Analysis — where profit was takable, how often, and where it died
Filterable per sport. Each chart finds conditions where the highest profit was takable,
where profit was most frequent, and where net and frequency turned into danger zones.
Profit is the $1-share-equivalent in cents, without costs.
Counts before rates; every rate is x% (a/b). Nothing here predicts the next game.
Chronological execution sells on a later executable target touch before the observed ladder kill; otherwise the position settles. Unknown deadlines remain unknown. The hindsight comparison lets eventual winners sell on later prints.
A random-chance profit baseline is unavailable.
No matches are left out.
A1 · Coverage and outcomescandidate: buy at 0.10× the opening price, hold to the end · per sport × cohort · every rate is x% (a/b)
What this chart says
Each row is one sport and one cohort in this window. Void, pending and unknown are counted, never dropped. Fill rate is fills / opportunities. Win rate is game won / (game won + game lost). Net is the sum of unit-price differences in cents ($1-share-equivalent), without costs. Buy-and-hold is opening price to confirmed payout on the same matches. All sports uses the rows above — not a pooled fake sport.
ASCII snapshot of this table (copyable; hidden until opened)
Coverage strip by weekhow many matches have tape this week — not profit
Q17 · net / stake versus rank-gap bucketdecision-time Timeportal score gap · E/F floors · sampled max marked · public-file proxy
A2 · Highest profitcandidate: buy at 0.10× the opening price, hold to the end · x = actual fill price in 5¢ bands · y = minutes after the start when the fill happened · colour = net cents
What this chart says
One candidate, replayed once per eligible match. Colour is total net cents, zero-centred, shared by every sport panel. Each cell prints fills and net. The outlined cell is the greatest net among cells with at least the min-opportunities floor (default 10); cells under the floor stay drawn but hatched. It is compared with the next qualifying cell and with buy-and-hold on the same population. Never sum alternative cells. Tooltip: fill price/time ranges, exits, pending, permitted ceiling (hindsight). Click a cell for cell.php, then odds.php.
ASCII snapshot of this chart (copyable; hidden until opened)
A3 · Most frequent profitcandidate: buy at 0.10× the opening price, hold to the end · same axes as A2 · colour = profitable / opportunities · count/denominator first
What this chart says
An opportunity is an eligible match scheduled for this candidate, before knowing whether it fills. Missing coverage does not remove a known opportunity. Print profitable count/denominator first, then the percentage and net. Outline and headline only cells with at least the min-opportunities floor (default 10); a 1/1 cell stays drawn but hatched. Unknown outcomes have a lower bound (known profitable / opportunities) and an upper bound (known profitable plus unresolved / opportunities). No coin-flip baseline.
ASCII snapshot of this chart (copyable; hidden until opened)
A4 · Danger mapscandidate: buy at 0.10× the opening price, hold to the end · two synchronised maps · worst net · highest loss fraction
Worst net (zero-centred). All sports draws separate panels, never a pooled panel.
What this chart says
ASCII snapshot of worst-net map
Highest loss fraction (losing trades / opportunities). Same min-opportunities floor as A3; cells under the floor stay drawn but hatched and are not outlined.
What this chart says
ASCII snapshot of loss-fraction map
Beside each outlined cell: game won/lost/void/pending crossed with sold at target, reached but unsold, never reached, observation incomplete. Bounce then die means a game loser touched the selected target before kill, remained unsold, and subsequently lost value. After-kill touches cannot rescue losers.
A5 · Candidate comparisoncandidate: buy at 0.10× the opening price, hold to the end · profit first · profitable-trade rate only in this comparison · neighbours · weekly strip
What this chart says
ASCII snapshot of this comparison
A7 · This setting beside doing nothingcandidate: buy at 0.10× the opening price, hold to the end · same matches · two do-nothing baselines and the four neighbouring settings
per-match differences against
Every setting in the table above is replayed on the same matches under the same
execution rule. “Filled by both” is the number of matches where this setting and that one
both bought; the two net columns beside it are counted on exactly those matches, so the
difference is a paired one. Matches only one of the two filled are counted separately and never
folded into the difference. Click a header to sort; untick a match to leave it out everywhere.
ASCII snapshot of this comparison (copyable; hidden until opened)
A8 · Which groups made or lost money under this settingcandidate: buy at 0.10× the opening price, hold to the end · basket × sport · ordered by return on stake · net dollars beside it
height
The first row is every match in this selection counted once, whatever basket it
sits in. A group with fewer filled matches than the floor prints its counts and “too few”,
is drawn hollow and is listed after the ranked groups. A group that staked nothing has no
return to read. Click a header to sort.
ASCII snapshot of this table (copyable; hidden until opened)
A6 · Owner decisioncandidate: buy at 0.10× the opening price, hold to the end · not live · money rules change only on superadmin
What this chart says
ASCII snapshot of the rule
loading match outcomes…
Legacy views — these do not depend on the candidate controls above (kept for reference)
Legacy sport × basket win-rate table with reached ≤28¢ / ≥81¢ columns. Candidate-independent — the buy/sell candidate controls above do not change these numbers.
1 · Legacy win-rate table (same matches as A1)
cohort
sport
What this chart says
Each row is one sport and one basket (who called the favourite). Subtotals sit under each sport; the last row is every sport together.
Win rate % is won / (won + lost) — void and open do not enter the denominator. Hover a count for the first ten match titles (links open the odds page). Click a column header to sort.
ASCII snapshot of this table (copyable; hidden until opened)
Coverage-by-day canvas: how many games started per UTC day (coverage, not profit). Candidate-independent — the buy/sell candidate controls above do not change this chart.
Matches per day, all of them, and how many of the total were woncount chart — SPX / NAS100 baselines do not apply
resolutionseries
cohort
sport
⋯ drag height ⋯
What this chart says
Each solid line is how many games in that basket started in that UTC day (or week / month). Each dashed line is how many of those we won.
The x-axis is real time spanning the selected window; empty buckets are drawn as 0 so gaps stay visible. Click a line to read just that series; click empty space to deselect. Wheel zooms time; drag pans. The burned legend collapses on click.
ASCII snapshot of this chart (copyable; hidden until opened)
Old A2 split table: favourite buy-window low (win.lo) histogram, won vs lost. Candidate-independent — the buy/sell candidate controls above do not change this chart.
3 · Where did the favourite’s price bottom out while bets were open?buy-window low (win.lo) · 5¢ bins · won vs lost · low while bets were open (before the buy window closed)
cohortoutcome
cohort
sport
⋯ drag height ⋯
What this chart says
Each bar is a 5-cent bin of the favourite’s lowest price while bets were still open (win.lo — the low before the buy window closed). Never the full-tape low: a loser always prints 0¢ after settlement.
Won is a solid bar; lost is hatched. Void and open games are not in these bars. The vertical mark is the Reached-low X¢ control.
Click a bar to identify won or lost; click empty space to deselect.
ASCII snapshot of this chart (copyable; hidden until opened)
Legacy ladder-backtest ranking: strategy × exit rows from the old backtest. Candidate-independent — the buy/sell candidate controls above do not change these numbers.
Legacy backtest ranking (A5 is the candidate comparison above)
cohort
sport
Backtest not generated yet.
Top 15 strategies by total net (units of stake).
Net is the sum of per-fill returns minus 1. Every strategy’s 95% bootstrap CI includes zero on the TP-only sample — direction, not proof.
For TP-only · won / TP-only · lost the table is conditional fill/return among matches that already had that outcome, not a tradable P&L.
Net and win rate by how deep we buyx = k (fraction of open / of pregame)
xseries
⋯ drag height ⋯
The shipping ladder, rung by rungfills · HOLD net · SHIP net · win % (right axis)
series
⋯ drag height ⋯
Sweet spot (descriptive — not a forecast)
Proposed per-match rule (TP-only favs)
What this chart says
backtest not generated yet
ASCII snapshot of this backtest (copyable; hidden until opened)ASCII snapshot of this view (copyable; hidden until opened)
Public bundle only — prices, outcomes and baskets. Window ends at the latest match with data, never wall-clock now. Timestamps are UTC ISO; tooltips add a KST twin.