A stability filter that discards any task whose configurations disagree with themselves sounds conservative. In one evaluation suite of 159 tasks, Erik Hill found that a strict version of this filter kept fewer usable tasks as he added repetitions, and that its statistical signal weakened when he pooled the data. The models did not change. The measurement rule and the way it handled accumulated data did.
What the strict rule did
Hill compared two models on a frozen suite of 159 tasks. His first rule, called strict, discarded a task whenever a configuration disagreed with itself across repetitions. On repetitions 1–3, it produced a 7–1 comparison over 8 informative tasks, with p=0.070. Because it had thrown away 13 tasks in that window, Hill wrote a second rule, rate, that tolerated a minority of disagreeing repetitions. On the same repetitions, rate yielded 13–2 over 15 informative tasks, with p=0.0074.
As an Amazon Associate I earn from qualifying purchases.
Hill acknowledges that he wrote rate after seeing the strict result. That sequence is the first reason to read the improvement cautiously. The second is the replication below.
Do these 3 things before closing this tab:
1Repair Windows errors before they cause bigger problems2Fix the driver behind crashes, sound loss and screen glitches3Clear out junk files and repair common Windows errorsThe preregistered replication
Hill then preregistered a replication on repetitions 4–6 and predicted that strict would again fail to reach significance. It did not fail. It produced a 9–1 comparison over 10 informative tasks, with p=0.0215. Hill describes this as stronger than he predicted, not as a prediction that came true.
#1 Best Overall
- MADE FOR THE MAKERS: Create; Explore; Store; The T7 Portable SSD delivers fast speeds and durable features to back up any endeavor; Build your video editing empire, file your photographs or back up your blogs all in an instant
- SHARE IDEAS IN A FLASH: Don’t waste a second waiting and spend more time doing; The T7 is embedded with PCIe NVMe technology that brings fast read and write speeds up to 1,050/1,000 MB/s¹, making it almost twice as fast as the T5
- ALWAYS MAKE THE SAVE: Compact design with massive capacity; With capacities up to 4TB, save exactly what you need to your drive – from large working files to game data and everything in between
- ADAPTS TO EVERY NEED: Whether using a PC or mobile phone, count on the T7 for extensive compatibility²; It’s a true team player when it comes to heavy-duty application usage or file-saving
- HI RESOLUTION VIDEO RECORDING: Record Ultra High Resolution (4K 60fs) videos directly onto the T7 Portable SSD with your favorite camera or mobile devices; Supports iPhone 15 Pro Res 4K at 60fps video and more³
The results so far, as Hill reports them:
| Window | Rule | Repetitions | Discarded tasks | Informative tasks | Win–loss record (as reported) | p-value (as reported) |
|---|---|---|---|---|---|---|
| First comparison | strict |
1–3 | 13 | 8 | 7–1 | 0.070 |
| First comparison | rate |
1–3 | not stated | 15 | 13–2 | 0.0074 |
| Preregistered replication (fresh data) | strict |
4–6 | not stated | 10 | 9–1 | 0.0215 |
| Pooled windows | strict |
1–6 | 17 | 8 | not stated | 0.0703 |
Every figure is from Hill’s post on DEV Community, which does not show a publication year in the article header; it reads “Posted on Sep 23.” The discard counts for the replication window and for rate are not stated in the post.
Why pooling lost power
The title’s “twice the data, less power” describes what happens when the two three-repetition windows are combined under strict. Discards rose from 13 in the first window to 17 in the pooled six-repetition set. Informative tasks fell to 8 in the pooled set, down from 10 in the fresh window. The p-value moved from 0.0215 to 0.0703, which crosses the conventional 0.05 threshold in the wrong direction for a claim of difference.
Rank #2
- Get NVMe solid state performance with up to 1050MB/s read and 1000MB/s write speeds in a portable, high-capacity drive(1) (Based on internal testing; performance may be lower depending on host device & other factors. 1MB=1,000,000 bytes.)
- Up to 3-meter drop protection and IP65 water and dust resistance mean this tough drive can take a beating(3) (Previously rated for 2-meter drop protection and IP55 rating. Now qualified for the higher, stated specs.)
- Use the handy carabiner loop to secure it to your belt loop or backpack for extra peace of mind.
- Help keep private content private with the included password protection featuring 256‐bit AES hardware encryption.(3)
- Easily manage files and automatically free up space with the SanDisk Memory Zone app.(5). Non-Operating Temperature -20°C to 85°C
Hill’s explanation is mechanical. Under strict, a task is removed as soon as any configuration disagrees with itself. Every additional repetition gives a noisy task another chance to show that disagreement, so the discard set can grow faster than the set of tasks that remain usable. Hill reports that two repetitions averaged roughly 8.7 discarded and 10.3 informative tasks, and three repetitions yielded 13 discarded and 8 informative tasks. These are approximate averages from his post, not a fixed property of the suite.
Hill also wrote a line that generalizes the pattern to any filter of this kind: “If the discard count rises faster than your informative count, your filter is spending your sample size, and the direction of that trade is not obvious from the code.” He also argues that “A conservative rule is not a free choice.” Both sentences are his own words.
Rank #3
- Capacity Display Variance: 250GB external ssd often appears as around 232GB on Windows. MacOS can show full 250 GB capacity. This is binary calculation difference and doesn’t affect SSD hard drive actual physical storage
- 1050 MB/s Speed: Instantly access to your files with blazing-fast 10Gbps external SSD read up to 1050MB/s and write up to 1000MB/s. LED Light indicates USB SSD instant activity
- Data Security: Solid state drives S.M.A.R.T. health diagnostics and adaptive TRIM optimizing data block management ensures consistent write speeds and extends the longevity of the portable SSD
- USB-C & USB-A Cable: Both cables featuring rapid USB 3.2 Gen2, this USB SSD effortlessly bridges devices, enabling seamless cross-platform file transfers and backup between computers, smartphones, tablets and iPhone
- Always Fast: No slowdowns for large file transfers. With SLC caching (25% of current available capacity allocated as high-speed cache), this external SSD delivers steady 10Gbps for transfers within the cache capacity
What the evidence does and does not establish
- One suite, two models. The suite is made of trap questions, and Hill explicitly warns against reading the result as a capability ranking of the models.
- Timing is confounded with the pooling. The first comparison and the replication ran at different times. Pooling can mix sampling periods. Hill attributes the loss of power to the rise in discards but has not run the analysis that would separate that explanation from the timing difference.
ratemay simply be the better rule. Hill concedes that the discard analysis could be a defense of a mistake. He treats the fresh-data replication as evidence against that reading, but describes a single replication with two models as limited evidence.- Implementation details are the author’s account. He says the
rate_marginparameter was fixed at 0.5 in advance and that the--alphaand margin flags were later removed from the command line. These details come from the post and have not been independently checked. - No independent review. I found no independent replication or statistical review of these results.
How to check whether your filter is discarding too many tasks
- For each configuration, log the number of discarded tasks and the number of informative tasks after every batch of repetitions, not only at the end of the run.
- Compute the change in each count when you add a batch. If discards grow faster than informative tasks, the filter is consuming sample size.
- Report the paired result at each window separately, then pooled. A fresh-data result and a pooled result can disagree, and you should see both.
- Record the threshold and margin parameters in your run configuration before you look at outcomes, and keep that record with the results.
- If you change the rule after seeing results, label it as post hoc and rerun it on fresh repetitions before treating the gain as established.
The linked pi-eval repository documents a deterministic grading setup with fixed predicates, suite fingerprinting, and handling for inconclusive comparisons. It is useful for seeing how a suite and its comparisons are structured, but it does not confirm Hill’s findings.
Independent reader supportYour contribution helps us test, update, and keep practical guides available for everyone.The trade-off in one comparison
The two rules answer the same question under different assumptions. The table below uses only what Hill reports.
Rank #4
- Easily store and access 2TB to content on the go with the Seagate Portable Drive, a USB external hard drive
- Designed to work with Windows or Mac computers, this external hard drive makes backup a snap just drag and drop
- To get set up, connect the portable hard drive to a computer for automatic recognition no software required
- This USB drive provides plug and play simplicity with the included 18 inch USB 3.0 cable
- The available storage capacity may vary.
| Axis | strict |
rate |
|---|---|---|
| How within-configuration disagreement is defined | Any disagreement across repetitions discards the task | Tolerates a minority of disagreeing repetitions |
| Tasks discarded and retained | 13 discarded and 8 informative in repetitions 1–3; 17 discarded and 8 informative when pooled to 1–6 | 15 informative in repetitions 1–3; discard count not stated |
| Paired result and uncertainty | 7–1, p=0.070 (1–3); 9–1, p=0.0215 (4–6 fresh); p=0.0703 pooled | 13–2, p=0.0074 (1–3); not replicated in the post |
| Fresh or pooled | Both, as reported | Only the first window is reported |
| Parameters fixed before results | No margin parameter is described | rate_margin fixed at 0.5 in advance, per Hill; the rule itself was written after the first strict result |
A strict rule can keep unstable tasks out of your counts, and that protection is real. It also shrinks the sample you have left to compare. The post does not establish a universal threshold; it shows that the balance has to be measured, repetition by repetition, for the suite you actually run.
The Tool Desk
Outbyte PC Repair FREEClear out junk files and repair common Windows errorsFree Scan →Outbyte Driver Updater FREEFix the driver behind crashes, sound loss and screen glitchesFind Drivers →The original post is Erik Hill’s Twice the data, less power on DEV Community.
Quick Recap
Best Value
- NEARLY 2X FASTER THAN OUR PREVIOUS GENERATION(8) – move 1,000 high-res photos in under 60 seconds(6) with up to 2000MB/s transfer speeds(2).
- IP65 RATING AND UP TO 3M DROP PROTECTION(3) – protects against spills and drops.
- POCKET-SIZED – fits easily in pockets and small bags.
- SPACE TO OWN YOUR AI CONTENT – speed and capacity to download your high-res clips and photo edits.
- 256-BIT AES ENCRYPTION(4) – helps keep private files secure with password protection.
“
Product prices and availability are accurate as of the date/time indicated and are subject to change. Any price and availability information displayed on Amazon at the time of purchase will apply.
