Go to the Upload page. Zip the contents of your Steam userdata/<id>/2868840/ folder (or just the history subfolder) and upload it. Non-.run files are ignored automatically.
Yes. Duplicate runs are detected by content hash and skipped. You can re-upload anytime to add new runs.
Your username is deterministically generated from your earliest run file. The same runs always produce the same username. You can set a display name on your player page after uploading.
Your username is based on your earliest run file. If one device is missing old runs (e.g. fresh Steam install without cloud sync), it will generate a different username. Contact us to merge accounts.
Solo runs only (no multiplayer). Floor-1 abandons (quitting/restarting at Neow) are excluded. Real deaths with the abandon flag are kept.
Runs with non-standard build_id (modded game versions) are rejected on upload.
Yes. There is currently no way to detect cherry-picked uploads. The run files don't contain sequence numbers, so missing runs can't be identified. This is a community stats site, not a competitive ladder.
The leaderboard ranks players by overall win rate across all characters and ascension levels, with a minimum of 5 solo runs. Win rate includes all non-abandoned solo runs regardless of ascension — an A0 win and an A10 win count equally. This is a known limitation: a simple win rate doesn't account for difficulty differences between ascension levels. We plan to explore ascension-adjusted rankings once we have enough data to reliably estimate difficulty at each level.
The survival chart excludes runs with the Golden Compass relic, which changes the map layout from 48 to 50 floors. This keeps the chart comparable across runs, but means some runs are excluded from the chart but included in the table stats.
Run time comes from the game's run_time field. We show the median (not mean) across runs to avoid outliers from AFK sessions.
Each combat card reward (pick 1 of 3 or 4 cards) is treated as a multi-way match using a Plackett-Luce model (multinomial logit). The picked card "wins" against the others. Offers where all cards were skipped are excluded. Ratings are fitted per character by gradient ascent on log-likelihood, centered on 1500 = average card. Shop offers are excluded since they use a different selection mechanism (buying with gold).
Higher Elo = players pick this card more often when offered. A card at 1500 is average. Above 1500 means preferred, below means avoided. The rating accounts for which other cards were in the offer — a card that beats strong competition gets a higher rating than one that only beats weak cards.
Card descriptions are scraped from sts2.untapped.gg. Some icon-based resources are shown as symbols: ⚡ = energy, 🔷 = stars (Regent). A few cards with complex rendering may have minor formatting issues.
Built by dark-void. Feedback and suggestions welcome.