About This Analysis
This page provides a match-level analysis of Blitz games using the
Blitz Impact Model. Each match is reconstructed from individual
player entries and evaluated to show how well the pre-game prediction aligned
with what actually happened in the server.
What We Show
-
Actual Match Outcome – Final score, margin, and an
actual closeness label that indicates whether the game was about even,
slightly tilted, or completely stacked in favour of one team.
-
Model Prediction – Probability-based estimation of which
team should win, summarised as a verdict such as
“Too close to call”, “Advantage Red/Blue”, or “Stacked AF”.
The same column also shows whether the checker was right, a soft upset,
or a hard upset.
-
FRF / Certified Close Games – FRF
(Final Round Finish) games are those that stayed alive until the
final attack (Blue’s last attack in Round 3). These are surfaced as
CERTIFIED CLOSE GAME, and represent games that went the distance.
-
Certified Not Close - Games where, from the scoring alone,
we can tell the losing team had no possible way to win before the final
attack (mercy end). These are
marked as CERTIFIED NOT CLOSE.
-
Certified Stacked - High-margin wins (large score gaps)
where the result is clearly one-sided. These appear as
CERTIFIED STACKED.
Summary Table
The summary above the list aggregates statistics across all recorded games,
including:
overall model accuracy, counts of soft and hard upsets, frequency of
CERTIFIED CLOSE GAME (FRF) matches, frequency of
CERTIFIED NOT CLOSE and CERTIFIED STACKED
matches, win distribution between Red and Blue, average score margin, and
average model confidence.
Game List
Each row in the table represents a single match and includes:
the actual score and margin with a closeness label,
a combined Predicted vs Actual column (model verdict,
actual winner + margin, and checker-right / upset badge),
a Summary column showing whether the game was certified close,
not close, or stacked, and the player list.
The checker link lets you open a deeper breakdown of player balance for that match.
Legend
Summary
CERTIFIED CLOSE GAME
The match stayed alive until map 6 and was a Final Round Finish (FRF).
CERTIFIED NOT CLOSE
The losing team had no possible way to win before the final attack. The match ended early.
CERTIFIED STACKED
High-margin win (margin > 5). A decisive or overwhelming mismatch. The margin tells us this was one-sided.
Checker right ✓
Blitz Impact Model predicted the correct winner.
Soft upset
Model prediction was wrong, but the match was close or FRF-driven.
Hard upset ✖
Model strongly favoured the other team, but the opposite side won.
Actual Game Closeness (derived from outcome)
About even
Balanced match; outcome decided by small moments.
Edged it
Winner had a very small advantage; still a close game.
Slightly stronger
Winner showed modest superiority but the match was still competitive.
Too strong
Noticeable gap; the losing side struggled to keep up.
Clearly stronger
Winner had clearly superior performance; not a close match.
Stacked
One-sided game; teams were significantly mismatched.
Stacked AF
Extremely one-sided match; overwhelming imbalance.
FFS! Stacked AF
Comically one-sided; the match was effectively unwinnable.
Predicted (based on Blitz Impact Model verdict)
Too close to call
Model reads the match as effectively 50/50.
Slight edge
Model sees a mild advantage for one team.
Advantage
Model favours one team with moderate confidence.
Likely too strong
Model expects one team to be significantly stronger.
Clearly stronger
High probability of one team outperforming the other.
Stacked AF
Model sees overwhelming imbalance; extreme mismatch.