How Cut Watch works
Cut Watch estimates how players and final records may fare by simulating 50,000 possible completions of the most recent tournament state it has received from Cobra. It samples possibilities rather than examining every outcome, so its percentages contain sampling error and rare paths may be missed.
Every unresolved game uses the same baseline probabilities: 54.2% Corp win, 42.5% Runner win, and 3.3% draw. These probabilities do not account for individual players, decks, or matchups. Intentional draws use a separate heuristic when a draw appears useful to both players.
Published pairings and results are used as received from Cobra. Unknown future pairings are approximated. Forecasts are estimates, not guarantees.
Technical model
Simulated worlds
Each world starts from the latest Cobra snapshot, resolves published but unfinished games, generates the remaining Swiss rounds, and ranks the resulting standings using Cobra-compatible points-per-game, strength of schedule, and extended strength of schedule calculations.
Future pairings
Known pairings are used as published. For unknown rounds, Cut Watch approximates Cobra’s priorities for score proximity, rematches, side balance, and byes. Earlier future rounds use a faster local search; the final two use a broader matching search. Generated opponents may therefore differ from those Cobra eventually selects.
The faster approximation permits more worlds within the refresh-time budget. Cut Watch therefore prefers an estimate that converges closely on a slightly approximate pairing model to one that remains farther from Cobra’s exact model because too few worlds were sampled. This reduces sampling noise but does not remove pairing-model bias.
Intentional draws
Intentional draws are modeled with an adaptive heuristic, not from knowledge of players’ intentions. A sampled draw that leaves a player in the cut makes later offers more likely; an intentional draw that leaves them outside the cut removes that offer state. Each player’s decision is sampled separately, and an ID occurs only when both decisions favor it.
The offer probability is capped at 90% and can fall towards 55% when winning materially improves a player’s chance of finishing in the top half of the cut. Adaptive IDs are considered only for a published current pairing and the final simulated round.
Metrics
Each world contributes observations for players and final-record groups. “In Cut” is the average number of cut places occupied by a record. “Safety” is the share of observations at that record that made the cut. “Side Choice” is the share that finished in the top half of the cut. Rare record paths are omitted because percentages based on few observations are unstable.
Updates and freshness
Source checks
Opening an event shows the most recent complete forecast, which may be older than Cobra’s current state. While the page remains visible, Cut Watch checks Cobra roughly every ten minutes early in a round, three minutes when half the games are reported, and one minute when all games are reported.
Replacement forecasts
When new data is found, the previous complete forecast remains visible and is marked as updating until its replacement is ready. Partial forecasts are not shown. Hidden or inactive pages eventually stop checking, and completed events do not refresh automatically.
Reliance
A delayed source check, calculation, or browser refresh can leave an older forecast visible. The freshness bar shows the expected update window; it does not confirm that the displayed data matches Cobra’s current state. Check the forecast’s freshness and updating state before relying on it.