5 ms·

As context this was a big international fiasco. Nobody was acting maliciously but fans of RNG and other teams felt wronged. RNG didn't want to attend due to cov

by dematz 4y ago

As context this was a big international fiasco. Nobody was acting maliciously but fans of RNG and other teams felt wronged. RNG didn't want to attend due to covid restrictions on travel but Riot really wants Chinese viewers in international tournaments, hence all this online latency work. Pros on other teams were lied to by the false latency information which must be almost paranoia inducing to be told it's only 35ms even when it (correctly) feels higher. In the end RNG had to replay their games, which was not as bad as it could have been since they were vs much weaker teams, but still annoying. Anyways, by "scenarios where the actual ping was significantly lower than the target latency" do they mean like if the ping from Busan-server is 10 but they want it to feel like 35? I feel like that's the one scenario you'd care about, making the Busan players have 35 ping was the whole point. Edit - also the other scenario, where the actual ping is significantly higher than the target latency...well in that case your latency service would be some sort of magical time travel box! Adding latency is literally the only thing it can do right? It's not possible to lower it. Maybe the it depends what they mean by "significantly" but idk the target latency was 35ms so like at most they could be 35ms lower than that. I guess they didn't catch it because the bug was in the latency service itself which gave the wrong value to both the in game display and to the old network monitoring system. Idk in hindsight it's easy to say if you're going to introduce this new ping equalizing service it might break how you test ping especially because it hasn't been tested in soloq. But meh a lot easier to see what happened in hindsight after they write it all up.

vikingerik 4y ago
You can in fact have that magical time-travel approach to get apparent latency lower than the real number. Keep a history of the game state at every tick, so you can roll back to any previous instant. If real latency from one client is 50 ms, and you're targeting 35 ms, you roll back the game state by 15 ms each time input comes in from that client, and re-run the game logic from that point forwards and send the resulting state update to all clients. Those clients will see a bit of jumpiness in the game state, which isn't ideal, but may be the lesser evil compared to enduring more latency. (I have no idea what LoL in particular supports or implements, but that's the general idea.)
- meinersbur 4y ago
  That's just another kind of latency. Additional time has still passed from you (or the enemy player) doing an action, and you seeing the results of that action on your screen. By contrast, the server rolling back the game state is indistinguishable from the server always being 15ms behind and the clients running the simulation ahead of time (which they do anyway for fluency [1]). The "jumpiness" is called rubberbanding [2]. [1] https://www.gabrielgambetta.com/client-side-prediction-server-reconciliation.html https://www.gabrielgambetta.com/client-side-prediction-serve... [2] https://www.urbandictionary.com/define.php?term=rubberbanding https://www.urbandictionary.com/define.php?term=rubberbandin...
- jchw 4y ago
  This strategy is commonly referred to as being in the class of “latency hiding” as it doesn’t actually improve latency, it just allows you to reduce apparent latency. In practice it’s still useful but I think it’s mainly useful because it allows for local input to be buffered significantly less than if you were to compensate for lag by adding buffering only. However, that is not always useful. It is mainly useful in games where netplay is handled in a lock-step fashion, where each peer needs to keep every action coherent frame-wise. That makes sense for things like fighting games, but it doesn’t make all that much for say, puzzle games, even though those can be quite latency sensitive too. Either way, there is nothing magic. At the end of the day, the soonest you can see an action is still physically unchanged. Especially since the prediction model that makes the most sense is just assuming no action. That said, I dunno enough about MOBAs to try and postulate on how much it matters, and I have no idea if LoL is using lockstep or some other approach.
  - skocznymroczny 4y ago
    LoL is not using lockstep. Lockstep has two big disadvantages for a competitive game. Tt requires both client to have the full game state to run the simulation, which makes creation of maphacks trivial. The second issue with lockstep is that if one player is lagging, all players are lagging, which is also not something you'd like to have in a fast paced competitive game.
ear7h 4y ago
> I feel like that's the one scenario you'd care about, making the Busan players have 35 ping was the whole point I think the key word is "significantly". The post states that the tool was originally developed for matches where both teams would be remotely connected, so neither team would have a significantly different latency (you can usually enginneer this by the first approach they suggested, of picking a server in between two locations).
jumpytrading 4y ago
Source on RNG didn’t want to attend?
StopDarkPattern 4y ago
good ol - china fuckery. lol

by dematz 4y ago

vikingerik 4y ago

You can in fact have that magical time-travel approach to get apparent latency lower than the real number. Keep a history of the game state at every tick, so you can roll back to any previous instant. If real latency from one client is 50 ms, and you're targeting 35 ms, you roll back the game state by 15 ms each time input comes in from that client, and re-run the game logic from that point forwards and send the resulting state update to all clients. Those clients will see a bit of jumpiness in the game state, which isn't ideal, but may be the lesser evil compared to enduring more latency. (I have no idea what LoL in particular supports or implements, but that's the general idea.)

meinersbur 4y ago
That's just another kind of latency. Additional time has still passed from you (or the enemy player) doing an action, and you seeing the results of that action on your screen. By contrast, the server rolling back the game state is indistinguishable from the server always being 15ms behind and the clients running the simulation ahead of time (which they do anyway for fluency [1]). The "jumpiness" is called rubberbanding [2]. [1] https://www.gabrielgambetta.com/client-side-prediction-server-reconciliation.html https://www.gabrielgambetta.com/client-side-prediction-serve... [2] https://www.urbandictionary.com/define.php?term=rubberbanding https://www.urbandictionary.com/define.php?term=rubberbandin...
jchw 4y ago
This strategy is commonly referred to as being in the class of “latency hiding” as it doesn’t actually improve latency, it just allows you to reduce apparent latency. In practice it’s still useful but I think it’s mainly useful because it allows for local input to be buffered significantly less than if you were to compensate for lag by adding buffering only. However, that is not always useful. It is mainly useful in games where netplay is handled in a lock-step fashion, where each peer needs to keep every action coherent frame-wise. That makes sense for things like fighting games, but it doesn’t make all that much for say, puzzle games, even though those can be quite latency sensitive too. Either way, there is nothing magic. At the end of the day, the soonest you can see an action is still physically unchanged. Especially since the prediction model that makes the most sense is just assuming no action. That said, I dunno enough about MOBAs to try and postulate on how much it matters, and I have no idea if LoL is using lockstep or some other approach.
- skocznymroczny 4y ago
  LoL is not using lockstep. Lockstep has two big disadvantages for a competitive game. Tt requires both client to have the full game state to run the simulation, which makes creation of maphacks trivial. The second issue with lockstep is that if one player is lagging, all players are lagging, which is also not something you'd like to have in a fast paced competitive game.

ear7h 4y ago

> I feel like that's the one scenario you'd care about, making the Busan players have 35 ping was the whole point I think the key word is "significantly". The post states that the tool was originally developed for matches where both teams would be remotely connected, so neither team would have a significantly different latency (you can usually enginneer this by the first approach they suggested, of picking a server in between two locations).

jumpytrading 4y ago

Source on RNG didn’t want to attend?

StopDarkPattern 4y ago

good ol - china fuckery. lol