Chunk Availability Matrix (24 Chunks x Swarm Peers)
Dark = Missing | Bright Cyan/Green = Possessed
Does losing every seeder necessarily lose every piece?
Incomplete peers may collectively hold the entire file even when nobody has a complete copy. Inspect the minimum number of copies of each chunk, rather than relying only on seeder count or this page’s authored risk label.
Transfers grant whole Boolean chunks, without byte duration, file data or hash verification. Churn and optimistic allocations are stochastic. The health score uses fixed penalties and does not predict reliability or real download speed.
If A holds chunks 0–11 and B holds chunks 12–23, every one of the 24 chunks has one copy although neither peer is a seeder. Exchange may still be possible through permitted uploads. This model nevertheless labels zero-seeder states CRITICAL by rule; that label is not a measured collapse probability.
BEP 3 uses bitfields and have messages to advertise pieces, with hash verification before announcing a completed piece. A complete peer and collective piece availability answer different questions. Export now includes each peer’s chunk flags so the modeled availability can be checked directly. BEP 3: BitTorrent Protocol Specification
The score starts at 100, subtracts 40 without seeders, subtracts 40 for an absent chunk or 20 when the rarest chunk has one copy. Risk labels apply fixed thresholds. Those penalties are assumptions, not estimates fitted to measured swarms. Rarity is refreshed from current flags after transfers and churn.
A receiver records which uploader supplied a chunk; contribution ranking uses those cumulative records. The upload-slot parameter limits choices rather than representing MB/s. Pause before changing conditions and inspect the matrix; reset randomizes initial leecher pieces. Churn Test may leave a seeder because retention is probabilistic.
Actual piece advertisement and integrity verification; not these health penalties, risk thresholds or churn probabilities.