EPTLotV Beta Balance Update - October 23 - Page 3
| Forum Index > Legacy of the Void |
|
FireCake
151 Posts
| ||
|
FireCake
151 Posts
| ||
|
Cyro
United Kingdom20309 Posts
It is random since the same inputs don't produce the same outputs. (or there is something hidden i am missing ?) I think the problem here is that the same inputs do produce the same outputs (the engine is fully deterministic, that's the basis for the whole multiplayer and replay system) - but it's difficult for the player to consistently give a precice input because of timing and highly inconsistent input lag of the engine. Also sometimes, small variations can have chaotic results. To add to that, there are also effects in the engine (like the scan range problem demonstrated here - + Show Spoiler + | ||
|
FireCake
151 Posts
On October 24 2015 17:25 Cyro wrote: I think the problem here is that the same inputs do produce the same outputs (the engine is fully deterministic, that's the basis for the whole multiplayer and replay system) - but it's difficult for the player to consistently give a precice input because of timing and highly inconsistent input lag of the engine. Also sometimes, small variations can have chaotic results. To add to that, there are also effects in the engine (like the scan range problem demonstrated here - + Show Spoiler + https://www.youtube.com/watch?v=oxWYxysVBvI If the engine is fully deterministic then explains me these things : -how do you predict creep spread ? -if 2 melee units attack each other, let's say 2 zerglings, which one win ? -when an scv build a building how do you predict on which side of the building the unit goes ? -About the pathfinding, how do you know which path the units will take ? This thing has changed so many times, and still change, DK confirmed me they were going to change it again for LoTv (and obviously this kind of thing is never mentioned in patch notes...) -The burrow/unburrow time of almost all (or all ?) units vary, how do you determine how many time a lurker (for example) will need to burrow ? ... | ||
|
RoomOfMush
1296 Posts
On October 24 2015 17:38 FireCake wrote: If the engine is fully deterministic then explains me these things : -how do you predict creep spread ? -if 2 melee units attack each other, let's say 2 zerglings, which one win ? -when an scv build a building how do you predict on which side of the building the unit goes ? -About the pathfinding, how do you know which path the units will take ? This thing has changed so many times, and still change, DK confirmed me they were going to change it again for LoTv (and obviously this kind of thing is never mentioned in patch notes...) -The burrow/unburrow time of almost all (or all ?) units vary, how do you determine how many time a lurker (for example) will need to burrow ? ... Just because a human cant find these things out doesnt mean the engine is not deterministic. If it wasnt then the outcome of a 1 on 1 battle between 2 zerglings would change if you load the game or watch the replay. Actually, pretty much EVERYTHING a computer does is deterministic. The only things that can screw with this is bad programming of multi threaded applications. | ||
|
Big J
Austria16289 Posts
On October 24 2015 17:38 FireCake wrote: If the engine is fully deterministic then explains me these things : -how do you predict creep spread ? -if 2 melee units attack each other, let's say 2 zerglings, which one win ? -when an scv build a building how do you predict on which side of the building the unit goes ? -About the pathfinding, how do you know which path the units will take ? This thing has changed so many times, and still change, DK confirmed me they were going to change it again for LoTv (and obviously this kind of thing is never mentioned in patch notes...) -The burrow/unburrow time of almost all (or all ?) units vary, how do you determine how many time a lurker (for example) will need to burrow ? ... You are right in general that there are many small randomnesses + Show Spoiler + (is that how you say it? :D) Not sure how creepspread works, but it's probably random too, true. Same probably for the SCV. Both of those things could actually follow a pattern, or a pattern of multiple possible ones (chosen randomly?), but at least they seem random. About other things: If two units attack each other, the "older" unit hits first, as far as I know. So basically, if you take one of your initial drones from the mineral line, it should beat a new built one due to that. This should also be true for abilities, e.g. two high templar feedbacking each other. Pathfinding isn't random. But the exact engine blizzard uses is unknown to us (I think). If it works properly, the units should just take the shortest path, but I think there are some bugs with cliffwalkers like colossus and reaper that make them take the longer paths sometimes. | ||
|
FireCake
151 Posts
On October 24 2015 17:54 RoomOfMush wrote: Just because a human cant find these things out doesnt mean the engine is not deterministic. If it wasnt then the outcome of a 1 on 1 battle between 2 zerglings would change if you load the game or watch the replay. Actually, pretty much EVERYTHING a computer does is deterministic. The only things that can screw with this is bad programming of multi threaded applications. This is wrong. Its very easy to add some randomness in a game but still have the same actions everytime you load the game or watch the replay. How ? Simply use a seed, for example a number like the date at which the game has been played and use this number to "solve" every random things of the game. This is used in many video games for decades. Strictly speaking computers are not random (even multi threaded applications), but what i care, as a player, is to understand the game and to be able to predict what is going to happen. After years of play i still have no clue how to predict all the things i mentioned earlier. On October 24 2015 18:12 Big J wrote: You are right in general that there are many small randomnesses + Show Spoiler + (is that how you say it? :D) Not sure how creepspread works, but it's probably random too, true. Same probably for the SCV. Both of those things could actually follow a pattern, or a pattern of multiple possible ones (chosen randomly?), but at least they seem random. About other things: If two units attack each other, the "older" unit hits first, as far as I know. So basically, if you take one of your initial drones from the mineral line, it should beat a new built one due to that. This should also be true for abilities, e.g. two high templar feedbacking each other. Pathfinding isn't random. But the exact engine blizzard uses is unknown to us (I think). If it works properly, the units should just take the shortest path, but I think there are some bugs with cliffwalkers like colossus and reaper that make them take the longer paths sometimes. This is not "small randomness", it's huge, most of the lucky window mine happened because of it. The older unit used to win against a new one, but this has changed and it's no longer the case, at least in HoTs. Same for mana units. I can't prove that the pathfinding is random (although I highly doubt they don't use an heuristic algorithm otherwise it would be very costly ressources wise and the shitty servers would be already dead...), however, units don't take the shortest path for sure, i have many replays where the units take a stupid path that is way longer than the obvious correct path. | ||
|
Cyro
United Kingdom20309 Posts
On October 24 2015 17:38 FireCake wrote: If the engine is fully deterministic then explains me these things : -how do you predict creep spread ? -if 2 melee units attack each other, let's say 2 zerglings, which one win ? -when an scv build a building how do you predict on which side of the building the unit goes ? -About the pathfinding, how do you know which path the units will take ? This thing has changed so many times, and still change, DK confirmed me they were going to change it again for LoTv (and obviously this kind of thing is never mentioned in patch notes...) -The burrow/unburrow time of almost all (or all ?) units vary, how do you determine how many time a lurker (for example) will need to burrow ? ... -how do you predict creep spread ? I'm not entirely sure but it's done the same every single simulation (for every player in the game of every replay viewed) using the same initial data - which is likely just the initial position of the creep tumor. They don't send a coordinate for every square of the map that gets new creep spread to it, AFAIK. -if 2 melee units attack each other, let's say 2 zerglings, which one win ? I've seen data on this. I think there's sometimes a delay - if you have a unit like a marine then there are times when the unit that was created first was guaranteed to shoot first. -About the pathfinding, how do you know which path the units will take ? I do not know, but the engine recieves a move command, knows the position of the unit when that command was registered and the path taken is the same. Every time. I can't prove that the pathfinding is random (although I highly doubt they don't use an heuristic algorithm otherwise it would be very costly ressources wise and the shitty servers would be already dead...), however, units don't take the shortest path for sure, i have many replays where the units take a stupid path that is way longer than the obvious correct path. It's not the shortest path but it's the same path, given identical commands and initial conditions. That's easily demonstrated by looking at either a multiplayer game - your ally doesn't recieve data for the path your unit walks on, his client just knows where the unit is and recieves the coordinate of a move command. Watching a replay doesn't contain that data either - both in MP and watching a replay, it's a list of commands that are simulated with identical initial conditions and deterministic simulation so everything happens in the same way -The burrow/unburrow time of almost all (or all ?) units vary, how do you determine how many time a lurker (for example) will need to burrow ? As commented above, there seems to be a delay associated with this that could be random to the player at the time. There might actually be a few of them in the game, but many things that people assume to be random are actually not random but due to things like the scan range effect (video above), highly variable input lag (think something like 50ms +-25ms while some games manage 20ms +-5ms or less) and other effects. I'm not really sure of the extent of true randomness in the game. The one that always bothered me the most was stalker vs marine. Sometimes you could kite way better than other times, sometimes you'd take a hit, sometimes like 5 marines would hit your stalker. That was due to a combination of the highly variable (uncontrollable) input lag and the scan range effect, two things that were fully deterministic and abusable by a machine but not by a human. They didn't need a seed or anything like that, it was two very understandable effects that happened to be borderline impossible for a human to track effectively. The second part of that is fixed now but not the first part - and you wouldn't understand either of them without looking deeper into the engine. --------------------------------------------------- Strictly speaking computers are not random (even multi threaded applications), but what i care, as a player, is to understand the game and to be able to predict what is going to happen. After years of play i still have no clue how to predict all the things i mentioned earlier. Yeah i agree, sc2 is lacking in some ways there. The input lag is really the worst one for me. Sc2 only processes input once every ~50 milliseconds so if you give an input you might have to wait 45 milliseconds for it to be registered or you might have to wait 5 milliseconds. On top of that, there was the added delay of your ping plus a buffer (another 50 milliseconds)? so the lag is both extremely high for a twitchy RTS and also extremely variable. You can compensate very well for high latency but you can't compensate for variability unless you're a machine. It's predictable variability, but not by a human | ||
|
drifter420
10 Posts
On October 24 2015 13:23 Riner1212 wrote: yea im not buying this game. me neither. i am very disapointed about the game | ||
|
RoomOfMush
1296 Posts
On October 24 2015 18:35 FireCake wrote: This is wrong. Its very easy to add some randomness in a game but still have the same actions everytime you load the game or watch the replay. How ? Simply use a seed, for example a number like the date at which the game has been played and use this number to "solve" every random things of the game. This is used in many video games for decades. I dont think you know what the word "deterministic" means. What you describe is still completely deterministic. Randomness like the computer produces is not random at all. Its absolutely deterministic. It just feels random to humans because they dont know where the numbers come from. On October 24 2015 18:35 FireCake wrote:Strictly speaking computers are not random (even multi threaded applications), but what i care, as a player, is to understand the game and to be able to predict what is going to happen. Yes they are. Because once you have multithreading with several CPU's things become random if you dont program correctly. CPU's do not always run the same code with the exact same speed. Sometimes a calculation can take one or two nanoseconds more. On average its always the same, but individually its highly erratic. This comes as close to pure randomness as you can get since it is real physical randomness produced by collisions between electrons and atoms within the computer circuits. You just use the wrong words. The game is completely deterministic and not random at all. But some actions are very complex and rely on a multitude of variables that are not visible to the player which makes them seem random. Its unpredictable. But not random. | ||
|
Cyro
United Kingdom20309 Posts
| ||
|
StarscreamG1
Portugal1653 Posts
| ||
|
Gradient
Sweden15 Posts
The delay before a Siege Tank in Siege Mode firing when being dropped by a Medivac has been increased from 0.53 to 1.43 seconds. I know they don't want to hear this, but please can't they just not remove the drop thing and make the tank good again? Playing bio/tank is so fun in TvZ and right now it feels like we can't buff the tank because of drops but at the same time they keep on nerfing drops so they are useless anyway. | ||
|
TimeSpiral
United States1010 Posts
On October 24 2015 20:00 StarscreamG1 wrote: The liberator change was needed because of the maps. But can someone explain the lurker den time nerf? It already takes a lot of time. I get that there were one or two mineral patches, in the nat and some third bases, on some maps, that were one or two squares out of range ... so if the Terran tech-rushed and attacked a greedy Zerg, some damage could be done. So, nerf the casting range by a little bit ... But a 61% reduction initially? Then an upgrade in the fusion core to return it to a 30% nerf of the original version? And the tanks firing cool down after drop increased by 275%? I mean ... this seems extreme. | ||
|
[PkF] Wire
France24211 Posts
| ||
|
Big J
Austria16289 Posts
On October 24 2015 22:30 TimeSpiral wrote: These are humongous nerfs to Terran, lol. My god. I get that there were one or two mineral patches, in the nat and some third bases, on some maps, that were one or two squares out of range ... so if the Terran tech-rushed and attacked a greedy Zerg, some damage could be done. So, nerf the casting range by a little bit ... But a 61% reduction initially? Then an upgrade in the fusion core to return it to a 30% nerf of the original version? And the tanks firing cool down after drop increased by 275%? I mean ... this seems extreme. I thought the cast range was 9 before, can anyone confirm what number it was before the patch? And the liberator thing is also quite a buff in certain situations, especially defensively early on when you suddenly have a quite some extra resources from not researching the upgrade and still can use the defender mode. | ||
|
haiyeah
70 Posts
This is a buff to the unit if you ask me. The units that deal with it still deal with it (vikings have no real problems stopping liberator harass). Zerg can rely more so on Queens and spores with the reduced range, but have to keep on their toes. In most situations the liberator could do damage with the max siege range, and I still think it will be able to get away a lot with nerfed range, especially to opponents who aren't prepared. Not to mention this opens up some good timings in which liberators can be used to zone and protect your army much earlier and at less cost overall. Opponents will have to factor this into their unit choices more than they do now and hopefully as some have said this deals with issues like Nydus play. It will be interesting to see how it plays out. | ||
|
ohmylanta1003
United States128 Posts
On October 24 2015 13:23 Riner1212 wrote: yea im not buying this game. Thanks for fucking telling us! What an insightful comment. | ||
|
WhaleOFaTALE1
47 Posts
| ||
|
FeyFey
Germany10114 Posts
On October 24 2015 20:06 Gradient wrote: I know they don't want to hear this, but please can't they just not remove the drop thing and make the tank good again? Playing bio/tank is so fun in TvZ and right now it feels like we can't buff the tank because of drops but at the same time they keep on nerfing drops so they are useless anyway. I think they are still afraid that improving siege tank damage again will make Bio unviable in TvT. The nerf is from that time though where everyone ran head on into a giant group of sieged tanks, so not sure it was ever needed. | ||
| ||
