22 Hz
one update every 45.5 ms
What this test sends:
22 packets per second of 80 bytes, matching StarCraft II rather than a generic ping.
verified
SC2's simulation runs at 22.4 game loops per second on Faster speed - the number every build order and APM figure is measured against.
A note on the rate: StarCraft II's simulation advances 22.4 times per second on Faster speed - the figure every build order timing in the game is measured against. Our catalogue holds whole numbers, so this test sends 22 packets per second. That is deliberately stated rather than rounded silently, and it does not change what the measurement tells you about your line.
Run the test above for at least 60 seconds. Here is what each
number means for this game specifically.
Packet loss (the lag bar's cause)
Because both machines must advance together, a missing input cannot be skipped - it has to arrive. Loss therefore converts directly into the game pausing while it waits, and if it goes on long enough the other player is offered the option to drop you. There is no graceful degradation to hide behind.
Jitter (timings that stop being repeatable)
Build orders are drilled to the second and measured in game loops. A connection that delivers unevenly makes the gap between issuing a command and it taking effect vary, which is exactly the thing a practised timing cannot absorb. Ladder players feel this long before it becomes a visible lag bar.
Ping (and the region you are laddering in)
Your account is tied to a regional server, so your baseline latency is largely fixed by where you play rather than something to tune. Under 80 ms keeps micro comfortable. What matters more is that the figure is steady, because everything in this game is a rehearsed timing.
What you see in game, and which number on this page explains it.
A late-game StarCraft II match has hundreds of units, each with a position, an order queue and a target, all changing constantly. Streaming that state to both players the way a shooter streams player positions would be enormous, and it would also be unnecessary, because the game is fully deterministic: given the same commands in the same order, both machines compute exactly the same battle.
So the network carries commands rather than consequences. You issue an order, it is distributed, and both simulations apply it at the same game loop. This is elegant and cheap, and it has one hard requirement: every machine must have every command before it can advance. There is no equivalent of dropping a stale position update, because there are no position updates to drop.
That requirement is what produces the game's distinctive failure. When a command does not arrive, the simulation cannot proceed without inventing something, and inventing something would put the two machines permanently out of agreement. So it stops, tells you who it is waiting for, and resumes when the input turns up. It is a genuinely more honest design than a game that quietly guesses and corrects, and it is far less pleasant to be on the wrong end of.
The practical guidance follows directly. Steady latency is cheap here, because a constant delay is something a player rehearses around and every build order in the game was learned on some amount of it. Bandwidth is nearly free, because commands are tiny. Packet loss is expensive, because it is the only one of the three that can stop the clock - and jitter is expensive in a subtler way, because it attacks the timing precision that the entire competitive game is built on.
If the chart above shows packet loss or high jitter, work through
these in order.
Why does StarCraft II freeze instead of lagging?
Because both machines run the same simulation in step. A missing input cannot be skipped, so the game waits for it rather than guessing - which is why you get a pause and a named lag bar rather than units warping around.
What is 22.4 game loops per second?
It is the rate at which the simulation advances on Faster speed, and it is the unit every build order timing in the game is really measured in. It is why the community discusses timings in seconds that do not divide evenly into anything obvious.
Can my opponent drop me for lagging?
In a lockstep game the session cannot advance without every player, so there is a mechanism to remove a participant who is holding it up. That is the practical consequence of everyone sharing one clock.
Does packet loss matter more than ping in StarCraft II?
Yes. Steady latency simply shifts every timing by a constant you can learn. Loss stops the shared simulation until the missing input arrives, which is a categorically worse failure and the one that produces lag bars.