After almost 10 years. Its over.
I have discovered the culprit of mouse drifting over time when in motion.
Moreover, i have compiled and tested to confirm the remedy.
I have converted the working test into an .exe file for public testing upon release. And will be releasing the final video on this issue hopefully by today or tomorrow with the test program attached.
I will attach the file here later as well as the source code (with permission obviously) for further analysis and testing.
EDIT; after getting home and doing much more testing. The circle drifting has not entirely been eliminated, just greatly reduced.
e.g. without app running, after 10 circles my cursor is off the screen, with the app running i can make around 100 before the same result. So am deeper diving, because if it's not perfect, its not good enough. I'll keep this post updated as i progress.
EUREKA!
Re: EUREKA!
The three lords of BlurBusters
Vocaleyes - ChatGPT mouse drift expert
kyube - spiritual RAM instability detector
and myself, Hyote - lord of regedit
Vocaleyes - ChatGPT mouse drift expert
kyube - spiritual RAM instability detector
and myself, Hyote - lord of regedit
-
dermodemon
- Posts: 136
- Joined: 10 Aug 2023, 12:03
lord of placebo regedit bs you meanRe: EUREKA!
I literally have the best Windows on Earth. And I'm not schizoposting, this is factual information.
Re: EUREKA!
you also have the best shoes in the world.Hyote wrote: ↑11 Jun 2025, 13:10I literally have the best Windows on Earth. And I'm not schizoposting, this is factual information.
- Attachments
-
- hayoteeo shoes.GIF (31.91 KiB) Viewed 4782 times
-
dermodemon
- Posts: 136
- Joined: 10 Aug 2023, 12:03
Re: EUREKA!
once i was stupidly desperated trying to fix my floaty mouse and tried your "tweaks" from somewhere here. yes it broke my windows and i was getting bsods left and right. now i will never touch my system regedit values. especially from your guidesHyote wrote: ↑11 Jun 2025, 13:10I literally have the best Windows on Earth. And I'm not schizoposting, this is factual information.
Re: EUREKA!
Please find attached the app. Remember this isn't a fix, but a test. A fix would need to be developed for compatibility and performance sakes, this is just a proof of concept of the effect fixed point math has vs floating point integer rounding.
I personally observe an increase in mouse sensitivity when app is active, alongside a reduction in the time it takes for the cursor to reach the edge of the screen.
e.g. app off = 10 circles until cursor drifts off screen, app on = 100 circles for the same effect.
While this isn't "perfect mouse movement" it shows the impact that the dropped decimals from floating point has on mouse movement in comparison.
I personally observe an increase in mouse sensitivity when app is active, alongside a reduction in the time it takes for the cursor to reach the edge of the screen.
e.g. app off = 10 circles until cursor drifts off screen, app on = 100 circles for the same effect.
While this isn't "perfect mouse movement" it shows the impact that the dropped decimals from floating point has on mouse movement in comparison.
- Attachments
-
- RawMouse_FixedPoint_And_DriftTest.zip
- (6.27 KiB) Downloaded 99 times
Re: EUREKA!
I made things more sophisticated since the early days of sharing regedit tweaks. I agree it's much heavier in my batch files but they also shouldn't leave anyone with an unbootable system, provided you apply Windows 10 tweaks on Windows 10 and 11 tweaks on 11. Also you and Jimmy ask 100 questions and act like mad 5 year olds when others don't feed you directly with fixes. I could've been more harsh but I learned my lesson on how not to get a cooldown here as well.dermodemon wrote: ↑11 Jun 2025, 14:55once i was stupidly desperated trying to fix my floaty mouse and tried your "tweaks" from somewhere here. yes it broke my windows and i was getting bsods left and right. now i will never touch my system regedit values. especially from your guidesHyote wrote: ↑11 Jun 2025, 13:10I literally have the best Windows on Earth. And I'm not schizoposting, this is factual information.
Re: EUREKA!
continued...
"All mice, across all platforms, drift in the same direction relative to your input pattern."
This completely reframes the issue.
---
###
What that confirms:
* The issue is **not random**. It's **systematic** and **reproducible**.
* It’s **not tied to a specific operating system**, USB stack, or motherboard.
* It **follows your input patterns** — clockwise = bottom right drift, counterclockwise = top left.
* You’ve eliminated sensors, surfaces, power management, and platform-specific bugs.
---
###
What that *strongly suggests*:
This looks like a **delta rounding error**, **accumulation bias**, or **minor loss in input fidelity** during processing — and it’s happening across multiple devices and systems. That means the drift is not environmental noise or jitter — it’s **directionally biased accumulation**, either:
1. **In the way your movements are physically made**, or
2. **In how input deltas are interpreted or accumulated system-wide.**
---
###
Likely Candidates Now:
#### 1. **Human input + imperfect sampling resolution (aka motion bias)**
* All mice poll movement in **discrete steps** — they sample deltas (X, Y) at fixed intervals.
* When drawing continuous circles, you’re generating a lot of **small diagonal movements**, and **digitizing these movements into discrete samples** causes **rounding or truncation** errors.
* If those errors are *biased*, even slightly, they will **accumulate over time in the same direction** depending on motion pattern.
#### 2. **Coordinate rounding + fixed-point math drift**
* On many systems, cursor position is stored as **integers**, while the real-world movement is continuous.
* If a driver, OS input stack, or mouse firmware truncates decimal remainders (instead of storing sub-pixel precision), over time it can **leak small deltas**.
* Circles amplify this — as rounding errors are not evenly cancelled out due to motion bias.
#### 3. **System-wide angle quantization error**
* Movement at low angles (like subtle diagonals) can trigger a consistent rounding path, like always favoring +1 over -1 when sampling near the center between pixels.
---
###
How to Test This Theory
#### Reproducibility test:
* Open **MouseTester**.
* Draw **clockwise circles slowly** for 30 seconds.
* Then draw **anti-clockwise**.
* Repeat with:
* High speed vs slow
* Small circles vs large
* Zig-zags or L shapes
If you consistently see drift **in the same direction relative to movement**, the problem is **not noise or latency**, it’s **digitization bias**.
---
### 🛠 Can This Be Solved?
Sort of. You can:
#### Use **higher DPI** settings:
* Higher DPI = higher resolution in deltas = reduced rounding error.
* Test at 400 vs 800 vs 1600 vs 3200 DPI.
#### Use **better accumulation math**:
* Some custom input layers (like the app we built with fixed-point math) reduce this — but it still happens if sub-pixel remainders are rounded too harshly or not handled at all.
#### Use **apps or games with raw input and floating/fixed-point smoothing**:
* Raw input bypasses the OS input smoothing.
* Custom smoothing logic can preserve precision.
---
###
Final Conclusion
You are almost certainly observing a **cumulative digitization artifact** caused by:
* The **combination of your specific input patterns** (circles),
* **How input deltas are sampled**, and
* **How sub-pixel movement is rounded/truncated** by the mouse firmware or host system.
This artifact **survives across all hardware and OSes** because:
>
All systems must digitize continuous mouse motion into finite X/Y steps, and all are vulnerable to cumulative bias unless they counteract it in software — which most don’t.
"could this happen across all platforms mentioned?"
Yes — **absolutely**. All three (PC, laptop, PS4) share a **fundamental limitation** that makes them all vulnerable to:
> **Delta rounding error, accumulation bias, and minor loss of input fidelity**.
Let’s break down **why** this happens **across such different platforms**, and why it's a **universal issue** under specific conditions like yours.
---
## 1. **All platforms process mouse movement as discrete deltas**
Whether it's a:
* **Gaming PC with high polling rate support**
* **Laptop with limited USB bandwidth**
* **PS4 with very basic mouse support**
They all do **this same thing**:
> Capture physical mouse movement as a series of **(X, Y) deltas** sampled at fixed intervals (e.g., 125Hz, 500Hz, 1000Hz, etc.)
Mouse movement is **not continuous** — it's measured in **steps**, and each step is **rounded to an integer**.
Even when polling is high or DPI is high, this quantization happens.
---
##
2. **Rounding and loss of sub-pixel precision is common**
On all platforms, cursor or in-game aim updates often use **integers** for screen position:
* If you move 1.4px, the system applies 1px and may **drop or mishandle the 0.4px**.
* Over time, if rounding always favors one direction slightly (due to motion pattern, angle, or internal math quirks), that bias **accumulates**.
This is called **accumulation bias**.
*And yes — it is especially exposed during repeated circular movement* (because the errors don’t cancel out perfectly).
---
##
3. **PS4 is especially limited in input precision**
* PS4 doesn't support advanced polling rates or driver customization.
* It treats most mice as basic HID devices.
* Input is processed through a simplified abstraction layer that doesn't preserve sub-frame or sub-pixel motion perfectly.
So even though the PS4 isn’t Windows or Linux, it still:
> Receives mouse movement in discrete deltas, rounds it, and moves the virtual cursor or camera.
Which means: **same problem, same result.**
---
## So, to answer you clearly:
| Platform | Uses discrete deltas | Integer rounding likely? | Prone to drift from circular input? |
| ---------- | -------------------- | ------------------------ | ----------------------------------- |
| Windows PC | Yes | Very likely | Yes |
| Laptop | Yes | Yes (same stack) | Yes |
| PS4 | Yes | Yes (even simpler) | Yes |
---
## Conclusion
> Yes, delta rounding error, accumulation bias, and input precision loss **can and do exist across all these platforms** — because they all share the same *core mechanism* of sampling mouse movement in discrete steps and updating position in integers.
And since your drift is:
* Directionally consistent
* Input-pattern dependent
* Reproducible across platforms
…you are likely observing the **mathematical edge case of real-world digitization** — not a “bug,” but a **limitation** of how input is processed everywhere.
Hardware Recommendations to Reduce Drift
Since your issue is precision loss + rounding bias, the goal is to maximize sensor resolution and reporting accuracy:
Mouse Recommendations
Mouse Sensor Notable Features
Razer Viper V3 Pro Focus Pro 35K Top-tier tracking, ultra-stable polling
Logitech G Pro X Superlight 2 Hero 2 Excellent DPI scaling and smooth response
Zowie EC2-CW PixArt 3370 No smoothing, no filtering – very raw feel
Pulsar X2V2 PAW3395 Highly accurate, minimal smoothing, very low latency
Look for mice with PAW3395 or Hero 2 sensors and true 1000Hz+ polling. These minimize quantization loss.
It seems this is where we hit our current dead end. but I feel i've provided enough information for now and i'm tired. lol.
"All mice, across all platforms, drift in the same direction relative to your input pattern."
This completely reframes the issue.
---
###
What that confirms:* The issue is **not random**. It's **systematic** and **reproducible**.
* It’s **not tied to a specific operating system**, USB stack, or motherboard.
* It **follows your input patterns** — clockwise = bottom right drift, counterclockwise = top left.
* You’ve eliminated sensors, surfaces, power management, and platform-specific bugs.
---
###
What that *strongly suggests*:This looks like a **delta rounding error**, **accumulation bias**, or **minor loss in input fidelity** during processing — and it’s happening across multiple devices and systems. That means the drift is not environmental noise or jitter — it’s **directionally biased accumulation**, either:
1. **In the way your movements are physically made**, or
2. **In how input deltas are interpreted or accumulated system-wide.**
---
###
Likely Candidates Now:#### 1. **Human input + imperfect sampling resolution (aka motion bias)**
* All mice poll movement in **discrete steps** — they sample deltas (X, Y) at fixed intervals.
* When drawing continuous circles, you’re generating a lot of **small diagonal movements**, and **digitizing these movements into discrete samples** causes **rounding or truncation** errors.
* If those errors are *biased*, even slightly, they will **accumulate over time in the same direction** depending on motion pattern.
#### 2. **Coordinate rounding + fixed-point math drift**
* On many systems, cursor position is stored as **integers**, while the real-world movement is continuous.
* If a driver, OS input stack, or mouse firmware truncates decimal remainders (instead of storing sub-pixel precision), over time it can **leak small deltas**.
* Circles amplify this — as rounding errors are not evenly cancelled out due to motion bias.
#### 3. **System-wide angle quantization error**
* Movement at low angles (like subtle diagonals) can trigger a consistent rounding path, like always favoring +1 over -1 when sampling near the center between pixels.
---
###
How to Test This Theory####
Reproducibility test:* Open **MouseTester**.
* Draw **clockwise circles slowly** for 30 seconds.
* Then draw **anti-clockwise**.
* Repeat with:
* High speed vs slow
* Small circles vs large
* Zig-zags or L shapes
If you consistently see drift **in the same direction relative to movement**, the problem is **not noise or latency**, it’s **digitization bias**.---
### 🛠 Can This Be Solved?
Sort of. You can:
####
Use **higher DPI** settings:* Higher DPI = higher resolution in deltas = reduced rounding error.
* Test at 400 vs 800 vs 1600 vs 3200 DPI.
####
Use **better accumulation math**:* Some custom input layers (like the app we built with fixed-point math) reduce this — but it still happens if sub-pixel remainders are rounded too harshly or not handled at all.
####
Use **apps or games with raw input and floating/fixed-point smoothing**:* Raw input bypasses the OS input smoothing.
* Custom smoothing logic can preserve precision.
---
###
Final ConclusionYou are almost certainly observing a **cumulative digitization artifact** caused by:
* The **combination of your specific input patterns** (circles),
* **How input deltas are sampled**, and
* **How sub-pixel movement is rounded/truncated** by the mouse firmware or host system.
This artifact **survives across all hardware and OSes** because:
>
All systems must digitize continuous mouse motion into finite X/Y steps, and all are vulnerable to cumulative bias unless they counteract it in software — which most don’t."could this happen across all platforms mentioned?"
Yes — **absolutely**. All three (PC, laptop, PS4) share a **fundamental limitation** that makes them all vulnerable to:
> **Delta rounding error, accumulation bias, and minor loss of input fidelity**.
Let’s break down **why** this happens **across such different platforms**, and why it's a **universal issue** under specific conditions like yours.
---
##
1. **All platforms process mouse movement as discrete deltas**Whether it's a:
* **Gaming PC with high polling rate support**
* **Laptop with limited USB bandwidth**
* **PS4 with very basic mouse support**
They all do **this same thing**:
> Capture physical mouse movement as a series of **(X, Y) deltas** sampled at fixed intervals (e.g., 125Hz, 500Hz, 1000Hz, etc.)
Mouse movement is **not continuous** — it's measured in **steps**, and each step is **rounded to an integer**.
Even when polling is high or DPI is high, this quantization happens.
---
##
2. **Rounding and loss of sub-pixel precision is common**On all platforms, cursor or in-game aim updates often use **integers** for screen position:
* If you move 1.4px, the system applies 1px and may **drop or mishandle the 0.4px**.
* Over time, if rounding always favors one direction slightly (due to motion pattern, angle, or internal math quirks), that bias **accumulates**.
This is called **accumulation bias**.
*And yes — it is especially exposed during repeated circular movement* (because the errors don’t cancel out perfectly).---
##
3. **PS4 is especially limited in input precision*** PS4 doesn't support advanced polling rates or driver customization.
* It treats most mice as basic HID devices.
* Input is processed through a simplified abstraction layer that doesn't preserve sub-frame or sub-pixel motion perfectly.
So even though the PS4 isn’t Windows or Linux, it still:
> Receives mouse movement in discrete deltas, rounds it, and moves the virtual cursor or camera.
Which means: **same problem, same result.**
---
##
So, to answer you clearly:| Platform | Uses discrete deltas | Integer rounding likely? | Prone to drift from circular input? |
| ---------- | -------------------- | ------------------------ | ----------------------------------- |
| Windows PC |
Yes | Very likely | Yes || Laptop |
Yes | Yes (same stack) | Yes || PS4 |
Yes | Yes (even simpler) | Yes |---
##
Conclusion> Yes, delta rounding error, accumulation bias, and input precision loss **can and do exist across all these platforms** — because they all share the same *core mechanism* of sampling mouse movement in discrete steps and updating position in integers.
And since your drift is:
* Directionally consistent
* Input-pattern dependent
* Reproducible across platforms
…you are likely observing the **mathematical edge case of real-world digitization** — not a “bug,” but a **limitation** of how input is processed everywhere.
Hardware Recommendations to Reduce Drift
Since your issue is precision loss + rounding bias, the goal is to maximize sensor resolution and reporting accuracy:
Mouse RecommendationsMouse Sensor Notable Features
Razer Viper V3 Pro Focus Pro 35K Top-tier tracking, ultra-stable polling
Logitech G Pro X Superlight 2 Hero 2 Excellent DPI scaling and smooth response
Zowie EC2-CW PixArt 3370 No smoothing, no filtering – very raw feel
Pulsar X2V2 PAW3395 Highly accurate, minimal smoothing, very low latency
Look for mice with PAW3395 or Hero 2 sensors and true 1000Hz+ polling. These minimize quantization loss.It seems this is where we hit our current dead end. but I feel i've provided enough information for now and i'm tired. lol.