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 |
|---|
Opponent snapping into a different position mid-combo? Drive Impact reactions that were clean offline and not online?
Street Fighter 6 uses rollback netcode, which is the best available answer to playing a frame-precise game across a network - and it is frequently misunderstood as a solution to bad connections. Rollback hides latency extremely well and hides packet loss very badly. Understanding which of your problems it can absorb, and which it cannot, is the difference between blaming the game and fixing something.
What this test sends: 60 packets per second of 50 bytes, matching Street Fighter 6 rather than a generic ping. verified
Rollback netcode (Capcom's own): inputs are exchanged once per 60 fps frame, peer to peer, and rolled back on mismatch.
Run the test above for at least 60 seconds. Here is what each number means for this game specifically.
Rollback works by predicting the opponent's next input and correcting when the real one arrives. When packets are lost, the correction is large and late, which you see as the opponent snapping into a completely different position. Ping produces a small, smooth prediction error; loss produces the jarring one.
A stable connection means the game predicts a consistent number of frames ahead and you stop noticing. An unstable one changes that depth constantly, so corrections vary in size and the match never settles into a rhythm. This is why a stable 70 ms often plays better than an erratic 40 ms.
At sixty frames per second, one frame is about 17 ms, so a 50 ms round trip is roughly three frames of prediction. That is well within what rollback conceals for most interactions. Punishing a Drive Impact on reaction is where the frames start to matter, and where a clean line is worth more than a close one.
What you see in game, and which number on this page explains it.
| What you see in Street Fighter 6 | Look at | What it means |
|---|---|---|
| Opponent snaps to a new position | loss | A large rollback correction after inputs failed to arrive on time. |
| Match never settles into a rhythm | jitter | Prediction depth is changing constantly instead of staying stable. |
| Reactions that work offline fail online | ping | Frames of prediction added to a genuinely reaction-based interaction. |
| Only some opponents feel bad | ping | You are measuring a different peer each match. Use the connection filter. |
A fighting game cannot afford to wait. At sixty frames per second, holding the game back until the opponent's input arrives would add the full round trip to every action, and the genre's entire skill structure is built on frame-level timing. Rollback solves this by refusing to wait: the game assumes the opponent will keep doing what they were doing, simulates forward on that assumption, and when the real input arrives it rewinds and re-simulates the affected frames.
This works remarkably well because the assumption is usually right. Players hold directions, hold buttons, and change state less often than a per-frame model implies, so most predictions are correct and no correction is needed. When one is needed and the input arrived roughly on time, the correction spans a few frames and is small enough that the eye reads it as normal movement. This is why rollback made online fighting games viable in a way that delay-based netcode never did.
The mechanism has a specific weakness, and it is not latency. Rollback's cost scales with how wrong the prediction was and how long it took to find out. A steady high ping means predicting further ahead but finding out on a predictable schedule. Packet loss means finding out late and being more wrong, so the re-simulation covers more frames and the visible correction is large - the opponent appearing somewhere they could not have walked to.
That inverts the usual advice about connection quality. In this genre, a stable connection with modest ping beats a fast one that drops packets, and the community's long insistence on cables is an accurate piece of folk engineering rather than superstition. It is also why this test's loss and jitter figures are the ones to read here, and why the ping number - the one most people quote - is the least important of the three.
If the chart above shows packet loss or high jitter, work through these in order.
The Symptom: Some matches are unplayable from the first round.
The game displays a connection quality indicator before a match and lets you restrict who you are matched with. Tightening that filter costs you search time and saves you matches that were never going to be fair. It is the most useful setting in the menu for anyone playing seriously.
The Symptom: Frequent large corrections mid-combo.
Wireless retries produce exactly the bursts of loss that rollback cannot smooth over, and the fighting game community has treated a cable as basic courtesy for a decade for that reason. A cable does not lower your ping much; it removes the loss, which is the part that actually ruins matches.
The Symptom: Links and combos drop online but not in training.
Rollback rolls the game state back and re-simulates it, which is invisible when predictions are right. If your execution is reliable offline and unreliable online, the variable is the connection rather than your hands - and this chart will show you whether the line is the cause.
The Symptom: Bad sets when someone else is streaming.
Fighting game traffic is tiny - a few inputs per frame - which is precisely why it loses to a saturated upload queue. Router QoS that prioritises small interactive packets removes a very common cause of matches that feel worse than the ping figure suggests.
No. Rollback conceals latency by predicting inputs and correcting them, which works well when the real inputs arrive on time. Lost packets mean the correction comes late and large, which is the visible snapping players complain about. Loss is the metric to watch here.
Because the game is re-simulating frames when a prediction turns out wrong, and a large correction in the middle of a link changes what you were reacting to. Consistent connections produce consistent corrections, which is what makes execution feel offline-like.
Yes, and not because of ping. Wireless links retry silently and produce bursts of loss, which is the one thing rollback cannot absorb. A cable typically changes your ping very little and changes your match quality a great deal.
At sixty frames per second, one frame is roughly 17 ms. Under about 60 ms is comfortable for nearly everything; beyond that the reaction-based interactions get harder. Stability matters more than the number in either case.