endereye generates per-player qualification probabilities during Last Chance Qualifiers and Midseason Showdowns, updating after each seed to answer the question: who is actually in trouble right now?
Visit https://lcqtracker.vercel.app, where it updates in real time for the currently active event!
Each event state is simulated 10,000 times. For each simulation, players are ranked per seed based on a power score derived from their Elo, win rate, best time, and average completion time. Players with a large gap between their best and average time are more likely to DNF in any given seed. Eliminations are applied at the scheduled cut points, and the top 4 finishers qualify.
The model scores each player's expected performance per seed using:
- Elo — a compressed version of their season rating, weighted lightly to avoid over-relying on ladder performance
- Win rate — scaled by number of matches played, so low-volume players don't get outsized credit
- Best time — how fast they can go on a good run, relative to the field
- Average time — how consistently they finish, weighted more heavily in the first half of the event
Beyond probabilities, the model provides exact answers where possible:
- Clinch score: the minimum placement a player needs this seed to guarantee survival at the next cut, under worst-case assumptions for all other players
- Safe: whether a player is mathematically guaranteed to survive the next cut regardless of this seed's result
- Can still win: whether a top-4 finish is still mathematically possible
These are computed deterministically, not from simulation, so they carry no sampling error.
The model has been backtested across Seasons 7–10 for both LCQ and MSS:
- AUC 0.886 vs. baseline 0.762 — statistically significant lift over raw Elo (p < 0.001, bootstrap 1000 iterations)
- 0 clinch violations — no clinch guarantee was incorrect
- 0 safe violations — no player flagged as safe was subsequently eliminated
Calibration is slightly overconfident in the 10–80% survival range (actual rates run ~3–6% below predicted), likely due to underestimating DNF variance in the competitive field.