User talk:Serge Yseron/Archive11

In doors we trust

They all died. Not us :)

Serge Yseron 00:07, 17 October 2011 (UTC)

Cliff hanger

After adding a parameter to provide context specific filters to my ray trace fonction something went straight to my mind: i am assuming that all entities have the same physical abilities. But what if a grandpa could run around to go where cliff hanger would have trouble himself getting there ? Nonsens. I will add a layer system that will avoid to have duplicate nodes into a separate web for entities that cant use some of the nodes.

Serge Yseron 01:02, 17 October 2011 (UTC)

I have finished the preliminary tasks that consist to separate the 2 methods ( grid / web points ) into 2 separate source files. The pf job now call the method that match the pathfinding type specified in the client request where grid and web infos share a common storage space under an union. I have also finished the bsp that allow a fast retrieval of the nearest node for a given pos with non uniformly distributed web nodes, althought this is really only required when spawning entities or when selecting the starting point for a path calculation. Althought it is possible to add a full transform matrix to webs in order to support path finding on moving chunks of grounds, i will keep the exemple simple. The 2 methods will be used simultaneously in the application exemple ( in fact it is already the case, but method 2, althought now having all the datas it need, still have an empty FindWebPath() as opposed to the already functionnal FindGridPath(). Yseron - 90.15.176.114 00:36, 19 October 2011 (UTC)

The 2 methods are now operationals and run simultaneously. I will finalize and track bugs for method 2 in the days to come. 128 units pathfind each other on the web based landscape while 128 units pathfind each other on the grid at 2700fps when nothing in frustrum, 600fps when the 4000+ web pts and its landscape + the right grid and all entities are drawn. Yseron - 90.15.185.172 23:57, 19 October 2011 (UTC)

Finished. Sources here.

Serge Yseron 13:18, 22 October 2011 (UTC)

Nodes that cant see each other because of terrain topology have been eliminated. Overhead have been reduced by providing a quadtree to the collisions with the mesh. We got 125ms for ~6000 ray casts on 6400 polys. The sorting of the most aligned nodes now look decent and compact without extra variables(even with its double indirection because we want to keep the exclusion of already visited nodes) and is correct. The reset all in the static bits field that was reseting only the first chunk have been fixed even if we werent using it ( only the one used in the constructor was used and this one was already ok ). Special note: it is possible to replace the normalise() in the pf web job with a projection of the branchs onto a local-x-axis built via a cross product of the up axis with the reach vector in order to get the most aligned node but you will have to remember that you are then assuming that nodes are uniformly distributed. Yseron - 90.14.224.177 14:47, 23 October 2011 (UTC)

The paths generated are now clean enough that a bot user could use them. An ultimate bug in the CTArray have been fixed, it was discovered while adding the paths cleaner that eliminate un-necessary branchs, wich result in paths being...well, very much shorter. 2048 entities does not seem to bother the web based method, but we are still far from the standards in the industry wiwth 10.000 entities and sliced paths calculations. Serge Yseron 22:11, 26 October 2011 (UTC)

Before uploading the sources last evening i did make a sad last minute change whithout testing it (no comment). Fixed, sources here. By setting the paths best score to the size of the latest registered path (wich in fact is the correct thing to do) you can still gain some fps. Yseron - 80.11.87.230 08:53, 27 October 2011 (UTC)

The path cleaning algorithm have been improved and have now the strict minimal number or recursions while always continuing from the last extraction point. The condition for the related SwitchToThread() have been fixed from ( iteration & 8 ) to ( ( iteration & 7 ) ==0 ) which smooth things when exceeding 4096 entities. I go 2048 entities pathing non stop above 140fps with scene rendrered, 405 fps with pure calculations and no scene rendered while pathing always the farthest node (it's the equivalent of all the mobs of 4 gw1 areas pathing the other side of the map non stop). Note for self: replace the scratchpad in the WC. Yseron - 90.15.51.109 06:05, 28 October 2011 (UTC)

I reduced the number of priorities to 32 and added an acceleration table. The sheduling of prioretised objects is now on nitro, wich comes quite handy when you got 2048 entities switching priorities on a regular basis. Yseron - 90.14.101.229 11:48, 29 October 2011 (UTC)

While preparing the addition of the 3rd method (naviguation meshes) i made a review of the scheduling and fixed a bug. The prioritization is now even more faster. Sources have been updated. Yseron - 90.9.254.144 21:54, 3 November 2011 (UTC)

3rd method should be finished in a week or so. I also made a review of the bsp tree used for the web based method which now does not create empty branchs any more allowing the strict minimum of recursions while parsing it, and greatly reduce the complexity of creation and the number of variables that were involved.

Yseron - Serge Yseron 15:02, 6 November 2011 (UTC)

Someone complained that the latest build did not work any more with 2048 entities. Fixed. Yseron - 80.11.87.230 10:04, 8 November 2011 (UTC)

Nearest-partition vs parsing-all-web-nodes when pos outside of world bb: saw it, updated. Yseron - 90.29.49.78 18:44, 8 November 2011 (UTC)

Because i dont use svn for small personal projects i lost some of the fixes that i had made at work (at lunch time). Everything merged back and running. Yseron - 80.11.87.230 09:25, 9 November 2011 (UTC)

The format used for the navigation mesh is the COLLADA 1.4.1 one. It is exported with the latest version of Maya (first things first: if no nav mesh, then no nav mesh based path finding). Yseron - 90.48.128.214 00:36, 10 November 2011 (UTC)

I have finished the serialization of the previous datas formats so that i dont have to generate them each time i launch the sample. There was so much dependencies that i almost decided to go with a binarisation system but the ptrs/indices conversions are quite straightforward with small datas like these. I also added the construction of collada meshs areas wich you can see on the picture above, as they are outlined. Creating areas as convex N-Gons reduce the number of areas we would have by using triangles directly, and eliminating redundant collinear shapes edges speed up detections, wich means that it is fast. 3rd method will be finished in less than a week since now come the easiest part. It will run simultaneously with the other methods as usual. Yseron - Serge Yseron 22:12, 12 November 2011 (UTC)

Convex areas generation have been put on nitro by testing only 3 corners involved in a shape expansion. Yseron - 80.11.87.230 15:18, 14 November 2011 (UTC)

---

Finished. Sources have been updated. I wil continue to work on it for a few days then it will be time to do something else unless one of our artists make me the castle and its nav mesh like he said he will.

Serge Yseron 23:00, 15 November 2011 (UTC)

3 major fixes + 1 optim (gained 60~70fps). Updated. Yseron - 90.15.178.246 02:32, 18 November 2011 (UTC)

Video here. Yseron - 90.15.50.196 04:30, 20 November 2011 (UTC)

They were already right

I took a look back at a site dedicated to designers that i first saw in 2007 (it's always interesting to look at things you dont know a dam thing about once in a while). I came across this site while seeking as much external analysis as i could on the alliance battles gameplay in gw. Now, looking at what gw2 designers anounced for world pvp, i realize that this guy, without even knowing gw, gave me the answer that is about to be implemented in gw2... 4 years later. Yseron - 90.14.98.143 21:58, 17 October 2011 (UTC)

X-men

I'm usually not fan of xmen movies because i dont have afinities with modern teen movies where tiny girls in leather suits fight hulks with cheap hard rock music. But this one movie: X-men origins was close enough to what i experienced when i did read my first marvel comic 27 years ago, the golden era when reading comic books was still considered abnormal by many parents. Back then beeing a "geek" because of these comics was not easy to live. Look where we are now... Go go daddy Stan lee ! Yseron - 90.14.98.133 20:15, 22 October 2011 (UTC)

PS: please bring back bad guys bad ass introductions ! Yseron - 90.14.98.133 20:25, 22 October 2011 (UTC)

GW2 and water

How to click a location into the void of the ocean to give the path finder a target position: you cant. Logic deduction > train your hands for long distance travels under the water if no clickable obstacle in sight, unless the auto run / auto swim button is available. Yseron - 90.14.224.177 14:56, 23 October 2011 (UTC)

I got it: if clicking instead of using the movements keys, the intersection of a ray starting from the player camera with the camera frsutrum far clip plane will give the direction and activate the auto swim. Simple. In fact you dont even have to calc an intersection, the proj matrix suffice to build a vector using the mouse pointer. Yseron - 90.14.224.177 15:11, 23 October 2011 (UTC)

They anounced that the "click to move" wont be present in gw2. So in the end, we will have to train our hands so that they dont hurt after some hours of play, despite what i said. Yseron - 109.213.29.70 23:25, 23 October 2011 (UTC)

The guildwars of our childhood

1 ) Necromancers here (who said dhuum ?)
2 ) This one have it all: an asura portal, Daniel Dociu and a bad ass boss with hundred blades.

But well, it wasnt really interactive...

90.29.179.34 19:49, 24 October 2011 (UTC)

Oh, the good ol' days. Koda 10:19, 26 October 2011 (UTC)

Startcraft 2

Flocking in action. Yseron - 90.14.101.229 12:12, 29 October 2011 (UTC)

Interesting lectures

- this one for those who will join the adventure one day.
- this one to see what people sometimes do with gw.
Yseron - 90.15.48.97 13:38, 31 October 2011 (UTC)

Could be a hoax

But if it's true, the most manly man of men of the anime history will be back with lots of james bond girls and his manly psycho gun(there is a slight recoil but you dont really aim), the bondage sisters and other super jazzy felicity von delight. Yseron - 90.9.127.2 13:18, 1 November 2011 (UTC)

AISeek and Kinapse

Over 9000 since some times now... Yseron - 90.9.254.144 22:51, 3 November 2011 (UTC)

This century will be multi-core or will not be. Yseron - 90.29.179.201 18:43, 5 November 2011 (UTC)

Note to new ogre users like myself: the performance monitor eat itself quite some fps so dont forget to minimize it while monitoring the fps otherwise you get things like 3200fps when you in fact have 4500fps without the ogre monitor (wich paradoxally is there to measure the performances of your application). The difference become less noticable as the fps gets lower wich might be the reason they didnt code it (they provide their sources) with the usual minimal-measurment-impact we are used to with commercial middlewares. Yseron - 90.29.179.201 19:26, 5 November 2011 (UTC)

Remy Grendizer

Save the earth !. Yseron - 80.11.87.230 15:41, 4 November 2011 (UTC)

It's official

When women look at the new superman, they look at the eyes. Yseron - 90.29.179.201 21:01, 5 November 2011 (UTC)

Still a lot of discussions around this

But whatever will come with the next century, our children might know an interesting era. Yseron - 90.15.178.3 20:52, 6 November 2011 (UTC)

Fast food

Somehow, if a simulation could run a crowd ordering random menus at a fastfood with various configurations for comparisons, we would know once and for all if the actual ratio (customers/cashiers+coockers) is justified. But if you do find that it is, please dont publish the result, because a fast food that is not fast sucks. Really. Yseron - 80.11.87.230 10:30, 8 November 2011 (UTC)

UDK

Toyed a bit with it. One word: *cant speak*. Yseron - 80.11.87.230 15:11, 8 November 2011 (UTC)

Oh Oh Oh

Hi ah hi ya oooohhh. An horror movie where Jason would be the good guy trying to survive in a town full of sickier things than himself would be ecstatic. Yseron - 90.15.48.179 19:31, 11 November 2011 (UTC)

Johny in the boot

Johny est dans le coffre. Yseron - 90.15.52.30 21:42, 13 November 2011 (UTC)

Why it's a good idea to encode your videos with H.264

Here. Yseron - 80.11.87.230 11:13, 14 November 2011 (UTC)

New Adobe update including support for H.264: thanks for the reactivity guys :). Yseron - 80.11.87.230 15:51, 14 November 2011 (UTC)

Considering H264 is not that new at all, 'reactivity' is a pretty broad term to use. Koda 15:58, 24 November 2011 (UTC)

plugin-free 3D wikis ?

why not ?. Yseron - 80.11.87.230 17:21, 16 November 2011 (UTC)

GW2 urbanism

Please, i want to look hot. Yseron - 90.14.97.200 14:52, 18 November 2011 (UTC)

Jeff Strain Undead Lab's pre production finished

Like zombies always say: "AaaaAAAAaaaAaaahhrrrr.....". More elaborate discussions here. Yseron - 90.15.176.152 22:39, 18 November 2011 (UTC)

Cool ingame footage with the lightning we like so much in horror movies. Yseron - 90.15.176.152 23:25, 18 November 2011 (UTC)

COLLADA

Made the collada importer a little more generic so i can load colladas containing multiple meshes continaing multiple sub meshes per material like this one. Infos on the geometry parts are gathered first, then verts/normals/tris buffers are allocated once. Less memory fragmentation and 26.000 polys files load instantly that way, but i still some work to do on it. It should be noted that xml formats were hugely not memory friendly with past portable consoles, but they all come now with a minimum of 16Mb. Typically, if you still had no choice but to use xml files, depending on the media these files were on, and depending on their size you would try 1) to right-allocate it, or 2) to load it in a reserved space for temp contents, or 3) (more rare) to not load it into ram at all, wich mean that you then had a considerable number of seeks, wich was a killer when CDs were still used, and could even lead to a TRC failure. But the cool thing with PCs is that you can code like a pig and still not get performances issues unless you greatly exagerate. That's why i did not come yet with a separate heap that dont need to be locked for each agent thread. Yseron - 90.15.49.244 15:53, 20 November 2011 (UTC)

I now comply with the essentials collada format specifications that i need, hence i do not rely any more on fields appearing in a specific order and with a specific number of them. I also fixed the shared normals and vertices along with the meshes references to other meshes sources (althought i wonder if collada references depth can exceed 1, i will have to check the specifications again), and succeeded to keep the loading and instanciation of a 26000 tris mesh under 2s. The potential latency due to case sensitive operations have been eliminated with this (you can simply mask the bit 6 if you only deal with the alphabet, you will then always work in lower case, but do not use a substraction because the char may already be in lower case). Precompiled headers have been activated because it begin to become somewhat necessary. Sources available here. Yseron - Serge Yseron 21:56, 22 November 2011 (UTC)

A bad variable rename had led to some portals not to be created any more. It have been fixed but no luck if you downloaded the previous version. Time to get some sleep so that at least i still see what i type when i will wake up. Yseron - 90.29.48.23 00:52, 23 November 2011 (UTC)

I did add adaptive semantics (it is derived from the one i had made for mathxpress) so that parsers benefit from a single bitwise & operator when checking redefinable semantics (so that you dont have to touch your parsers when a bracket or any oher semantic changes of meaning). I am in the process of replacing clusters of bloated characters tests with tiny single semantic checks but the chore mechanic is already donwloadable and included in the collada parser. Yseron - 109.213.153.211 16:52, 26 November 2011 (UTC)

Vij

A 2 meters diameter life saver, Poncho Villa's bollocks, and a book titled "exorcism for the noobs". This is aint gay exorcism. Yseron - 80.11.87.230 13:06, 23 November 2011 (UTC)

GW2 path engine

Look who is there :=) -> ArenaNet and NCSoft for another title than gw2. Yseron - 80.11.87.230 13:13, 25 November 2011 (UTC)

Curiously enough it isnt mentioned on the gw2 wiki page at the time i am typing this. They choose another one already ? Yseron - 80.11.87.230 17:06, 25 November 2011 (UTC)

Just above you got this one but its 3 time the price of the one choosen by arenanet. Here we retrieve the convex areas we are fond of in the screenshots. There are other softwares that will cost you more but on another hand you pay for a highly professional reactivity of the support (resolution of reported bugs within 24 hours). Yseron - 80.11.87.230 13:34, 25 November 2011 (UTC)

Catch'em now

It's raining full movies. Yseron - 109.213.31.248 03:54, 27 November 2011 (UTC)

Busted by scientific police, guy ! Sometimes i'm happy to not wait for marketing to market things for me, that way i can see movies a few others also see. Those who will see each clip will be self rewarded, those who will go directly to clip 3 will miss something. Yseron - 90.15.183.170 20:08, 29 November 2011 (UTC)

Z fitness

is for zombie fitness. I know... I didnt want to believe it neither. Yseron - 90.48.3.177 23:47, 30 November 2011 (UTC)

Spelling

Because. Yseron - 90.48.3.177 00:21, 1 December 2011 (UTC)

K, i get out. Yseron - 90.48.3.177 00:25, 1 December 2011 (UTC)

Disco Babe canceled

I know it's old news. Each time i remind it i think to myself "goodbye hot bondage". Yseron - 109.213.30.34 22:32, 3 December 2011 (UTC)

Buggy Woogy

Serge Yseron 16:50, 4 December 2011 (UTC)

Open Pinder

(...) Android does not support System V IPCs, i.e. the facilities provided by the following standard Posix headers:

<sys/sem.h> /* SysV semaphores */
<sys/shm.h> /* SysV shared memory segments */
<sys/msg.h> /* SysV message queues */
<sys/ipc.h> /* General IPC definitions */

The reason for this is due to the fact that, by design, they lead to global kernel resource leakage.
For example, there is no way to automatically release a SysV semaphore allocated in the kernel when:
- a buggy or malicious process exits
- a non-buggy and non-malicious process crashes or is explicitly killed.

Killing processes automatically to make room for new ones is an important part of Android's application lifecycle implementation.
This means that, even assuming only non-buggy and non-malicious code, it is very likely that over time, the kernel global tables
used to implement SysV IPCs will fill up.

At that point, strange failures are likely to occur and prevent programs that use them to run properly until the next reboot of the system.

And we can't ignore potential malicious applications. As a proof of concept here is a simple exploit that you can run on a standard Linux box today:

(...)
removed given example code for obvious reasons
(...)

If you run it on a typical Linux distribution today, you'll discover that it will quickly fill up the kernel's table of unique key_t values, and that
strange things will happen in some parts of the system, but not all. (You can use the "ipcs -u" command to get a summary describing the kernel tables
and their allocations)

For example, in our experience, anything program launched after that that calls strerror() will simply crash. The USB sub-system starts spoutting weird
errors to the system console, etc...
(...)
Yseron 80.11.87.230 10:23, 6 December 2011 (UTC)

Some time you type NCSoft and Collada in google

And here is what you get. Yseron - 90.15.184.24 22:25, 8 December 2011 (UTC)

Energy system tossed and revisited due to complains

at the end in this video. Yseron - 90.15.184.24 22:52, 8 December 2011 (UTC)

The laughable side of modern marketing

Things we all find to be obvious except power-traders. Yseron - 80.11.87.230 13:53, 9 December 2011 (UTC)

Dancing was Matt

New travels are in the making.

While holding the ashes of Matt, you travel 33% faster, are immune to cripple, and are granted the grand master cartographer title and all allies whithin dance-shot gain +10 in broadway-booty-shaking. Yseron - 90.15.61.100 19:57, 10 December 2011 (UTC)

One last thing: the what the hell skill description. GO GO GO GO GO !. Yseron - 90.15.61.100 20:01, 10 December 2011 (UTC)

Peaks

The gw experience reaches its peak when you buy a scrol for Urgoz or the Deep only to find out that like the vast majority of other outposts, they are empty. Massive amount of casual Players tried to get involved. Oh boy, they tried. The worst thing anet could have done is to discourage them so that they stop trying. A single article who give hints about how to survive the first room would have sufficed to keep them wanting to try and play these areas they will never bother to enter again. After all, the more people play your game the more you should be happy, right ? Whatever they decide to do with gw1 in the futur, it wont have a long lasting effect, wich would be useless now that gw2 is comming anyway. Yseron - 90.15.59.52 22:41, 10 December 2011 (UTC)

Briefing

Eva: when i say it, we start running
Team: hooray !
Eva: but in the direction of the green thing
Team: ...

Serge Yseron 03:05, 17 December 2011 (UTC)

Why is it always Zhed who gets the worst beating? You would begin to think enemies are programmed to go after the most implausible creature first...Koda 22:01, 17 December 2011 (UTC)

On this particular trip i was a little too light hearted and made him target clumps of foes in unsafe locations that's why he ended beaten like a dead horse. With the subsequent tweaks he finaly prevailed over his mates by scoring a zero death. On another hand i witnessed Norgu trying to get in touch range in order to cast a spell. Dont ask... Yseron - 90.15.63.76 22:55, 17 December 2011 (UTC)

Testing things alone is a bad idea. It reveal things the average pug dont want you to know.

Serge Yseron 02:32, 18 December 2011 (UTC)

F...ing portal

First and last attempt because of a f...ing portal.

Serge Yseron 22:24, 18 December 2011 (UTC)

Spirits cannot hold down pressure plates, no matter how angry they are. Next time I will bring a 'volunteer' because it went surprisingly well so far. I am wondering why all the Deep groups went so bad in those old days. Koda 11:14, 19 December 2011 (UTC)

Dear Santa Claus

This year i have been a good girl so if it's possible, i would like this book.

Eva 90.15.185.181 02:57, 19 December 2011 (UTC)

The fun side of beeing a pony

Serge Yseron 21:48, 19 December 2011 (UTC)

Elvis vs vampires

Elvis will be back in Bubba Nosferatu: Curse of the She-Vampires. For those who dont know the prequel Elvis recruit an imposter before retiring and then kick some mummy's ass 20 years before Ash becomes the guy we know.

Sing for the King, baby. Blackout Joe 90.48.1.95 23:41, 20 December 2011 (UTC)

Goodbye AB

So long... Yseron - 80.11.87.230 12:52, 21 December 2011 (UTC)

Experience

Definition. Serge Yseron 20:41, 22 December 2011 (UTC)

WoC sho estate: fingers in the nose

Hint:

Scavenger / famine... got it. (personally i prefer to cast when they can only wand wich occurs right after 8 secs)
You need only at leat the first spirit.

Serge Yseron 17:47, 2 January 2012 (UTC)

This one is more direct but it's not really a build, because, well, ursan...


Serge Yseron 01:51, 3 January 2012 (UTC)

Van Ki Shot

For this 4rth day i am about to finish vanquishing the third and final continent. I have aqcuired the certitude that infuriating heat is an invivation for warriors to kd kournan teams that got 2 healers. Great dwarf weapon greatly speed up things in the process (or weapon of agression), and when thinking of it, i should have used that when doing woc: minister sho estate (opening: warrior head to critical targets that must be taken down quickly while i make spirit last the longer possible, if i am mesmer secondary i rupt showers or revives with "complication" preferably, we ignore melee until step 3, only keeping them empeded. step2: if there is a second relevant target, warrior chain to this one. step3: we quitely take down the remaning melees or rangers). Yseron - 90.48.129.33 20:05, 4 January 2012 (UTC)

3/3, Finished. Yseron - 90.15.61.44 09:17, 6 January 2012 (UTC)

When looking back at the thousand screenshots i took during this, i can see i have been forced to bring back some concepts to life:

a) 2 monks that got a minimum of 80 armor (95 when suffering from conditions) are not interesting for HM foes who ignore them. Suddenly, there are two party members that are not to be healed, which immediately reduce the energy they spend to keep the team alive by 33% (because a 60AR take double effort to keep alive, and while beeing kept alive, party members with higher AR are not healed, wich will demand even more energy later). Add to this that now Talkora is paragon secondary with 24 armor buff and party wide regen, that dunk cant be knocked down and disable attacker skills through shield bash: we are now back to the situation that allowed me to do sorrow furnace with a 3 men team when no pve skill existed. So to resume, they have less energy because they got no staff, but since they need less energy overall, they still got 33% more energy than what they now need: it improved. The only mistake Anet perhaps made here is to assume that all PVE monk have a minimum of PVP experience.

b) A more active play can significantly ease situations where 2HM healers would complicate things. By making warriors or assassins tank with no offensive skills, the players that regularly complain about HM are in fact preventing themselves from being effective. What they call an overpowered healer is in fact a healer that is left alone so he can quitely do his job. Now observe this: in HM the AI kit melees wich means he dont heal his mates while kiting. If he does not kitt, the warrior apply enough pressure that the healer is forced to heal himself, leaving his mates at your mercy. Anet have been extremely cautious with the content of HM by ensuring that a HM healer that is hard to take down mean his mates can be taking down regardless, and that monks that are easy to be killed mean that you will have trouble killing his mates without extracting the monk first.

c)What disappoint one-build-for-life players is the very concept of a build: you build as you see fit, and with sugar on top, you can manage a library by saving them. So whenever Anet will try to pull players away from the old holy-trinity mentality, these same players will fail when in fact they got tons of skills that can handle these situations efficiently. It happens in every content made recently (one year old or so). Yseron - 90.14.101.182 19:44, 6 January 2012 (UTC)

Good monks are very rare. I am certainly not one, but what I do know is that PS easily replaces a Healer's Boon monk on its own on hard mode, and most people do not realize this. They rather have their monks empty their energy pools spamming heals than actually reducing damage to about 20%. Koda 01:29, 7 January 2012 (UTC)

I think they are confonted to situations that occur so fast that proting a single target in a time laps of a few secs would result in that unique target being alive after those 5 secs. In other word, i think they random-pug often. However, after the whole team survived 4 successive searing flammes in HM (bad aggro due to an error from me), i am tempted to think that even with prot spirit, paragon secondary with party wide buffs is not a bad idea. There were only 3 heros with less than 110 armor when it happened thought (me and the 2 monks had 114, the warrior was 144, the paragon was around 134), and they had 84 armor. The team health did not drop much below 75%. Talkora did land "stand your ground" and "never surrender" immediately. After this experience, i realised we could rush the folowing groups that had a Rubis Jin, wich accelerated the vanquishing, and made me realise that Anet really had made the paragon well suited for Nightfalls. Diversion seem to be the best response on rubis Jins thought (or signet of humility). Yseron - 90.14.101.182 03:21, 7 January 2012 (UTC)

That is, of course, because you are good at this game. Most pugs are not. Yes, I apparently like to torture myself by joining random groups for z-quests and getting frustrated at the bazillion eles who think that because Shockwave got buffed, they should shadowstep to the nearest mob they see and empty their energy pool. Koda 17:04, 8 January 2012 (UTC)