RadarFooty BackAccountRO

Cascade

What a cascade of conditions really covers: the model's estimate for the upcoming matches and what happened in the rounds played. It shows what happened, not what will happen. Match filters →

How a cascade works

A cascade puts the same condition on several matches, in kick-off order, with one variant for each place where the condition might NOT happen. Exactly one variant happens, or none: if the first “no” comes before the match where the cover starts, none does.

Example: 4 matches, cover starting at match 2

  1. all 4 happen
  2. first “no” at match 2 (match 1: yes)
  3. first “no” at match 3 (1 and 2: yes)
  4. first “no” at match 4 (1–3: yes)
  5. none: the first “no” is already at match 1

Build a cascade

Pick the condition, how many matches, from which match the cover starts and how the matches are chosen. For each variant you see how likely it is by the model's (calibrated) estimate, and at the top how likely it is that none happens. A cascade does not change any match's probability, it only splits it.

There are no upcoming matches with a model estimate right now (usually during a break).

The Elo grid cascade

A cascade built from the Elo grid: for the chosen condition we take the best positions of the round, by their history in the rounds before, and put them in kick-off order. The first 4 are the base and must all come true; from match 5 on, the cascade shows where the first “no” comes. The table below shows, for each played round, what happened — not what comes next.

Round 23 Oct – 26 Oct · 2+ goals

No estimated chance: not all 4 base matches have the model's frozen estimate.

  1. 1Cagliari – Bologna0-2P41 · 74% · base✓
  2. 2Mainz – Bremen2-3P21 · 79% · base✓
  3. 3SC Paderborn – HSV1-0P29 · base✗
  4. 4Augsburg – Union Berlin1-1P22 · 76% · base✓
  5. 5Arsenal – Everton2-0P2 · 72%✓
  6. 6Inter – FiorentinaP1 · 83%·
  7. 7Hull City – BrentfordP43·
  8. 8Marseille – Le HavreP3 · 79%·
  9. 9Sunderland – Leeds UnitedP45 · 73%·
  10. 10Freiburg – BayernP7 · 86%·
  11. 11Real Sociedad – LevanteP27 · 84%·
  12. 12Barça – Real MadridP19 · 84%·

In progress.

Every round

Covered in 3 of 5 decided rounds.

  • 23 Oct – 26 Oct✓✓✗✓|✓·······in progress
  • 20 Oct – 21 Oct—only 1 match
  • 16 Oct – 19 Oct✗✓✓✓|✓✓✓✓✓✓✓✓base failed
  • 9 Oct – 12 Oct·✓✓✓|✓✓✓✗✓✓✓✓not settleable
  • 18 Sep – 20 Sep✓✓✓✗|✓✗✓✓✓✗✓✓base failed
  • 15 Sep – 17 Sep—only 8 matches
  • 11 Sep – 14 Sep✓✓✓✓|✓✓✓✓✓✗✗✓covered at 10
  • 4 Sep – 7 Sep✓✓✓✓|✗✗✗✓✓✓✓✓covered at 5
  • 1 Sep – 3 Sep—only 2 matches
  • 28 Aug – 31 Aug✓✓✓✓|✓✓✓✗✗✓✗✓covered at 8
  • 25 Aug – 27 Aug—only 4 matches

Tracked cascades ❄

For each round, the matches and the model's estimates are frozen 1–2 hours before the first match (❄) and do not change afterwards. After each match you see ✓ if the condition happened and ✗ if it did not. The first ✗ decides the cascade: if it comes at or after the match where the cover starts, its variant happened; if it comes before, none happened. For example, with cover starting at match 4, a ✗ at match 2 means “none”. The next round's cascade then moves to the top.

Rounds decided: 2 · none: 1 of 2 (50%; 95% interval: 9.5%–91%) · average “none” estimate, at freezing: 79%

  1. 23–26 OctFirst “no” at match 4 ⇒ its variant happened (covered)✓✓✓✗✗········“none” estimate: 80%
  2. 16–19 OctFirst “no” at match 1, before match 4 ⇒ no variant happened✗✓✗✓✗✓✗✓✓✗✓✓✓“none” estimate: 79%

What the record shows

The same cascade applied to the rounds played in the last two seasons, with estimates rebuilt (↺) only from the matches before each match day: how often no variant happened, next to what the model had estimated. Rounds with too few matches are left out and counted separately.

80 rounds with at least 13 matches (32 with fewer, left out)

None happened: 63 of 80 (79%; 95% interval: 69%–86%)The model's estimate, before the matches: 79%2024-25: 29 of 39 · 2025-26 (clean test): 34 of 41

What happened

  1. First “no” at match 45
  2. First “no” at match 53
  3. First “no” at match 62
  4. First “no” at match 72
  5. First “no” at match 82
  6. First “no” at match 92
  7. First “no” at match 101
  8. None happens63

How to read it

How to read the numbers on this page without giving them a meaning they do not have.

Updated: 25 Oct, 07:20 UTC