Speed Test Credibility & RM Grab Index
ReleaseMatch runs libtorrent segment download tests on the Hero Recommended torrent only. This page documents how RM Grab Index and speed credibility labels are derived—so you can interpret scores on episode and movie pages.
RM Grab Index
A composite score from 0 to 100 for the Recommended release on that page. Four dimensions are scored 0–100 each, then combined with fixed weights:
| Dimension | Weight | Input | Scoring logic | IG ref |
|---|---|---|---|---|
| Speed | 35% | Phase 2 avg & max KiB/s | Log curve on avg speed (≈15 at 1 KB/s → ≈100 at 5 MB/s); +5 if peak ≥ 2× avg | S-06 |
| Reachability | 25% | Phase 1 peers_total, test status | ≥10 → High (100) · 3–9 → Medium (65) · 1–2 → Low (35) · 0 / timeout / error → Unreachable (0) | A-01 |
| Connect rate | 20% | peers_reachable / peers_total | Round(connected ÷ observed × 100), capped at 100 | A-02 |
| Freshness | 20% | Hours since last test + validity level | ≤24h Fresh · ≤48h Valid · ≤72h Stale · >72h Older — see Speed credibility | A-03 |
Formula: Grab = round(speed×0.35 + reach×0.25 + connect×0.20 + freshness×0.20), clamped 0–100.
Overall tiers
| Score | Tier label | Meaning |
|---|---|---|
| ≥ 90 | Excellent | Strong measured download experience |
| 75 – 89 | Great | Reliable for most users |
| 60 – 74 | Good | Usable with some caveats |
| 40 – 59 | Fair | Mixed signals—check breakdown |
| 20 – 39 | Weak | Limited swarm or stale data |
| < 20 | Poor | Not recommended for grab decisions |
Speed test credibility (A-03)
Every test timestamp is classified by age. This drives the freshness badge, validity label, and the freshness sub-score inside Grab Index. Older than 72h is labeled “Older”, not “Expired” — we do not assert the swarm is dead.
| Status | Age window | Validity | Freshness sub-score | Page guidance |
|---|---|---|---|---|
| Fresh | ≤ 24 h | High | 100 | Safe to cite as live measured evidence (S-07) |
| Valid | 24 – 48 h | Medium | 78 (−5 if validity Medium) | Still useful; re-test when convenient |
| Stale | 48 – 72 h | Low | 42 (−15 if validity Low) | Peers/speed may have changed |
| Older | > 72 h | Uncertain | 12 | Swarm may have shifted — re-test before relying (not marked expired) |
| Unknown | No timestamp | Unknown | 0 | No libtorrent test on record |
Speed test metrics
Shown in the collapsible Speed evidence panel and measured metric cards:
| Metric | Definition | How it is measured |
|---|---|---|
| Test region | Egress node region of the libtorrent run (e.g. Japan · Osaka) | slot_speed_summary.test_region from RM_SPEEDTEST_REGION |
| Tested at | UTC time of last successful slot test | Written to speedtest result / summary timestamps |
| Avg speed | Mean KiB/s over 256 KB segment samples | Phase 2 libtorrent download (strategy A2) |
| Max speed | Peak KiB/s in the same run | Phase 2; spread vs avg shown when available |
| Peers observed | Distinct peers seen in tracker/DHT | Phase 1 peer sampling |
| Peers connected | Peers that completed handshake | Phase 1; pair shown as connected / observed |
| Connect rate | connected ÷ observed | Percentage; feeds Grab connect dimension |
| Latency | Round-trip style delay summary | Derived from peer handshake timing when available |
Peer reachability (A-01)
Reachability is not a separate crawl—it is derived from Phase 1 peer counts and test status:
| peers_total | Reachability | Sub-score | Typical interpretation |
|---|---|---|---|
| ≥ 10 | High | 100 | Healthy swarm for partial download test |
| 3 – 9 | Medium | 65 | Enough peers to attempt measurement |
| 1 – 2 | Low | 35 | Fragile swarm—speed may be unstable |
| 0 or timeout/error | Unreachable | 0 | No usable peers during test |
Method note on each page states the libtorrent segment strategy (e.g. 256 KB, A2). Indexer seed counts (B-02) are shown separately and do not replace measured peers (A-02).