Summary
Shields do not seem to grant a longer active parry window, which appears to be 13 frames regardless of shield usage.
Method
Testing was done on 28175. Two equipment sets were tested. No enchantments, facets, rings or other items were used. The sets were as follows:
- a Pig Sticker and empty off-hand,
- a Pig Sticker and Wooden Shield.
For each set, I recorded around eight minutes of attempts to parry the dummy at 60 fps. I used vsync and inspected frame timings to minimise frame-counting errors. Then, stepping through the footage of failed and successful parries frame-by-frame, I determined, counting from the first animation frame:
- A: the latest frame in which I could still take damage,
- B: the earliest frame in which I could successfully parry,
- C: the latest frame in which I could successfully parry,
- D: the earliest frame in which I could take damage again,
- E: the earliest frame in which I could move after initiating the parry.
The differences B–A and D–C should be 1 frame. This is both a sanity check (values must be non-negative) and a sign that enough data has been acquired (differences of 2 or more would suggest I should’ve recorded more). C–B+1 is the active parry window.
Errors for A to D should be within 1 frame. This means differences B–A and D–C may be 0.
While E makes recovery comparable, moving tends to be the slowest recovery cancel (compared to attacking, dodging, or blocking) in Souls games. Beginning of movement is difficult to determine due to acceleration and a lack of a trackable fixed point. Testing dodge would have been smarter.
Results
| Set | A | B | C | D | E |
|---|---|---|---|---|---|
| Pig Sticker, empty off-hand | 4 | 5 | 17 | 18 | 26 |
| Pig Sticker, Wooden Shield | 6 | 7 | 19 | 20 | 46 |
This is an active parry window of 13 frames for both sets.
Differences B–A and D–C look sensible.
The window begins and ends two frames later for the Wooden Shield relative to the visible start of the parrying animation.
Discussion
The two-frame shift between shield and no shield may be explained by animation differences. This is supported by the identical length of the window. Also see below; in an older game version, startup for shields was closer to my data for the estoc.
From frame timings and how frequently specific frames occur in my data, I would also consider an active parry window of 14 frames plausible, but decidedly not 12 frames.
Time until walking is possible is very short for empty off-hands and considerable for the shield. If anyone feels like testing earliest-cancel frames, please do.
As a point of reference, non-special shields in Dark Souls PtDE have 7 active parry frames at 30 fps, equivalent to 14 frames in 60 fps.
There’s still a chance frame data differs by weapon type or even by weapon.
Additional Testing
On a hunch, I tested one set in the crucible_legacy branch by parrying one of the Risen.
| Set | A | B | C | D | E |
|---|---|---|---|---|---|
| Blood-Rusted Sword, Wooden Shield | 4 | 5 | 15 | 16 | 44 |
An 11-frame parry window nearly matches Dammitt’s database. Speculatively, frame data in Dammitt’s database is based on an old build and a 60-fps metric. It does not match current builds in either a 30-fps or 60-fps metric.
I expected a larger difference between legacy and live, but that just shows the impact of a small difference in frames.
(I also tested cancelling the recovery by dodging, which was possible from frame 35. Testing moving really wasn’t optimal.)
@Silas-Inservio-Pax, @sBAM: Please forgive the ping. I believe this may be of interest to you.
