BIG IF TRUEAll issuesCode & data on GitHub

#002 T1: the Korean esports team that transforms in October

Elo ratings for every professional League of Legends team since 2011, every Worlds from 2014 to 2026 simulated in its real format, a test of T1's Worlds-only over-performance, and this year's teams replayed under seven tournament formats. Also in Italian: T1: la squadra coreana di esport che a ottobre si trasforma.

Read the articleShort versionLeggi in italianoSubscribe on Substack

Figures

Click an image for full resolution.

BIG. The 19 teams qualified for Worlds 2026, by Elo rating on 10 October 2026
BIG. The 19 teams qualified for Worlds 2026, by Elo rating on 10 October 2026. Script: code/bit/fig_big.py.
IF. Each champion's pre-tournament chance, expected against actual wins, and T1's wins above expectation
IF. Each champion's pre-tournament chance, expected against actual wins, and T1's wins above expectation. Script: code/bit/fig_if.py.
The predictions. Each team's chance of reaching each round of Worlds 2026
The predictions. Each team's chance of reaching each round of Worlds 2026, from 100,000 simulated tournaments. Script: code/bit/fig_odds.py.
TRUE. Title chances of the four leading contenders under seven formats
TRUE. Title chances of the four leading contenders under seven formats. Script: code/bit/fig_true.py.

Check the article

pip install -r requirements.txt
make check        # re-derives 46 printed numbers from data/, PASS/FAIL each, in both editions
make figures      # redraws every figure, English and Italian

The derived data are in the repository; make figures needs the raw Leaguepedia snapshot in raw/ (from the Zenodo record, or make data, which fetches today's version). On the shipped data all 46 numbers reproduce, and the figure scripts redraw the published figures (titles and margins can shift by a few pixels between matplotlib versions). make all reruns the whole analysis in a few hours; see the README.

Data sources

DatasetProvider
Professional League of Legends results, 2011–2026 (snapshot of 27 September 2026)Leaguepedia contributors, CC BY-SA 3.0, via the Hugging Face dataset gptilt/lol-esports-matches
Six series results, 27 September to 10 October 2026news reports (Sheep Esports, Dot Esports, Terra, Strafe, Pichau Arena)
National-team football resultsmartj42/international_results on GitHub, CC0
Global Power Rankings; lolproelo.com ratingsRiot Games; lolproelo.com. Not redistributed (no licence found): values blanked, scripts recollect them

What is checked, and what is not

All 46 checks run from the shared data. Not checkable from the package: the comparisons with Riot's rankings and lolproelo.com (their values are not redistributed), five files produced by one-off commands during the analysis, and a byte-identical rebuild of the games table. The README lists them.

How to cite

Cite the archive, not the post. One Zenodo record holds this issue's code and data together: Gallotti, R. (2026). BIG IF TRUE #002: T1: the Korean esports team that transforms in October [Code and data] (v1.0.0). Zenodo. DOI: 10.5281/zenodo.23326955.

Cards on the table

This analysis was written and run by Claude Opus 5.5 under the author's direction, not re-read line by line and not peer-reviewed. The bug log records the errors found and fixed, and KEY_NUMBERS.md lists every number in the article. Found a difference? Open an issue.

Related peer-reviewed work: De Bois et al., IEEE Transactions on Games (2025); González-Díaz, Gossner & Rogers, Journal of Economic Behavior & Organization (2012); Sziklai, Biró & Csató, Computers & Operations Research (2022). Full references in the article.

Ratings summarise results: they know nothing of rosters, injuries, patches or the draft, and games within a series are treated as independent.