LOWSET-01EXPERIMENTAL

Low-set statistical signal and operational gate

LOWSET-01 measured whether a constrained candidate family changed the frequency of a registered witness condition. The statistical direction was consistent across eight comparisons. The operational implementation, however, was approximately 97.94 times worse than the scanner under a gate that required at least an equivalent 100-fold improvement. The signal is therefore published as an observation, not as faster mining.

Published: 2026-08-14
Updated: 2026-08-22
Authors: BTC PoW Lab
Replication: Internally audited; operational gate failed

Mining advantage claimNO DEMONSTRATED MINING ADVANTAGE

Local structure, statistical signal and operational advantage are evaluated separately.

Question

Does the registered low-set construction produce a measurable witness-rate difference, and can that difference survive end-to-end implementation cost?

Hypothesis

A stable statistical enrichment may exist, but it is only operationally relevant if the complete implementation clears the frozen scanner-equivalent cost gate.

Why this matters

Statistical enrichment and useful mining work are different claims. Measuring both prevents a local signal from being mislabeled as economic or computational advantage.

Method

Compare the frozen T0 and M20 populations, audit the full hash count and B16 witnesses, compute a confidence interval with multiplicity correction, then evaluate the implementation against the preregistered operational gate.

Experimental setup

Baseline: Conventional scanner performing equivalent candidate evaluation.
Setup: Eight registered comparisons over 68,719,476,736 audited hashes, with B16 witness counting and a separate implementation-cost measurement.
Hardware: Classical CPU/GPU research pipeline; no stock-ASIC speedup test.
Dataset: Internal audited run manifests and witness counts; publication package under review.

Results

Audited hashes68.72B

Observed ratio1.515×

Operational gateFAILED

T0–M20 ratio: 1.515×; all 8 of 8 registered comparisons were favorable.
95% confidence interval: [1.270, 1.863]; Holm-adjusted q = 0.00889.
68,719,476,736 hashes audited and 1,048,845 B16 witnesses recorded.
Operational gate failed: the implementation was approximately 97.94× worse than the scanner, where the frozen gate required an equivalent improvement of at least 100×.

ANALYSIS / MODEL

Scientific analysis

LOWSET-01 separates three questions that are often conflated: whether a registered event is enriched, whether that enrichment survives independent comparisons, and whether exploiting it costs less than equivalent conventional evaluation. Passing the first two does not imply passing the third.

The comparison was frozen as T0 versus M20 before interpretation. Across eight registered comparisons, the observed direction was favorable in all eight. The aggregate rate ratio was 1.515, meaning the registered witness appeared about 51.5% more often in T0 than M20 under the experiment's counting convention.

The implementation gate is deliberately independent of statistical significance. It charges candidate construction, dispatch, verification and fallback work. Under that accounting, the operational implementation failed; the site therefore preserves the signal while rejecting the mining-advantage claim.

Statistical signal and operational gate

Rate ratio1.515×95% CI [1.270, 1.863]

≠

Operational gateFAILEDThe signal does not imply speedup

FIG / EVIDENCERegistered enrichment and the separate operational gate. The statistical interval is above the null ratio of 1, while the implementation verdict remains failed.

ANALYSIS / EQUATIONS

Mathematical formulation

RR = p(T0) / p(M20) = 1.515

The rate ratio measures relative witness frequency; RR > 1 is enrichment, not a speedup.

CI₉₅(RR) = [1.270, 1.863]

The reported interval excludes 1 for the registered aggregate comparison.

qHolm = 0.00889

The Holm step-down correction controls family-wise error across the registered comparison family.

advantage = useful work / total charged work

Operational relevance requires the full ratio—not witness frequency alone—to beat the frozen baseline gate.

ANALYSIS / STATISTICS

Statistical reading

The sign pattern was 8/8 favorable comparisons. This consistency supports the direction of the registered observation but does not establish portability to other grammars, headers or witness definitions.

The confidence interval and Holm-adjusted q-value address sampling uncertainty and multiplicity. They do not measure implementation overhead, ASIC compatibility or energy efficiency.

A valid replication must preserve the cohort definition and report denominators, not only favorable witnesses. Selective reporting after viewing outcomes would invalidate the registered error control.

ANALYSIS / SCOPE

Validity and scope

Internal validity: hash and witness counters were audited over 68,719,476,736 evaluations.
External validity: limited to the frozen T0/M20 construction and B16 witness definition.
Operational validity: failed. No stock-ASIC or end-to-end speed advantage was demonstrated.

SOURCES / METHOD

Methodological references

Interpretation

The observed signal is scientifically interesting and worth isolating. It is not evidence of faster Bitcoin mining because the end-to-end implementation cost erased the local statistical result.

Limitations

The signal is tied to the registered construction and witness definition.
The operational comparison did not demonstrate a deployable advantage.
No stock-ASIC benchmark has validated the construction.

Reproduction

Reproduction requires the frozen candidate grammar, registered T0/M20 comparison, identical witness definition and independent seeds. A valid reproduction must report both statistical and operational gates.

Artifacts

Audit manifests and a redacted reproduction package are being reviewed for release.

Research-use notice

Experimental content is provided for research and educational use, without warranty. Validate independently before relying on it. Read the full disclaimer.