|
Russian Federation4235 Posts
Now, one might wonder why I post this on the strategy forums, but in fact this could affect one's strategy and generic useful facts usually go in this subforum.
I. What is latency?
Latency is a wait command issued between "User Interface receives command" and "command is executed in-game". With high latency it's hard to micro because you have to micro in advance by like 0,5 seconds. However, there are actually two sources of latency: actual latency between peers (also called ping) and in-game latency setting.
II. Ping.
Ping is measured by sending a packet of information to the peer and measuring the time taken to receive response. It's being measured in milliseconds, so a ping of 100 means it takes 0,1 seconds to exchange information between peers. What effect does it have on the gameplay is not an easy question to answer. That's because the built-in latency system in StarCraft usually tends to compensate by creating artificial delay between command execution so that you never feel the ping. However, if ping exceeds the latency delay, the game becomes unstable - framerate drops and it becomes slower, but (and this is important) it might or it might not increase the command delay. So, in most cases, for high pings you'd want to increase the latency. Setting it to high is usually enough, in some cases, you need to set it to extra high, but that means you have a really bad connection to the peer and would want to find another playing partner.
III. Latency.
Now, to the most important part - there are actually four, not three latency settings:
- Extra High Latency
- High Latency
- Low Latency
- Single-player Latency
EDIT: It has been proven that there are more settings and Low Latency over B.net is different from low latency on LAN. The point is about LAN, though, so B.net will be left out of the boat at this point.
Now, the command delay for the last setting is zero. Yet, this is the only setting that has zero delay. What is important is that command delay for LAN latency is NOT zero and (this is purely observational, but should be within limit of error) has an equivalent ping of about 200. Yes, 200 milliseconds between issuing an order and it's execution. Is this large? I can say that it is. If you want to test the impact, create a UMS (it will not take longer than 10 minutes) that focuses on doing some precision micro, for example, dragoon mine defuse micro (Free - style, like in PP06) or zealot dance micro (move - hit - move). Run that map in single-player, practice for some time to succeed to some extent. Then, run the same map on UDP LAN. Feel the difference. It's there and it's significant.
IV. Conclusions.
1) Whenever you make or download a micro training UMS, don't play in sigle-player mode! The techniques you might take a hold on will not work online.
2) This game is a little harder than it seems and those micro feats done by the pros take abit more skill to do despite them "playing on LAN".
3) Average human reaction time to "slightly expected" event is about 300 milliseconds. Those 200 add up to a grand 500 which is quite a large time period for intense micro. So good micro is not only gained by fast hands, you need to expect stuff. The point of micro training is that you need to expect what will happen next, predict it, and react before it happened. If you make a reaver harass, for example, you want to pick up your reaver right after it fires. If you don't have the timings in your brain, it will take 300 ms to react to the reaver shot and 200 more for it to load into the shuttle. If you have the timings, you will order the pickup like 100 ms before the shot happens, saving you 400. Almost half a second. That tank might have fired in that time.
4) Maybe Blizzard should implement more LAN-friendly latency settings for StarCraft II, like a 50 ms latency which is like totally unnoticeable but covers the possible pings on LAN. WarCraft III's engine totally failed latency settings.
UPDATE:
A really nice experimental work by SonuvBob here:
On September 12 2007 05:48 SonuvBob wrote:
Just tried the FPVOD test on LAN w/low latency, in both single and multiplayer.
In MP (UDP) it's 200-250ms (13-14 frames at 60fps) between when the right-click circle first shows up and when the SCV starts to react.
On B.net it's 450-650ms (27-39 frames at 60fps).
In SP it's 80-100ms (5-6 frames at 60fps).
That was tested with a drone or SCV (time between first frame with the right-click circle and the first frame the drone reacts), as well as unit creation (time between mouse-up on unit icon and appearance of unit in queue).
|
Are you sure that low latency setting on Battle.net means the same thing as low latency using LAN? I've always had the impression that the two are different.
|
I don't think you are right when you assume there are only 4 latency settings. I just played a number of games on LAN high latency settings, and it had significantly less latency then I would have encountered on battle.net with low latency.
|
i like the look of your post, sadly the content is totally wrong :-/
have you ever played bw MP over LAN (or even hamachi should do)? you WILL notice the difference...
|
muta micro on bnet even with low latency is still really hard, so much less response...
|
Canada291 Posts
|
Hong Kong20321 Posts
On September 11 2007 21:15 IefNaij wrote:
You are wrong.
gg pwned? 
|
Low Latency is nowhere near LAN.
LAN is no latency. Go play a game on LAN vs a friend or something then go on bnet and try low latency. HUGE difference.
Nice idea, but you apparently don't know what you're talking about. Even hamachi has less delay than bnet, and that's because the lat settings for each are completely different.
|
hamachi is pretty close to lan irrc, bnet is miles away from that
yeah, you are wrong.
|
Battle.net was originally made for people with dial-up, expecting that people had at least a ping of 200. That's the main reason why the 'low latency' is actually pretty high on Battle.net.
And yeah, running it through UDP is different. I know, because I only have 10ms to Battle.net and the latency is always there no matter who I play with.
|
I have several IRL friends which I play a lot with. Sometimes we play over battle.net and sometimes over hamachi. I can assure you that there is a HUGE difference when playing over Hamachi, it's a different game. We have good connections and live pretty near each other geographically so you can easily compare our hamachi gaming to real LAN.
|
Russian Federation4235 Posts
Ok, I accept defeat since I've made a few tests with B.net right now. The low latency setting for B.net is indeed higher and you all can have my apologies for providing wrong info.
However, that is not really the point of the OP. The point is the difference between the single-player latency and low latency on LAN. It is there and that is something I'm willing to stand for. Editing the OP now, but the admins still might close the thread since it originally contained wrong info.
|
Maybe someone could find the (more or less) exact times by recording short FPVODs with low latency, one using UDP on LAN and one on B.net, and checking the time in between click and response.
|
Does anybody know how B.net works? It keeps track of the state (i.e. positions of units, HP, upgrades, etc.) of the game based on the commands you and your opponent send and reports that information back to you after a certain delay that acts as a buffer period (delay determined by your latency setting)? So when your local state is out of sync with the opponent, it forces the person ahead to wait in that 45 sec countdown screen. Greater latency means less getting out of sync but greater wait times before responses. Or do I have a part wrong? Is instead the "server" client that's starting the game tracking the game's state?
Ethernet is extremely reliable and fast (100 Mbps nominal generally for point-to-point connection, depending on what type Ethernet), yeah, so LAN latency should be pretty low.
For LAN, how is it implemented?
|
I belive bw is constantly sending small packages of information to b.net and the latency detriminates the size of the packages, and therefore also the frequency with which they are sent.
|
Im sorry to have to contradict you. Your lan setup might be crappy, defect or just plain clustered but, in no circumstances ever, should you get a 200ms delay on your Lan.
|
On September 12 2007 00:58 BluzMan wrote:
Ok, I accept defeat since I've made a few tests with B.net right now. The low latency setting for B.net is indeed higher and you all can have my apologies for providing wrong info.
However, that is not really the point of the OP. The point is the difference between the single-player latency and low latency on LAN. It is there and that is something I'm willing to stand for. Editing the OP now, but the admins still might close the thread since it originally contained wrong info.
still u r talking shit
low latency on bn is higher than extra high on lan
gg no re
|
On September 12 2007 04:37 DOgMeAt wrote:Show nested quote +On September 12 2007 00:58 BluzMan wrote:
Ok, I accept defeat since I've made a few tests with B.net right now. The low latency setting for B.net is indeed higher and you all can have my apologies for providing wrong info.
However, that is not really the point of the OP. The point is the difference between the single-player latency and low latency on LAN. It is there and that is something I'm willing to stand for. Editing the OP now, but the admins still might close the thread since it originally contained wrong info. still u r talking shit low latency on bn is higher than extra high on lan gg no re No need to be an ass about it. He just admitted he was wrong.
|
So what is better when playing over iccup? should one setup High or Low latency ??
I play in iccup and my ping is always 220. impossible to micro dragoons well....
|
Russian Federation4235 Posts
On September 12 2007 04:36 Famehunter wrote:
Im sorry to have to contradict you. Your lan setup might be crappy, defect or just plain clustered but, in no circumstances ever, should you get a 200ms delay on your Lan.
Did you miss it? I was talking about LAN without a peer at all. Running single-player missions. C'mon, just go and test it and if it shows nothing, I'm gonna eat my hat.
On September 12 2007 05:22 saitox wrote:
So what is better when playing over iccup? should one setup High or Low latency ??
I play in iccup and my ping is always 220. impossible to micro dragoons well....
You should be playing with the lowest latency setting possible. What is possible is determined by the game running smoothly.
|
|
|
|
|
|
EPT
