Measuring…
The verdict for this game appears once the first full second of packets is in.
Comparing with other tests to this server…
Measuring…
The verdict for this game appears once the first full second of packets is in.
Comparing with other tests to this server…
Are you sure you want to delete all results?
| Date & Time | Duration | Avg Ping | Avg Jitter | Packet Loss | Actions |
|---|
Losing a round to one hit you were sure you blocked? Opponent snapping across the screen mid-Roman Cancel?
Guilty Gear -Strive- combines two things that interact badly with an unreliable connection: rollback netcode, which is the best available way to play a frame-precise game online, and a damage model in which a single confirmed opening can take most of a health bar. Rollback handles latency well. What it does not do is make a mispredicted frame cheap - and in this game, one mispredicted frame at the wrong moment is the round.
What this test sends: 60 packets per second of 40 bytes, matching Guilty Gear -Strive- rather than a generic ping. verified
Rollback netcode (GGPO-style): one input exchange per 60 fps frame.
Run the test above for at least 60 seconds. Here is what each number means for this game specifically.
Rollback predicts the opponent's input and re-simulates when the real one arrives late. In a game with long, grindy rounds a bad correction costs a little health. In Strive, the interaction that got corrected was frequently the one that decides the round, because a single confirmed hit converts into most of a health bar.
Blocking a mix-up is a trained reaction to a specific visual cue at a specific frame. Consistent latency lets you adapt to it; varying prediction depth means the cue arrives a different number of frames early each time, and the reaction you drilled stops transferring. This is why an unstable 40 ms feels worse than a steady 70 ms.
One frame is roughly 17 ms at sixty frames per second. Strive complicates the arithmetic because Roman Cancel briefly slows the game's own clock, so the relationship between milliseconds and frames is not constant during those moments. A stable line is what keeps that manageable.
What you see in game, and which number on this page explains it.
| What you see in Guilty Gear -Strive- | Look at | What it means |
|---|---|---|
| Opponent snaps position mid-pressure | loss | A large rollback correction after inputs failed to arrive on time. |
| Blocking reactions that work offline fail online | jitter | Prediction depth is varying, so the visual cue arrives at a different frame each time. |
| Whole rounds decided by one exchange | loss | High damage per opening turns a single bad correction into the round. |
| Timing feels off during Roman Cancel | jitter | The game's clock slows there, so instability is felt more sharply. |
Rollback netcode is usually discussed as though its cost were uniform: some frames get re-simulated, there is a small visual correction, and the game carries on. That framing is fine for a title where rounds are long and health bars deplete gradually, because a corrected exchange costs a small fraction of the outcome. It stops being adequate when the game is built around single decisive interactions.
Strive is built exactly that way. The design intent was to make every opening matter, and it succeeds: a confirmed hit frequently converts into most of a health bar and a positional advantage that carries into the next round. That is a legitimate and popular design decision, and it also means the variance introduced by a network correction lands on the most important moments rather than being spread thinly across a long match.
The series' history makes the improvement easy to see. Earlier Guilty Gear titles used delay-based netcode, in which both machines simply wait for the other's input before advancing a frame. That is honest and consistent but it adds the full round trip to everything, and it makes distance genuinely unplayable. Rollback replaced waiting with prediction, and it is the reason cross-country matches are viable at all.
None of that changes what prediction cannot do. A packet that does not arrive is not a slightly later input; it is an unknown one, and the re-simulation that follows spans more frames and moves more state. In a game where a single interaction is the round, that is the difference between an online experience that resembles offline and one that resembles a coin flip - which is why loss, not ping, is the number to read on this page.
If the chart above shows packet loss or high jitter, work through these in order.
The Symptom: Rounds lost to single corrected exchanges.
Wireless produces bursts of loss, and a burst during a mix-up is a health bar in this game specifically. A cable barely changes your ping and removes the bursts entirely, which is the trade that matters in a title with this damage model.
The Symptom: Some opponents are consistently unplayable.
Strive shows connection quality before you commit to a match. Declining the ones that read poorly costs a little search time and saves sets that were decided by the network rather than by either player - which is a better use of an evening than grinding through them.
The Symptom: Blocking works in training and not online.
If your reactions hold up offline and collapse online, the variable is the connection. Adapting your habits to compensate for jitter teaches you timing that is wrong everywhere else. Measure first, then decide whether there is anything to adapt to.
The Symptom: Bad sets that line up with someone streaming.
The game sends a handful of tiny packets per frame, which is the traffic profile most easily starved by a saturated upload. Router QoS that prioritises small interactive packets removes it without affecting anything else in the house.
Yes, and it was a deliberate change of direction for the series - earlier Guilty Gear titles used delay-based netcode, where both machines wait for inputs instead of predicting them. Rollback is the reason online play in Strive feels closer to offline than its predecessors did.
Because Strive's damage per confirmed opening is unusually high. A network correction that would cost a few percent of a health bar in another fighting game can cost most of one here, so the same connection quality has a much larger effect on results.
Loss, clearly. Rollback is designed to hide a few frames of latency and does it well. It cannot hide packets that never arrive, because the correction then spans more frames and becomes visible as the opponent snapping to a new position.
The mechanic is the same, but it briefly slows the game's own clock, so the number of real milliseconds in a frame changes during it. On an unstable connection that makes an already demanding timing window less consistent than it is offline.