Team tendency rates
The strongest support market was not a separate prediction engine. It protected the exact-score cloud because every disclosed score stayed under 3.5.
Project Turtle / settled four-score score lab / July 4, 2026
Colombia vs Ghana
This file is now settled. Turtle locked four exact-score claims before kickoff; Colombia won 1-0, so the score card hit through the first score.
ResultColombia 1-0
Four-score cardHit
Official record3/5
Containment4/5
Evidence figures / settled postgame analysis
The 1-0 hit came from the low-event control branch.
Colombia-Ghana is Turtle's first clean settled example of the four-score workflow. The four-score card kept the exact-cell leader 1-0, two Colombia-control extensions, and the 1-1 draw branch. The match landed on the leader.
Locked score cloud
1-0
leader / landed
2-0
control extension
2-1
Ghana-score branch
1-1
fourth-score
The four-score card hit through the first score. That distinction matters: four-score accuracy and four-score accuracy stay separate.
Three public exact-score claims.
Fourth score in the same C4 card.
The locked leader landed and Under 3.5 support settled correct.
Pregame / postgame analysis
The pregame analysis and postgame analysis stay in one match file.
Pregame analysisFour-score card: 1-0, 2-0, 2-1, 1-1.
Turtle locked the exact-cell leader 1-0, kept the Colombia-control branches 2-0 and 2-1, and preserved 1-1 as score 4.
The public rule was fixed before kickoff: four-score accuracy is scored against all four exact-score claims.
The reason 1-0 stayed in the four-score card was mechanical: it was both the low-event exact-cell leader and a branch-coherent Colombia-control score. Turtle did not need to choose between the leader and the cushion family; the final card preserved 1-0 first, then carried 2-0 and 2-1 as extensions.
Postgame analysisFinal score: Colombia 1-0 Ghana.
The exact-cell leader landed. Four-score card is a hit; containment is also a hit because the score cloud already contained the final score.
The carry-forward lesson is to protect a branch-coherent exact-cell leader when the match profile is low-event. Under 3.5 worked because it supported every score in the cloud.
The theorem/card evidence is therefore aligned with the result: the selected score was not a late story, it was the first disclosed score score before the match settled.
Locked Turtle four-score analyst card
1-0, 2-0, 2-1, 1-1.
Notation is COL-GHA. All four scores are the Turtle prediction card. Score 4 is a normal C4 member, not a smaller side bucket.
The lock is visible so the theorem cannot move after kickoff.
$10 on each of the four score predictions.
Insurance/support is separate and only used after the score cloud is locked.
Low-event Colombia branch: a single Colombia goal plus match control beats the draw/transition alternatives.
Fails if Ghana's transition branch lands or Colombia open the match into a cushion.Colombia control branch: the model asks for a cushion when Colombia attack meets Ghana concession risk.
Fails if Ghana score once or Colombia settle for control after the first goal.2-1 is selected mechanically by branch mass, exact-cell probability, and historical bucket support.
Fails if the selected branch is wrong.Score 4 branch. Under the July 6 protocol it is part of the four-score card.
Fails normally; it exists to preserve the draw-danger route as part of the same card.Settlement analysis / July 4, 2026
Final: Colombia 1-0 Ghana. Turtle C4 hit.
This section is postgame. It settles the locked card without changing the theorem above: the official score card was 1-0, 2-0, 2-1, and the final was 1-0.
Final score matched 1-0, the exact-cell leader.
score 4 did not need to settle because C4 landed.
Under 3.5 protected every Turtle score in the cloud.
| Metric | Value | Analysis note |
|---|---|---|
| regulation_result | 1-0 | Colombia 1-0 Ghana after regulation. Jhon Arias scored the only goal in the 14th minute. |
| four_score_card | 1-0, 2-0, 2-1 | Four-score accuracy is scored only against the three score exact-score claims. |
| score_4 | 1-1 | Score 4 is part of the four-score card. |
| four_score_settlement | HIT | The final score 1-0 was the exact-cell leader and score-card score #1. |
| four_score_settlement | HIT | Four-score card hit because the final score was in C4. |
| support_market_call | Under 3.5 / Over 3.5 NO hit | Live support call favored Over 3.5 NO / Under 3.5 at 0.67; it settled correct because the final had one total goal. |
| model_lesson_exact_cell | validated | R2 protected the eligible exact-cell leader 1-0; this is the branch that settled. |
| model_lesson_cushion | overstated but controlled | 2-0 and 2-1 remained live Colombia-control branches, but the match did not produce the second Colombia goal. |
| model_lesson_totals | low-total thesis validated | The exact cloud stayed under 3.5; support-market selection matched the score thesis. |
| match_stats | COL 20 shots, GHA 8 shots | Guardian live report recorded Colombia 20 shots to Ghana 8, with shots on target 2-3. |
| public_record | four-score ledger 4/5 | Public ledger after settlement uses C4 as the normal card. Small sample; not proof of edge. |
| postgame_source | https://www.theguardian.com/football/live/2026/jul/04/colombia-v-ghana-world-cup-2026-last-32-live | Used only for settlement; not used in pregame prediction generation. |
Compact score orbital
Raw Poisson grid before branch proof.
Support-market layer
Totals and BTTS only orbit the score card.
Support markets are not the model. The live support call favored Under 3.5 / Over 3.5 NO because every Turtle score in the locked cloud stayed below four total goals; that call settled correct.
| Market | Side | Model | Fair | Max entry | Stake | Role | Decision |
|---|---|---|---|---|---|---|---|
| Over 1.5 goals | YES | 61.2% | 0.61 | 0.55 | 0-5 | Supports 1-1, 2-0, 2-1 type clouds; weak if card is mostly 1-0/0-0. | consider only if price is below max_entry_price |
| Over 2.5 goals | YES | 34.1% | 0.34 | 0.28 | 0-5 | Only protects if the exact card needs three goals or more. | not a default support unless score cloud has 2+ high-total scores |
| Over 2.5 goals | NO | 65.9% | 0.66 | 0.60 | 0-5 | Supports 1-0, 1-1, 2-0 type clouds; dangerous if score 4 is 2-2/3-1. | best support candidate if the C4 stays low-total |
| Both teams to score | YES | 37.1% | 0.37 | 0.31 | 0-5 | Only protects if Ghana goal appears in two or more score scores. | secondary; exact-score cloud comes first |
Model math
The expected-goals construction.
blended Colombia attack
blended Colombia defensive weakness
blended Ghana attack
blended Ghana defensive weakness
final expected Colombia goals
final expected Ghana goals
Formula provenance. Turtle uses the standard independent Poisson score model after constructing lambdas from the match ledger. The lab-specific part is the branch-aware selection theorem.
Blend attack
A_i = 0.40 GF50_i + 0.25 GF10_i + 0.35 GFwc_i
Blend concession
D_i = 0.40 GA50_i + 0.25 GA10_i + 0.35 GAwc_i
Raw lambdas
lambda_raw_COL = (A_COL + D_GHA) / 2
Context gates
lambda_i = lambda_raw_i x pace x pressure/transition x block_drag
Score cells
P(COL=a, GHA=g) = Pois(a; lambda_COL) x Pois(g; lambda_GHA)
Data analysis
Rows Turtle is actually using.
The raw last-50 ledgers are preserved for proofing. The visible analysis is reduced to features that actually enter the model.
| Team | Rows | W-D-L | GF/M | GA/M | Scored | Conceded | Clean sheets | BTTS | O2.5 |
|---|---|---|---|---|---|---|---|---|---|
| Colombia | 50 | 31-12-7 | 1.860 | 0.800 | 41/50 | 26/50 | 24/50 | 23/50 | 27/50 |
| Ghana | 50 | 16-14-20 | 1.220 | 1.320 | 34/50 | 33/50 | 17/50 | 23/50 | 24/50 |
Sample split
Last-50, last-10, and current World Cup rates
| Team | Sample | W-D-L | GF/M | GA/M | Scored | Conceded | BTTS | O1.5 | O2.5 | O3.5 |
|---|---|---|---|---|---|---|---|---|---|---|
| Colombia | last_50 | 31-12-7 | 1.860 | 0.800 | 82.0% | 52.0% | 46.0% | 66.0% | 54.0% | 32.0% |
| Colombia | last_10 | 6-2-2 | 1.600 | 0.800 | 80.0% | 50.0% | 50.0% | 70.0% | 60.0% | 30.0% |
| Colombia | world_cup_2026 | 2-1-0 | 1.333 | 0.333 | 66.7% | 33.3% | 33.3% | 33.3% | 33.3% | 33.3% |
| Ghana | last_50 | 16-14-20 | 1.220 | 1.320 | 68.0% | 66.0% | 46.0% | 70.0% | 48.0% | 28.0% |
| Ghana | last_10 | 2-2-6 | 0.600 | 1.500 | 60.0% | 70.0% | 40.0% | 60.0% | 30.0% | 10.0% |
| Ghana | world_cup_2026 | 0-0-0 | 0.000 | 0.000 | 0.0% | 0.0% | 0.0% | 0.0% | 0.0% | 0.0% |
Result-bucket math
Exact scores depend on the branch.
Avg GF 2.548; Avg GA 0.581
Modes: 1-0 (8/31); 2-1 (5/31); 3-0 (4/31); 3-2 (3/31); 2-0 (3/31)Avg GF 0.750; Avg GA 0.750
Modes: 0-0 (6/12); 2-2 (3/12); 1-1 (3/12)Avg GF 0.714; Avg GA 1.857
Modes: 0-1 (3/7); 1-2 (2/7); 2-3 (1/7); 1-3 (1/7)Avg GF 2.438; Avg GA 0.500
Modes: 1-0 (6/16); 2-1 (2/16); 5-0 (2/16); 2-0 (1/16); 3-2 (1/16)Avg GF 0.786; Avg GA 0.786
Modes: 0-0 (6/14); 1-1 (5/14); 2-2 (3/14)Avg GF 0.550; Avg GA 2.350
Modes: 1-2 (6/20); 0-2 (5/20); 0-1 (3/20); 1-4 (1/20); 0-3 (1/20)Mathematics / Turtle Pro
Definitions, data, lemmas, theorem.
This is the formal selection proof. The exact card is not a sentence of confidence; it is the result of candidate universe, probability kernel, branch rules, and exhaustion.
Candidate table S10
| Rank | Score | Branch | P(s) | I(s) | Status | Rule | Deduction |
|---|---|---|---|---|---|---|---|
| 1 | 1-0 | B_home_win | 17.5% | 0.673938 | SCORE | R2 exact-cell leader | 1-0 has the largest P(s) among eligible S10 scores. |
| 2 | 0-0 | B_draw_low | 12.7% | 0.604272 | EXCLUDED | R1 hard gate | EXCLUDED_BY_R1_TOTAL_ACTIVITY_GATE |
| 3 | 2-0 | B_home_cushion | 12.1% | 0.525311 | SCORE | R3 favorite-cushion representative | 2-0 maximizes I(s) inside B_home_cushion. |
| 4 | 1-1 | B_draw_low | 12.0% | 0.563192 | SCORE_4 | R5 fourth-score | 1-1 is the best excluded thesis-compatible danger branch. |
| 5 | 0-1 | B_away_win | 8.6% | 0.456191 | EXCLUDED | R2-R6 exhaustion | 0-1 loses branch or index comparison to the disclosed representatives. |
| 6 | 2-1 | B_home_btts | 8.3% | 0.523646 | SCORE | R4 complementary scoring branch | 2-1 is the best Colombia-win score that admits one Ghana goal. |
| 7 | 3-0 | B_home_cushion | 5.6% | 0.313272 | EXCLUDED | R2-R6 exhaustion | 3-0 loses branch or index comparison to the disclosed representatives. |
| 8 | 1-2 | B_away_win | 4.1% | 0.278204 | EXCLUDED | R2-R6 exhaustion | 1-2 loses branch or index comparison to the disclosed representatives. |
| 9 | 3-1 | B_home_cushion | 3.8% | 0.264521 | EXCLUDED | R2-R6 exhaustion | 3-1 loses branch or index comparison to the disclosed representatives. |
| 10 | 0-2 | B_away_win | 3.0% | 0.217273 | EXCLUDED | R2-R6 exhaustion | 0-2 loses branch or index comparison to the disclosed representatives. |
Let G={0,...,7} x {0,...,7}. S10 is the first ten scores after sorting by P(s): {1-0, 0-0, 2-0, 1-1, 0-1, 2-1, 3-0, 1-2, 3-1, 0-2}.
For s=(a,g), P(s)=Pois(a; lambda_COL) Pois(g; lambda_GHA), lambda_COL=1.3838, lambda_GHA=0.6833.
Branches are deterministic: B_home_win, B_home_cushion, B_draw_low, B_draw_high, B_away_win, B_home_btts.
I(s)=0.45 exact probability index + 0.25 bucket support + 0.20 branch mass index + 0.10 current form fit + 0.04 emotional index.
Colombia last-50 W-D-L is 31-12-7; GF/M=1.860, GA/M=0.800.
Ghana last-50 W-D-L is 16-14-20; GF/M=1.220, GA/M=1.320.
Recent form: Colombia scored in 8/10 last-10; Ghana scored in 6/10 last-10.
Current World Cup rows: Colombia n=3 with GF/M=1.333; Ghana n=0 with GF/M=0.000.
Hard scoring gates remove only scores contradicted by near-perfect recent and current-tournament scoring evidence.
The eligible exact-cell leader, i.e. the score with maximum P(s), must enter C4.
If P(Colombia win)>P(draw), choose the best I(s) representative from B_home_cushion.
If BTTS probability is at least 0.30 and Colombia win exceeds Ghana win, choose the best Colombia-win score with one Ghana goal.
Choose the fourth score by the best nonzero draw/danger branch; it is part of C4.
If fewer than three score representatives exist, fill remaining score slots by highest eligible I(s). Scores with total 0 are removed if P(over 1.5)>=0.55.
1-0 is disclosed because it lies in B_home_win, passes R1, and is selected by R2 exact-cell leader with I=0.673938 and P=0.175122.
2-0 is disclosed because it lies in B_home_cushion, passes R1, and is selected by R3 favorite-cushion representative with I=0.525311 and P=0.121170.
2-1 is disclosed because it lies in B_home_btts, passes R1, and is selected by R4 complementary scoring branch with I=0.523646 and P=0.082795.
1-1 is the fourth C4 score because it is the best excluded danger branch with I=0.563192.
0-0 is excluded by R1 hard gate: EXCLUDED_BY_R1_TOTAL_ACTIVITY_GATE
0-1 is excluded by R2-R6 exhaustion: 0-1 loses branch or index comparison to the disclosed representatives.
3-0 is excluded by R2-R6 exhaustion: 3-0 loses branch or index comparison to the disclosed representatives.
1-2 is excluded by R2-R6 exhaustion: 1-2 loses branch or index comparison to the disclosed representatives.
3-1 is excluded by R2-R6 exhaustion: 3-1 loses branch or index comparison to the disclosed representatives.
0-2 is excluded by R2-R6 exhaustion: 0-2 loses branch or index comparison to the disclosed representatives.
Under Definitions 1-4, Data 1-4, and Rules R1-R6, C4={1-0, 2-0, 2-1, 1-1}.
Data room
Analysis files for proofing.
Public sources used for this page. Betting or private staking decisions are logged separately from public claims.