Steam
Action RPG · Updated 7 September 2026

Elden Ring parry timing: every parry window in milliseconds

The widest parry in this game is 200 milliseconds — and it opens 200 ms before the blow lands.

Every wiki publishes the parry table in frames, which is a unit nobody can feel, and every guide says to time it well without saying what the time is. Here it is in milliseconds, next to the one number that decides whether you can use it: how far apart your own presses land. A window this narrow is shorter than a human reaction, so a parry is never a response — it is a prediction, and consistency beats speed.

Press when the blow lands

Not when you think you should parry — when it actually lands. The lead is arithmetic and this page can do it for you; what it cannot do is guess how far apart your presses fall, which is the half that decides everything. 8 attempts before the numbers mean anything, and the swing is a different length every time so you cannot count it.

Click, tap or press Space on the line
300 ms earlythe blow300 ms late
Your scatter—
Your bias—
Lead needed—
As you pressed—
With the lead—
Attempts0
0 of 8

Press when the blow reaches the line.

Keep and share this result

Save the numbers to your private Device Passport in this browser, or copy a card that shows exactly what was measured.

Open passport

Nothing is uploaded unless you contribute. Cards and links contain only the numbers shown above.

What this will not tell you Whether you would have parried that boss. It times your press against the moment this page drew the blow landing, and between that draw and your eye sits your own display, which no browser can see. So the scatter transfers — the spread of a hand is the spread of a hand on any screen — and the bias carries your screen's latency inside it. That is why the page leads with the scatter and treats the bias as the half you can simply stop applying. It also makes no claim about which attacks in the game are parryable at all; many are not, and nothing on screen says so.

Every parry in the game, in milliseconds

The frame counts are published everywhere and they are on the game's own 30 fps animation clock, so one frame is 33.3 ms. Read as time, the table says something the frame column cannot: the widest ordinary window is 200 ms and the tightest is 100, and a human reaction to something unexpected is around 250. You are not reacting to the swing. You are predicting it.

Parrying with Window Press this early Recovery A miss costs
Storm Wall, against a projectile 333 ms 0 to 333 ms 13f 767 ms
Buckler Parry 200 ms 100 to 300 ms 12f 700 ms
Parry with a small shield 200 ms 133 to 333 ms 11f 700 ms
Parry with a medium shield Same numbers as any fist, any claw, any thrusting sword, Storm Wall against a weapon, Carian Retaliation, Thops’s Barrier against a weapon. 200 ms 133 to 333 ms 13f 767 ms
Parry with a dagger 167 ms 100 to 267 ms 13f 700 ms
Parry with a curved sword 133 ms 100 to 233 ms 14f 700 ms
Golden Parry 100 ms 133 to 233 ms 16f 767 ms
Thops’s Barrier, against a spell 100 ms 100 to 200 ms 17f 767 ms
The one you can actually react with Storm Wall, against a projectile has no startup at all and 10 active frames — 333 ms, more than half again the widest weapon parry, with nothing to lead. It is by a distance the most forgiving deflection in the game and nobody talks about projectile parrying as the easy one.

The parry everyone recommends is the hardest one in the game

Golden Parry gets recommended because its bubble reaches further out, which makes it forgiving in space. Nobody checks what it costs in time, and the answer is both halves at once.

100 msits window

Half the Buckler's

3 active frames against 6. The tolerance you are given to place the press is exactly halved, and the press is the whole skill.

16fits recovery

A third longer than the Buckler's

Against 12 frames. So it is harder to land and it leaves you open for longer when you do not — the two costs point the same way instead of trading off.

67 msmore commitment

Per attempt, every attempt

767 ms from press to free, against 700. Over a boss fight that is a different animal standing in front of you.

Which is not an argument against it Reach is real, and a parry that catches an attack you would otherwise have been out of position for is worth something. The argument is against learning on it. The Buckler has the shortest startup, the widest window and the shortest recovery of anything with a startup at all — it is the tool to measure yourself against, and this page's bench defaults to it for that reason.

A light roll buys you no extra invulnerability at all

This is the most repeated piece of wrong advice about this game, and the wiki states the correction in one line that nobody quotes: light and medium rolls are identical, and light only travels further. The entire invulnerability penalty for rolling heavy is one frame — 33 milliseconds.

Roll Equip load Invulnerable Recovery What it actually costs
Light roll under 30% 14f · 467 ms the same as medium Nothing, if you can carry it. The armour you gave up to get here is the cost.
Medium roll 30% to 70% 14f · 467 ms the same as light None. This is the same invulnerability as a light roll, for whatever armour fits under 70%.
Heavy roll over 70% 13f · 433 ms almost twice as long One frame of invulnerability — and a recovery animation that roughly doubles, slower stamina regeneration, and more fall damage.
Overloaded 100% and over none no roll at all You cannot roll or backstep. The character staggers forward with no invulnerability of any kind.
So what is heavy roll actually punishing you for The recovery, not the dodge. A heavy roll's recovery animation is close to twice as long, which is the window a boss's second swing arrives in — and it slows stamina regeneration and raises fall damage on top. Dropping under 70% is worth doing. Dropping under 30% buys distance and nothing else, and the armour you take off to get there is a real price paid for a benefit that is not invulnerability.
Where the sources disagree, and it does not change the answer The wiki read for this page on 2026-09-21 gives 14 invulnerable frames on a light or medium roll and 13 on a heavy one. A second source gives one fewer across the board — the same gap, counted from a different boundary frame. Sources differ by one frame on the absolute count and agree exactly on the comparison. Both agree on the two things above: light and medium are the same, and heavy costs a single frame.

The four bands, and the one that stops you rolling

Equip load is a percentage of your maximum, and the bands are hard edges rather than a slope: 29.9% and 30.0% are two different rolls. The band most people never see is the fourth one.

100%and over

Overloaded removes the roll entirely

Not a slower roll — no roll and no backstep. The character staggers forward with no invulnerability of any kind, which is a different state from heavy rather than a worse version of it.

OffCamera

Auto Target

With it on, landing a hit while not locked on drops you into lock-on anyway — so the camera snaps to something you did not choose, usually in the middle of a roll. This is the setting behind most 'the camera fought me' complaints.

Well below your free-lookCamera

Camera sensitivity (locked on)

Two separate numbers. A high free-look sensitivity is good for reading a room fast, and the same number while locked on makes the small corrections a boss fight needs impossible. Set them apart deliberately.

On for big weaponsCamera

Manual Attack Aiming

Lets you steer a heavy swing while locked on rather than having the game point it. Matters for colossal weapons where the arc decides whether you hit one enemy or three.

22keyboard binds

The reference, verified and dated

Every default on keyboard and on 16 pad actions, with all 5 settings and the reasoning behind each. Open the control reference →

Everything this site knows about Elden Ring

Two pages, and they are not interchangeable: one is the reference you look a binding up in, the other measures the hand that has to hit a 200 ms window. Both figures below are counted from what each page holds.

Asked about parrying

The questions this page can answer.

How long is the parry window in Elden Ring?

Between 100 and 200 milliseconds, depending on what you are parrying with. The published frame data is on the game’s 30 fps animation clock, so one frame is 33.3 ms: a Buckler Parry and most shields are 6 active frames (200 ms), a dagger is 5 (167 ms), a curved sword is 4 (133 ms) and Golden Parry is 3 (100 ms). The one exception is Storm Wall against a projectile, which is 10 frames — 333 ms — with no startup at all.

Why do my parries in Elden Ring feel too late?

Because the window opens after the press, not on it. Every parry has startup frames — 3 or 4 for almost everything — before it can deflect anything, so the button has to go down between 100 and 233 ms before the blow lands. A player pressing at the moment of impact is pressing a full window late, every single time, and it feels like bad timing rather than a missing lead.

Is Golden Parry the best parry in Elden Ring?

It reaches further than anything else, which is real, but on timing it is the hardest tool in the game: 3 active frames against the Buckler’s 6, and 16 recovery frames against the Buckler’s 12. Half the window and a third longer to punish a miss, with both costs pointing the same way. It is a fine tool once your press is steady; it is the wrong one to learn on, which is the opposite of how it is usually recommended.

Does light roll give more i-frames than medium roll in Elden Ring?

No. The game’s own wiki states it outright: light and medium rolls have identical invincibility and identical recovery, and light only travels further. The whole invulnerability penalty for rolling heavy is one frame — 33 milliseconds. What a heavy roll actually costs is the recovery animation, which is close to twice as long, plus slower stamina regeneration and more fall damage.

What happens if you go over 100% equip load in Elden Ring?

You stop being able to roll or backstep at all. Overloaded is not a slower roll, it is no roll: the character staggers forward with no invincibility frames of any kind. The bands are hard edges rather than a slope — under 30% is light, 30% to 70% is medium, over 70% is heavy, and 100% and over is overloaded — so 29.9% and 30.0% are two different characters.