. Also the fact that this isnt a problem on project1999 means that its already been solved and I am not sure that they took the time to get 10,000 lines of data per level to solve it. I dont mind power leveling toons to certain levels and getting them the same gear to do the testing but I want to know that if I do it at lvl60 will that be sufficient or are you wanting something higher than that to do anything? Which is why I asked Cavedude what he meant by higher levels.

65 Raids parse very similar to Live on TGC. I agree the farther you go down the worse we scale but even at the lower levels it's honestly not that bad. It also needs to be stated that the era we are currently on cannot be parsed because it no longer exists on Live. So, we approximate. As it's been said above, you need hundreds of combat parses to form a true data set. A handful of battles will not do, because that same battle may play out much differently each time you try it. It's really a big picture type situation.

Combat sadly simply cannot just be fixed. Sony overcomplicated it, and so it's more a slow progression to dial in on the right formulas. KLS did a lot of that a while back, but really he got combat to a place that is really close, especially considering that we have probably a hundred or so rules that can be adjusted on the fly to tweak it further. I'm not saying there isn't room for improvement, there most certainly is. But, I wouldn't call combat broken, either.

You can't compare P1999 to EQEmu. They are emulating a much earlier time with a far less complicated combat system. Even if they provided their source/rules, I am sure much of it would be irrelevant to EQEmu.

But, as I mentioned above we do have tons of rules that govern a lot of areas in combat. If your C++ skills are limited you may be able to look into them to start tweaking. The rules can be changed and reloaded on the fly, without even restarting the server so it's pretty ideal. If you find values that you feel are closer, please share. Also, I am the first to admit NPC stats are off in the DB. We just don't have enough data to correct them all, since the important ones are hidden to the player.

How many fights would you need? I can get unlimited fight data using MQ2. I currently run my 90 sk mnk shm clr in brells rest without logging them out for weeks at a time. The longest I have left them in was three weeks without logging out. Every 4 or 5 days I run my shaman over and sell all the crappy and I usually get about 1mil plat every 5-7 days. My sk can solo pretty much anything so if you wanted me to I could deck out a tank in GoD gear and use my clr shm to keep it up indefinately. The only problem I see is that raid and group stuff in EQ has always been balanced differently. Remember the nerf to monks back in PoP era? The raid monks were effected least of all when they nerfed the mitigation and monks that were in group gear suffered greatly.

My goal is to get the parse charts to scale like live on my server and honestly I think a check like I talked about is needed. If it doesnt work then thats ok but if nobody tries it then we wont know. If I start writing this piece of code can you help me with the stuff I dont understand? Honestly most of my C++ experience is writing and modifying macros for mq2. I have started teaching myself C++ from my friends books he used it college but being that I am in college atm I dont get a lot of time to just sit down and learn it.

Edit: Just to note. I dont die ever. Ive even had people train my guys and they dont die =)

Edit2: Also it wouldnt be too difficult to recreate a similar era to what TGC is in. If you limit your gear and aas that were released during that era and stay in zones that were also during that era you will be really close. Another thought is that raid data will be harder to get since it has lockout timers and long respawn timers that you'll have to wait for on live.

You dont need to know any C++. Parse 3 different AC sets against a common mob 250 times every 5 levels(could probably do less at early levels). Then once you have all that data, create a mathematical formula that will produce those values given the inputs available. You will quickly find out that close is good enough.

I will get started then. Starting at lvl1 going to lvl65 every 5 levels 500 mobs. Because I wont be able to guarantee that they will always be even con.

Edit: I will be using a troll warrior to do this.

Do you want me to use any specific gear? or just the same gear for all levels?

or do you want like 100ac 200ac 300ac? Not character sheet but gear total ac.

Would have to be appropriate for the level(ie like 100,150,200 for level 10 ... 1000,1250, 1500 for level 50, etc).

Also if you are serious about doing it(good on you), then wait alittle bit and see what kind of consensus you get as to the best way to parse it. I just threw out that scenario because its going to yield more useful data than straight up parsing but I sure some of the devs who actually dealt with the combat code may have better ways.

sounds good. I'll start with levels 1 5 and 10 for now and do 100 200 300 ac. Checking the baz to see whats up and how much ill need to farm. shouldnt be too much.

I wouldn't mind putting some criteria together, but I need a few days to mull it over. Caryatis is spot on when he says get some consensus as to the best way to parse, consensus now would make accepting potential changes a lot easier. I wouldn't mind taking a look at live first to get some idea of baselines.

10,000 lines of combat sounds like a lot, but it is a reasonable figure to get a solid dataset. Getting 10,000 lines takes significantly less time than leveling up a char.

Any parses on live should be matched with identical parses on 0.8.0.

As far as what AC brackets to use, there should be a bell curve distribution of hits across 20 different damage values. As AC and PC level increases the bulk of hits will shift to the lower end of that 20, and as mob level increases the hits will shift to the upper end. To replicate the live formula's closely you need to identify the lower and upper end of that AC bracket at each level interval (5 level intervals seems reasonable), and parse 10-20 AC intervals (i.e 10-20 discrete AC values, not Starting AC + 20, Starting AC + 40 etc) in between. It's also necessary to understand the impact of PC level vs NPC level as a ratio rather than simply 5 levels +/-.

FWIW this is a pretty big task, once some criteria are established I imagine it would be more sensible to break the task up between more than one person if there are other volunteers ;-) Plus if people have chars at various levels on live it saves a lot of leveling up time.

I currently have 623 mobs parsed at lvl1 with 113ac. Im now doing 600 at lvl1 with 55ac. Then I was going to do 600 at lvl1 with no gear. Using the same wep the entire time.

Edit: I have my 90 cleric grouped with my war in the field of bone newbie area. I have my 90 sk killing trooper taer so he doesnt come down and kill my war =).

are all the mobs the same level? :P

No they are all even con and yellow. To get 600 even con mobs at lvl1 would take far too long. Thats why I did 600 so that it would even out some. Unless they have to be even con then Ill do it again. I dont see too much of a problem with level 1 data including yellow mobs.

So they're double and triple your level 1 char's level.. imagine how that scales at 30? :) Like I was saying don't rush out and parse.

They barely hit for more than even cons with 113 ac....

/GU Tanking summary for: Zial --- Total damage: 4268 --- Avg hit: 2 --- Swings: 4252 --- Defended: 0 (0%) --- Hit: 2063 (48.5%) --- Missed: 2189 (51.5%) --- Accuracy: 48.5% --- Dodged: 0 (0%) --- Parried: 0 (0%) --- Blocked: 0 (0%) --- Riposted: 0 (0%) --- Absorbed: 0 (0%)



Zial - vs - Combined: A scaled wolf pup
Attempts 4252 100%
Missed 2189 51.48%
Dodged 0 0%
Parried 0 0%
Blocked 0 0%
Riposted 0 0%
Absorbed 0 0%
1 1288 30.29%
2 265 6.23%
3 146 3.43%
4 149 3.5%
5 83 1.95%
6 44 1.03%
7 30 0.71%
8 27 0.63%
9 9 0.21%
10 15 0.35%
11 4 0.09%
12 3 0.07%

But the great thing about gamparse is i can choose which fights i want to examine and I know that all of the scaled wolf pups and decaying skeletons are even con so I hand pick all of those.

Here is all of the decaying skeletons and scaled wolf pups

/GU Tanking summary for: Zial --- Total damage: 1828 --- Avg hit: 1 --- Swings: 2684 --- Defended: 0 (0%) --- Hit: 1108 (41.3%) --- Missed: 1576 (58.7%) --- Accuracy: 41.3% --- Dodged: 0 (0%) --- Parried: 0 (0%) --- Blocked: 0 (0%) --- Riposted: 0 (0%) --- Absorbed: 0 (0%)


Zial - vs - Combined: A scaled wolf pup
Attempts 2684 100%
Missed 1576 58.72%
Dodged 0 0%
Parried 0 0%
Blocked 0 0%
Riposted 0 0%
Absorbed 0 0%
1 773 28.8%
2 143 5.33%
3 82 3.06%
4 64 2.38%
5 29 1.08%
6 4 0.15%
7 7 0.26%
8 5 0.19%
9 1 0.04%


The above had 5 large scorps and 1 militia skeleton in it which is why the hit was as high as 12.

Can only speak for myself but I'm not interested in data which isn't coming from tightly controlled testing. Going to pass (no pun intended) on this one for now.

Here is the same data that I first posted with the large scorps and the militia skel removed

/GU Tanking summary for: Zial --- Total damage: 3307 --- Avg hit: 1 --- Swings: 3970 --- Defended: 0 (0%) --- Hit: 1849 (46.6%) --- Missed: 2121 (53.4%) --- Accuracy: 46.6% --- Dodged: 0 (0%) --- Parried: 0 (0%) --- Blocked: 0 (0%) --- Riposted: 0 (0%) --- Absorbed: 0 (0%)


Zial - vs - Combined: A scaled wolf pup
Attempts 3970 100%
Missed 2121 53.43%
Dodged 0 0%
Parried 0 0%
Blocked 0 0%
Riposted 0 0%
Absorbed 0 0%
1 1223 30.81%
2 243 6.12%
3 134 3.38%
4 135 3.4%
5 67 1.69%
6 20 0.5%
7 16 0.4%
8 10 0.25%
9 1 0.03%

The fact is at level 1 the difference is negligible. Also this is with a solo war. The data is going to be skewed at higher levels unless you have a group because fights will take longer with a solo warrior than with say an sk pally mnk zerker rog. The warrior will take more damage because they cant dps to save their life.

You can see that with those 5 mobs removed the percentages are off by +- two. So if you actually knew how to do this mathmatically you would know how to spot outliers and remove them.

Any body who has ever played live can tell that levels 1-5 even db and yellow mobs are very similar but once you pass lvl10 they start to scale more. You are making the assumption that all of the data will be this way. When in fact at later levels the warrior wont be able to solo even con mobs unless heavily twinked in which case the whole problem to begin with is that the current system is based on being raid geared which messes up the everything that isnt a raid target and anyone who isnt in raid gear. You have to balance raid mobs seperately from the rest of the npcs because they are the minority and they have special combat flags that normal mobs do not get.

You balance group npcs with group gear. Then you balance raid npcs with raid gear. This is how sony had the game balanced until they started throwing out all of the all all op raid gear starting with velious. Then everyone who had raid gear from that era on could easily take down previous raid targets very few peoeple. It wasn't until they started make class specific and set gear that it came back into a better balance.

You don't really understand the point of this. The length of fight, etc doesn't matter. You are parsing the incoming damage, everything else is irrelevant.

Balancing does not enter the equation yet, you find the formulas and correct combat and then the server ops balance their content.

Really you want something like 3 or 4 hours(absolute min) at each data point(which may be around 250 mobs, maybe less maybe more). But the real key here is that the people in charge of committing your change, agree with your methodology. You can waste all that time if people think your data is useless so there is no point until a process is solidified.

Stuff like this isn't going to help you sell your idea to anyone.

Ok so the problem I have now is that not only are you not interested in following any sort of instruction as to how to source data, but now there is this spectre of scope creep looming. If you want to make changes to to the emu code I think you're going to have to do them yourself.

You're so lost it's insane. Just stop now, please.