Jump to content
Search In
  • More options...
Find results that contain...
Find results in...


  • Content Count

  • Joined

  • Last visited

About gaunsaku

  • Rank

Profile Information

  • Gender
  1. Snappier animations are always welcome! I'm a little confused by this update and the powers prediction one concerning warmup animations, however. The powers prediction one implies the length of an animation for a skill will be dependent upon network latency due to the server authority model being implemented (is it in-game now for all powers or just a select few?). Animation changes like the confessor's implies results upon use, like we typically expect from a game. Has the prediction model changed at all or will the animations still varying in speed/length based on latency to ensure 100% server authority accuracy across all clients?
  2. Yes, that is a problem. Ani-cancel, but require the full animation before doing damage/effect calculations. IRL with swordsmanship, I can be nearly 75% through a cut and still change, but that change results in not finishing the cut in the first place. RL has ani-cancel, so I think games should too
  3. Thomas is right in saying "the whole point is to telegraph what you are intending to do to other players, so you can respond." That I get and that I understand. Knowing how combos chain and being able to react to them in important in skill-based combat. But what Tera and BnS do is implement animation cancelling, which makes it even more skillful than what CF has implemented thus far. Parties involved in combat can cancel what they're doing and move into a counter, dodge, or escape. The ability cooldown kicks in, but it allows players to disengage, reset, and prolong the battle a bit. Furthermore, in Tera, some abilities (especially slayer ones) activate faster in a chain compared to executing on their own. CF's animation locks don't allow for this. You can get stuck in longer animations and certain abilities are not accessible unless done through the chain, and only the chain. Depending on how this animation warmup implementation goes, you're looking at maybe 4-5s before an ability 3-chain-steps in even gets fired off. I already feel champion chains, for example, take too long to trigger. Methodical combos like in BnS are one thing (ie, Kung Fu Master's 3RF), but hitting a button, waiting, and hitting it again after a noticeable pause can be a bit too much. CF shouldn't be Super Smash Brothers, but it shouldn't be chess, either. Now I will say this. The entire CF model for combat and animation is probably a good model to have for friendly fire. The way the playtest went last night, if FF was turned on, every group I ran into would've killed themselves. In FF situations, having a little more time to survey the landscape may make combat more forgiving. Or animation lock may make melee classes a rarity in FF campaigns due to their inability to get out of the way. Putting 4-5 ranged archetypes on a target versus 1-2 melee would be much more effective, especially if good melee abilities require a lot of time to setup because of their combo depth. We'll see how it goes!
  4. Hey Thomas! It's entirely possible I don't understand either the problem or solution! So let me clarify here as my time is not constrained by work or being a ninja The problem, as I'm interpreting it, is concurrency. Rather, it's trusted computational concurrency. The server is the only trusted source by the client is the relayer of state. Distributing data across actors/nodes (players and non-player voxels) such that they all have accurate concurrent states via a trust source (the server). The rest of my response is based around that, so if this problem definition is wrong, please let me know! Looking at the charts from the news article, any kind of client-server-animation greater than 50ms outside of ping is already quite slow. Server processing of power/location/physics data should be max 20-30ms and ideally in the 10ms range (IMO, YMMV). When you get into the 250ms+ realm, net ping included, I feel like something is going wrong unless the servers are at the south pole (no overheating, at least). You all probably feel the same way. But to tackle that problem of smoothness, concurrency, and (server) trust across all actors/nodes, you need to have: Server architecture keeping stateful awareness of all actors/nodes across the entire CW A way of receiving requests (chat/power/potion use/voxel impact/etc) from the client (net ping dependent) A way of checking the actor/node state for current position relative to the other actor/node involved (<5ms, CPU dependent) Resolving the request (<5ms, CPU dependent) Passing the resolution to all actors/nodes necessary (<5ms, CPU dependent) Reporting the resolution back to the appropriate clients (net ping dependent) If your team is able to share the internal test timing of request/resolve/return calls (steps 3-5) or embellish on how you all are threading state resolution on the server and client (ie, possible multithreading overhead with context switching/memory/scheduling queues/etc), that might give me a better understanding, because the charts in the news article make it seems like powers will function at the same speed as a Slayer ability in Tera when you mistime a swing and ruin the flow (the sword is really, really slow). With Orleans, it's not something you necessarily need to use or consider, more of a suggestion on how to handle the distribution concurrency problem with incredibly low latency (meaning net ping is really the only problem). Constant heartbeats across actors/nodes on the server which then distribute that information back to the clients for processing. I also threw out the storing of past actor/node states on the client, so that client-server resolution can remain invisible to the user while maintaining the true authority of the server. This is how Unreal implements their network architecture as I understand it, through function replication and state buffer that the server corrects. For example, I hop on my knight and see Valor's Champion trying to build a siege engine. I run up and attack. My client should execute the necessary functions for my attack (hitbox checks, attack speed, distance to target, etc) as if the client was a trusted source. The same calls are executed on the server, too, which does its magic and sends the corrected state of things back to the client. Thus if the client has cheats setup, this "prediction" model based off buffered client states don't cause popping (like the news post said) but just invisibly resets the client to the "true" state and displays the effect of the "true" execution. The vast majority of players will experience this near simultaneous state, while cheaters will have server corrections, so Valor won't escape my wrath even though he uses the popular cheat CrowRise. Net ping is the only real hangup; if I hit my attack key and don't swing or don't get results for 400ms, that's lag and I should kick my router. The handshake/verification/response on the client/server should not be the hangup. The one engineer also pointed out the goal is to have the client animation hit/end at the same time across all clients, so adding the warmup/delay gives the code time to get all that data distributed. Again, I don't know your guys' architecture or service model being used for updating clients, but if the solution is to vary the actual animation speed so that everyone is on the same page at the same time, that doesn't seem like the best method for ensuring smoothness. Hence something like Orleans or another distributed Actor Model implementation. I also have a question regarding this variable timing and projectiles. If I hop on my Confessor and launch an attack at Valor, how will the client/server authority work that out? Will you calculate power timing based on target range, projectile speed, and math? How would you even know what my target is? I might have aimed at Valor, but he moved and Ren was standing behind him, so I'm ok hitting Ren. Or will the warmup/concurrent model you're looking to implement result in my spell being delayed until the server has finished telling all clients to process the spell 50ms from now? Sorry for the long post. This is a complicated topic and we're not privy to all the code nuances. I just feel that, ultimately, the charts show animation and combat being too slow and some is bottleneck appears to exist outside of net ping. We want combat to be great, hence my thoughts/feedback/stupidity!
  5. I know it's been a long time since I've posted (damn RL), but I have to weigh in here because this model does not seem like the best way to go. I'm really afraid this is going to make combat even worse, which is something none of us want to see. The ultimate problem, as you said, it latency between client and server coupled with the need to trust the server in every way. What you guys need is a combination of actor awareness/distribution coupled with a buffer of client states/insputs such for validity checking. Rather than mess with the animations, keeping a client buffer state and using that to manage where the actor was, what they were doing, etc at the time of the server action request, and when the response is received you discard states older than the corrected response from the server, thus applying the actions invisibly to the world based on those positions from the past. You don't want to stop the world to handle this client-server relationship, which it seems like the animation locks and "build up" changes you feel like you need to make are for. Don't rework animations, rework the client-server model. I would also highly recommend you look into Project Orleans from Microsoft research. It's an implementation/derivative of the Actor Model Theory for massively scaleable virtual environments and was used in Halo 4 and Halo 5 (amongst other projects). It's also open source. While it was intended for cloud computing (Azure and stuff), it can be worked for use in a traditional server environment. Some links of note- Project Orleans GitHub Page Project Orleans Overview and some Videos on Microsoft's site Great blog post giving an overview of Orleans, the agent model, goes over core concepts, covers the latency problem, and more Plenty of other pages/links exist Please take a look and consider those options before you go too far down the rabbit hole of forcing the client into awkward combat mode to ensure client-server accuracy!
  6. I heard a rumor that next week's show would be Valor and Blixtev doing an hour of "What If" questions. Can this be confirmed or denied? Also understand that Valor doesn't want nameplates because in that Pro-PvP mind of his rests a little scared kittie kat. He doesn't want to end up like Todd, dying first in every encounter (too soon?). With friendly fire turned on, I might just accidentally see his nameplate and kick the snot out of him whenever we engage enemies. I mean, he and my other LotD guildmates keep questioning my PvP talents; if I suck that much, can you really blame a guy for killing his teammate?
  7. But if we have that large a % of testers, then the chances of us playing together is better than straight up PUGing.... Here's hoping pre-alpha folks are good at PvP!
  8. Question One - if all the LotD backers are playing together and absolutely dominate pre-alpha (while providing great feedback none-the-less) and this causes other alpha players to quit because they can't handle the harsh PvP environment, will you exclude us from future alpha tests? Or if some other guild besides LotD did that. I'm not sure who that would be, hence using my guild Question Two - how extensively do you want people to try and break things? Do you want folks to get out their memory management tools or is it too early for that? Regardless, this is all very exciting! I look forward to breaking things in testing when I get the opportunity
  9. Nice! Animations are flowing smoothly. Question - what kind of distance(s) are those? Are those grid blocks reference for any kind of distance? In the last combat question video you all talked about distances and how far things are. Just trying to get a sense of what you consider "20 meters" or "30 meters" to be
  10. I both group and solo depending on the time of day and how annoyed my guildmates are with me. Having the flexibility to do both will be greatly appreciated. But, more so than just the matchmaking system, I'm guessing the group discussion focuses more on the campaign world win conditions? It's one of the big unknowns: how individuals "work" within a campaign world that's not faction based (or even faction based ones to a lesser extent). On a different note, I love the idea of knocking projectiles out of the air. Any more insight on how that might work? I'm guessing spamming your skills and constantly attacking won't act as a "perma-shield" for projectiles in front of you. Also, something I just thought of with the whole projectile system and mass: have you run into any issues with projectile clusters having an negative effect on each other? If projectiles have physicality to them and you have 20 ranged people all aiming at the same centaur (for example), might their projectiles interfere or are you going with the model of "no two projectiles will never exist in the same space at the same time" ? Good stuff as always!
  11. Right, I'm not talking about that Discipline. I'm talking about disciplines like Werewolf and the like. I'm assuming that disciplines behave like in Shadowbane where they grant extra abilities and things, hence my question about resource mechanics and how that affects the disciplines
  12. I like the emergent gameplay ideas! Frontline "healer" is a cool concept, especially given mobility. Centaur size and their effect against FF and projectiles is also important to take into consideration from a tactical perspective. Good ideas. Questions! Does this mean disciplines will not be reliant on mana/rage/etc? Or will the discipline just make use of the mechanic of the archetype? Or will disciplines be limited to certain archetypes so-as to not mess with this resource mechanic? I'm just thinking that if there's, say, a Blademaster discipline and it gives special attacks that use mana, how would that work with an archetype that uses rage? How will the Legionnaire's Dodge ability affect other legionnaires? Maybe not at all given their equal size/mass? Along the lines of #2, any further developed thoughts on what physics "trump" what outside of iFrames? Would 5 legionnaires "dodging" into a line be a preferred meta? Why do you have to rag on Steve's? They get ragged on enough as it is Looking forward to future updates! Especially when the hitbox stuff is in place
  13. Yoh, I'm a bit confused by what you're saying. At first it seemed like you wanted there to be less variation because too much variation makes it difficult to "read" opponents, hence my reply. Then you said I'm misunderstanding and talked about complexities that are open for everyone to see but also how GW2 tried to do complex along with hidden and failed miserably. Bahamut then talks about LoL and also uses the dagger example of learning about your opponent, to which you counter with other esports that go with a much more simple approach. You also rail again on GW2's failed attempts at complexity. I'm with Bahamut on this, but for the sake of clarity, can you specify which of these position(s) you're taking? Or some other position? I included a quote to show why I say what I say. Hidden/Passive abilities make it difficult to form coherent strategy because of the myriad of possible combinations, so this is bad. The extent of "yomi" in a game should be fairly straight forward, based on limited combinations that people will know ahead of time Too many hiddens/passives/combinations-in-general leads to chaos and confusion, making a game bad. If you can't tell what an opponent will do based on the makeup of their avatar and the abilities they execute, that's bad. Balance between builds is important otherwise the game is bad. On the GW2 front, as part of a top tier guild who helped define and counter various metas in T1, I can tell you that strategies and tactics and organizing/managing small and large groups occurred regularly. We had some damn good commanders along with specs for everyone, checks, yelling at people non-compliant, and more. Organization and developing strategy in such a multi-faceted game can and has occurred. But you acknowledged that when you said "While perhaps at a high tip top level you could create order from chaos..." This is what we did because really good PvPers do that. It's how things have been since the beginning to varying degrees (based on how the game design allowed for it). It sounds like you want simple for the sake of easy discerning, limited options, and game balance. I can understand and appreciate why someone would want that. It does make things a lot easier from a PvP and game building standpoint. But that's not what people like myself want. If I wanted a simple setup where I know the basic combos and can turn off my awareness because I know that assassin is going to try and stab me Basic Instinct style, that's just like every other MMO out there pretty much. SB had variety, GW2 had variety, and both are up there from a PvP standpoint. Like Todd said, if you're playing a bad combination and you keep trying to make it work, that just means you're a bad player. I cannot be a successful computer engineer if I keep studying swahili. Life just doesn't work that way. Life gives us infinite flexibility and options while leaving how we implement up to us. Some are more successful than others. RNG is a factor IRL. Chaos is the norm. Yomi is what helps us navigate that chaos and the skills we nurture serve as a compass for doing so. If you nurture poor skills, you're going to have a rough time. Clausewitz's On War applies here as well in understanding chaos, differences in human beings, and formulating strategy/tactics out of that. I keep saying, this is simple, but not easy. You seem to want it to be simple and easy, which is where I must disgree with you
  14. I gotta disagree with this. Well, sorta disagree. You're thinking of yomi as actually being able to tell what's going on. That's a very basic level of yomi, something that most anyone can do just like most anyone can learn football formations and see what might be up. If you want that level of yomi, then by all means, remove passives and hidden mechanics and all that. Doing so will allow most players with a brain to see that the ranger with the bow will shoot arrows, the confessor will make me sinful, the assassin will backstab me, and the myrmidon will headbutt me to death. This is the typical class setup all MMO's have. From a PvP standpoint, that's basic and fairly dull. The next level of yomi is to "know" what those passives and hidden mechanics are through experience. This is more along the lines of what GW2 does. How well is up for debate, but it can't be argued that GW2 has a lot of options available. And you need to learn on the fly if they have that passive or that talent or that whatever which makes them different. You can't just assume they're specced D/D so they'll do XYZ. That's what yomi IRL is about, becoming aware of those "hidden" things, that kept in the shadow. This is another reason why Sun Tzu uses the ideograph 間 for spies: the sun is what illuminates and reveals that hidden in the shadow and it is the sun that can get between the cracks. Yomi is very deep, deep enough to be otherworldly as Todd's post discussed. That comes from knowing the unknown. That thrill of discovery is what makes PvP and dynamic content so much fun IMO. Then it's the bad PvPers who cannot change with that expanded awareness, just like it's the bad militaries that cannot change quickly with newfound intelligence. Yomi brings awareness. Awareness leads to strategy. Change is the tactic used for victory. Simple..... but not easy
  • Create New...