# S16 final pre-event freeze · 2026-10-15

[简体中文](final_pre_event_2026_10_15.zh_cn.md) · [Results](https://lol.frankblogs.com/) · [Frozen release](https://github.com/EdmondXing2008/lol-bp-prediction/releases/tag/m1-m4-final-preworlds-2026-10-15)

Patch26.20. **Input cutoff: 2026-10-15T15:18:24.838910+00:00, UTC**. Whole-tournament forecasts. New information after the event starts must not overwrite this delivery.

## Updated inputs

- Added all5 completed games from WE3–2BRO, DCGI semifinal, patch26.18. Knockout supplements:28games/6series;23have complete draft order. M1 S16 same-regime pre-event history:2,545games.
- October15 completed results, actual-role lineups, players and full ban sets checked against reports and reproduced official broadcast/stat images. GoL still showed a preview at collection: **exact action order and GoL fullstats for5games are missing**. Display order is not used as played-role order or invented action labels.
- Re-fetched the previous23games on GoL: order, bans, roles, players, patch and winner unchanged. Four October14games still lack independent full-order verification, with the secondary game3ban-list conflict retained. Nine games await complete independent BP verification; this is not a claim of full cross-source verification.
- Practice:103event-listed starters/substitutes,392known non-China accounts attempted;69new records,49qualifying. Capture-window total3,232; model-window total**3,227**, covering63/103players. Solo/Duo only, evidenced account identity and actual official role required. Original observation times preserved.50account reads incomplete/unresolved; Taiwan public registry incomplete; one dated account ownership dispute retained. Missing remains unknown.
- KR Master+ patches26.18/19/20 refreshed.26.20provider meta.match_count511,207→550,401(+39,194), **not verified independent matches**. Current patch incomplete. Matchup reference:254hero-role endpoints,3,176directional records,0failed endpoints; overlapping directions are not summed.

## Models and freeze

M1/M3/M4 replay existing S15-selected coefficients. M2 updates current fits with fixed previously selected settings, preserving historical evaluation. All current modules share the M1cutoff and checked source hashes; neutral drafts updated too. Classic/Full Fearless opportunities stay separate; blind picks excluded. Unknown order supplies no timing labels.

Current desktop/mobile UI retained, including chronological picks until roles are confirmed. Separate frozen payloads, SHA-256manifest and immutable Release support later comparison. S16Worlds outcomes are not used. Riot lists the first match for October15,11AM PDT/18:00UTC: [official primer](https://lolesports.com/en-GB/news/worlds-2026-primer).

## Largest changes since the previous published forecast

Sorted by absolute whole-tournament per-game pick+ban change, not BO5-at-least-once probability.

- Qiyana: 79.14% → 82.88% (+3.73 pp)
- Smolder: 1.84% → 5.16% (+3.31 pp)
- Alistar: 18.66% → 21.23% (+2.56 pp)
- Olaf: 60.55% → 58.16% (-2.39 pp)
- Pantheon: 40.92% → 39.57% (-1.35 pp)
- Kai'Sa: 6.44% → 7.75% (+1.31 pp)
- Ornn: 9.93% → 11.19% (+1.26 pp)
- Yunara: 10.22% → 9.02% (-1.20 pp)
- Annie: 21.36% → 22.54% (+1.18 pp)
- Nautilus: 31.79% → 30.64% (-1.16 pp)
- Camille: 21.03% → 19.89% (-1.14 pp)
- Twisted Fate: 20.24% → 19.15% (-1.09 pp)
- Malphite: 1.07% → 2.16% (+1.09 pp)
- Wukong: 24.00% → 25.05% (+1.05 pp)
- Seraphine: 26.20% → 27.23% (+1.03 pp)

## Reproduce / view

Check out the matching Release tag, verify and extract all frozen-data archives at repository root, then run `python tools/seal_pre_event.py --verify`. To view offline, extract complete_results.zip and open START_HERE.html. [Refresh audit](../s16_preparation/sources/refresh_2026_10_15/refresh_summary.json) · [Seal](../data/development/pre_event_freeze.json) · [October15 final report](https://news.zhibo8.com/game/2026-10-15/match2310212vgnative.htm).