Sources, methods & corrections

Three kinds of evidence

Network archive

The scheduled collector requests Blockchair Litecoin and Dogecoin statistics every 15 minutes. It retains the provider's source timestamp separately from our retrieval time. Repeated source timestamps are deduplicated; unsuccessful attempts are logged and do not create observations.

Snapshot fields include block height, reported mempool count, rolling 24-hour average transaction fee, estimated hashrate, difficulty and the provider's USD reference price. Dogecoin price and difficulty support reproducible merged-mining calculations; a failed Dogecoin read leaves those fields empty. Hashrate is the provider's estimate, not directly measured hardware output.

Fresh snapshots must have source timestamps no more than 15 minutes old and a reported tip block no more than 30 minutes old. This conservative rule can reject a valid network during a long block gap; it does not imply a network outage. All archive exports use Unix seconds in UTC. Litecoin fee amounts are LTC unless a column explicitly says litoshi; 100,000,000 litoshis equal 1 LTC.

The rolling fee series is not a fee recommendation. Consecutive 24-hour windows overlap. The chart uses hourly means of available snapshots and leaves long gaps open. Its zero baseline is shown; the collection log and CSV expose missing observations.

We request the latest 100 blocks each run and recheck those heights for changed hashes. Block records include first retrieval and last check times. Replacements are counted as observed revisions, not attributed to a cause. Deep reorganizations outside the checked window may be missed. Original block timestamps can predate the archive.

Interval statistics compare adjacent recorded heights only. Negative timestamp differences are counted separately and excluded; zero differences remain. Quantiles use linear interpolation at index (n − 1) × q on the sorted sample. These are miner-set timestamp differences, not first-seen-to-confirmation durations.

Blockchair API documentation · Observations · Run log · Reproduction script

Archive checks and recovery

A separate BlockCypher chain endpoint records a tip height, hash and reported time every 15 minutes. A same-height hash comparison needs a primary block check and secondary timestamp no more than 30 minutes old. Missing overlap, old responses and different hashes are separate results. The two requests are not simultaneous. This check does not validate either provider with our own node and never substitutes another source into the Blockchair time series.

Consecutive unique observations retrieved no more than 30 minutes apart are flagged when height decreases, USD price changes by more than 25%, difficulty by more than 50%, hashrate by more than 100%, or the mempool grows more than tenfold and exceeds 1,000. These are review thresholds, not proof of bad data. Flagged values remain visible. Missing collector runs, snapshots or block checks are highlighted after 35 minutes.

A daily online SQLite backup is restored into a separate temporary database. Integrity, required tables, row counts, latest observations and a SHA-256 checksum are checked before the backup is retained. Fourteen daily copies remain on the same server. This is archive recovery, not a claim about off-site disaster recovery or the article database.

Transaction checker

The form accepts a public 64-character transaction ID. Our server asks Blockchair's Litecoin transaction dashboard and displays the status, a bounded input/output breakdown and a value-flow diagram when the full amounts reconcile. Confirmations are calculated as provider tip height − transaction block height + 1 for a confirmed record.

Virtual size is ceil(weight / 4). The effective fee rate is fee in litoshis / virtual size. Coinbase and identified MWEB-related records receive a separate explanation; their rate is not interpreted as a standard payment fee. A not-found response describes this source, not every node or a wallet's private MWEB state.

Transaction IDs travel in a POST body, not a page URL. The tool does not save search history in the browser and does not load Google Analytics. The server caches the summary and up to 200 public input and 200 public output records for two minutes; expired rows are removed during subsequent lookups. Raw scripts and signatures are discarded. Rate-limit counters use a daily keyed hash of the connecting IP and expire; normal hosting access logs can still record a visit. Requests to the data provider expose the public ID to that provider. No recovery secrets are requested.

Input and output details

Integer litoshi values, row indices and transaction references are checked before display. A complete-value diagram requires every declared input and output, matching provider totals, and input total minus output total equal to the reported fee. Missing, malformed or conflicting detail data never becomes a guessed total. At most 200 individual records per side are returned; larger complete transactions retain their checked totals, with the first four rows drawn separately and remaining values grouped. Tables initially display 20 rows and can reveal more.

Address labels come from the source. Their owners, sender change and customer account credit are not inferred. A matching sum is an arithmetic consistency check, not independent transaction or signature validation. Coinbase and identified MWEB-related records use a separate explanation without an ordinary-payment diagram. Source times, rate limits and the existing no-history/no-analytics policy still apply.

Sampled first-observation confirmation study

The confirmation observer polls our wallet-disabled, pruned Litecoin Core 0.21.5.8 mainnet node every 15 seconds. Initial block download must be complete, validated blocks must equal known headers, at least one peer must be connected, and the current tip header must be no more than two hours old or in the future. The tip must remain unchanged during the read. These conservative conditions may suspend sampling on a healthy network during a long block gap.

Transactions already present at startup or after a gap are excluded. Among newly appearing mempool txids, the first eight hex characters interpreted as an integer must be divisible by 16. This deterministic rule targets a roughly 1/16 sample, not a random survey of senders. Core's default 300 MB mempool cap and its changing minimum fee influence admission and eviction. The observer supports at most 100,000 current mempool entries and 250,000 retained cohort rows; exceeding a resource cap suspends collection and creates an unavailable period.

First observation and inclusion times are collector UTC polling times, not miner header times or sender timestamps. A continuous interval must be positive and at most 30 seconds. Inclusion is matched only when exactly one new block extends the last observed tip. A missed interval, a multiple-block jump, a process restart or a chain change censors open waits. Disappearance without a matched block is recorded without guessing the reason. The same sampled txid is never reenrolled while its record remains in the 90-day working database.

After an interruption or detected reorganization, prior confirmations are withheld for a canonical block-hash recheck. Matching blocks restore the original record; changed blocks are excluded as reorganization outcomes. Rechecks process at most 50 distinct heights per poll. Private MWEB transfers, transactions confirming entirely between polls, baseline entries and transactions unseen by this node do not enter the timed cohort. This is a source of selection bias, especially for short waits.

The window selects records by first observation in the last seven days. Pending and censored observations remain visible in the outcome table and CSV. The histogram, median and 90th percentile condition on the currently confirmed subset only; they are not completion probabilities or network-wide wait estimates. Quantiles use linear interpolation at (n − 1) × q and require at least 20 confirmed records. The CSV is refreshed about once a minute; the page snapshot about every 15 seconds. Database pruning runs hourly and retains 90 days; the public live window remains seven days. No financial or personal sender data is inferred.

Protocol v2 adds immutable daily exports and a fixed 30-calendar-day study beginning at the next UTC midnight after the first valid collecting interval. Gaps remain in the calendar period. Daily raw polls determine coverage; the last snapshot determines cohort outcomes, labelled with its actual export time. Earlier cohort snapshots must not be summed together. Later reorganizations can supersede historical confirmations without rewriting those exports. The archive page links the protocol, files, checksums and report calculation. Synchronization RPC diagnostics may wait 20 seconds; eligible timing reads still must finish within 10 seconds.

Cohort state and summary source · RPC collection source · Official Core release

Local payment CSV reconciliation

Invoice and receipt CSVs are processed in the browser using integer litoshis, validated Litecoin address checksums and explicit output references (txid plus vout). Conflicting duplicate outputs and ambiguous invoice matches are excluded and flagged. Identical duplicates are counted once. Confirmed amount means the imported confirmation count meets the invoice requirement; no network verification is performed. Expiry comparisons use the imported reported receipt timestamp, and unknown times remain unknown. An arithmetic match is not a merchant acceptance decision.

Weekly reports

Reports summarize completed UTC weeks, Monday 00:00 through the next Monday 00:00. A report requires at least one valid observation on all seven days and valid network and block-window reads in at least 90% of the 672 quarter-hour slots. Repeated runs in a slot count once. A completed week below that threshold is withheld; its raw observations remain available. The report shows actual counts and expected 15-minute collection slots, so a week with gaps is visible. Figures are generated from the archived data using the formulas above; they are not manually inferred market commentary.

A frozen report records its generation time. The report freezes its input observations, blocks and run log in the same transaction. Weekly downloads reproduce that frozen input; the rolling archive may later reflect a rechecked block. Histograms include only adjacent, non-negative header-time differences. The longest snapshot gap includes the interval from week start to first observation and from last observation to week end. Hourly mempool values are unweighted sample means; they are not time-weighted estimates. Any material correction to a published report will be explained rather than silently presented as a new measurement.

Wallet recovery experiments

The versioned recovery study separates wallet-file restoration, deterministic address reproduction, balance discovery and signing authority. Core runs on isolated regtest; Electrum-LTC runs through its offline command line; Trust Wallet Core is tested as a WebAssembly engine. An engine test does not establish mobile-app behavior. The cross-implementation bridge imports BIP39-derived account keys, not BIP39 words through Electrum's graphical wizard.

Release 2026-09-27.2 adds same-chain Core history rescans, an old backup omitting a later imported key, and offline Electrum address-range preparation. It does not test pruned-history restoration, a clean mainnet node resync or online Electrum history discovery. Published data identifies exact software versions, fixed fixture positions, timestamps, negative controls and reproduction commands. Address matches and assertion counts are not population recovery rates. A verified message signature is not an accepted network transaction. Public synthetic fixtures must never be used for funds.

Preparation uses AI-assisted coding and drafting together with actual runs and scripted checks. No independent human review is inferred. The editor's profile explains responsibility and related projects.

Frozen node records and declared scenarios

The 2 October package queries our pruned Litecoin Core 0.21.5.8 node. Its 1,001 early headers retain signed timestamp differences, including negatives. A separate check selects only output zero of the sole coinbase transaction in 998 one-transaction early blocks; genesis and two multi-transaction blocks are excluded. Current unspent status cannot identify a lost key or owner.

The recent sample decodes 50 fixed mainnet blocks. It separates coinbase, HogEx and ordinary transparent rows; summed outputs still include change and self-transfers. HogAddr value is selected separately from peg-out outputs. Kernel records do not identify people or reveal hidden transfer amounts. Captures retain boundary block hashes and were rechecked after acquisition.

All raw files, input checksums, calculation source and limitations are in the frozen node package. Recomputing statistics verifies the export’s arithmetic; querying a second node would be a different verification step.

The declared arithmetic package contains 42 Scrypt capacity scenarios, six hypothetical card sale-cost cases, 18 synthetic treasury cases and a payment-report denominator calculation. Synthetic values are not issuer NAV, card quotes, measured power or available attack capacity. Provider-derived shares remain rounded; their denominator and exclusion rules are stated.

The workflow calculation package adds nine synthetic fee budgets and a seven-row inventory ledger. The blank CSV is a recordkeeping aid, not a tax return. Documentation checklists are labelled as source reviews and are not counted as controlled experiments.

Article and catalogue evidence labels share one registry. Controlled wallet tests, source-function tests, download verification, observations and models remain distinct. Dataset metadata describes the same downloadable records visible on the page; it does not represent a ranking score or a promise of search visibility.

Editorial responsibility and dates

Article bylines identify the named author or responsible editor shown on that article. An edited date means the record changed; it does not certify that every statement was independently reviewed. Where a fact-check record exists, the article shows its date and scope separately. We do not invent reviewers or attach expertise claims to an automated check.

Research pages distinguish calculations, public-source observations and actual software tests. A documentation review does not become a hands-on test. Real recovery phrases, private keys and private wallet exports are never appropriate evidence to publish. Reproduction fixtures may contain deliberately public, synthetic test material, prominently marked as unsafe for funds.

Material corrections should state what was wrong, what changed and when. Report an issue through the contact page, including the page URL and the relevant source or calculation. View the public corrections log.

Commercial links and related sites

The wallet comparison introduced here uses direct documentation links without affiliate tracking and does not sell ranking positions. Its order is not a recommendation or a security score. Sponsored material and affiliate links must be disclosed where they appear.

Links to related projects such as Forex Mechanics and Currency Converter are included only when they help with the task and are identified as related projects. A cross-link is not independent corroboration of a claim.