SIGNAL PATH

From match to number.

What happens between a kickoff somewhere in Europe and a figure in your area — every station, and above all the points where data leaves the path again.

Seven stations, four exits

The trace runs top to bottom. At four points something branches off — matches, forecasts or bets leave the path there rather than being quietly passed along. Those branches are the interesting part.

01

Collect

Several independent sources deliver fixtures, results, detail statistics, odds and lineups. Every match arrives more than once — one copy per source.

  • fixture
  • result
  • shots · corners · cards
  • odds
  • lineup
Exit 1

Results disagree. When two sources report a different final score, the match is not scored — it is flagged. Nothing is guessed.

sources
6
competitions
71
matches total
17,718
of those finished
13,080
02

Understand

Raw events become form: how a side has been shooting, fouling and winning corners lately — and how good the opposition was while it did.

  • rolling form
  • opposition strength
  • home advantage
  • player availability
teams
1,362
players
18,210
team observations
18,884
player observations
162,721
feature rows
375,781
rule
pre-kickoff only
03

Compute

One calculation per market, one probability per selection. The figure that gets published is then calibrated against bets that have already been settled — never against ones still open.

markets
13
models in use
9
forecasts total
79,016
calibration
settled only
04

Select

Not every forecast is published. This is where the day is decided — and above all, whether the outcome will be checkable at all.

Exit 2

Duplicate fixture. The same match exists once per source. Only one record goes forward, or the identical forecast would appear twice in your area.

Exit 3

No statistics source. A market that needs shots on target or cards is published only where those figures can actually arrive later. Otherwise it would produce an analysis nobody can ever score.

forecasts total
79,016
published
1,166
share
1.5 %
05

Publish

Before kickoff, timestamped, unchangeable afterwards. Every member receives, from that same list, the number their plan provides for — highest rated first.

analyses so far
1,166
Essential
3 a day
Pro
5 a day
Expert
8 a day
days of history
730
after kickoff
unchangeable
06

Settle

At regular intervals the system checks which matches have finished and scores them against verified results. Hits and misses feed the same public rate.

Exit 4

Statistic missing. When a single figure never arrives, the analysis reads “under review” and keeps being retried. If it stays missing for a week it is discarded — neither a hit nor a miss.

scored
1,030
hits
708
misses
322
hit rate
68.74 %
under review
0
discarded unscored
25
07

Learn

The system works out where it was weak, drafts variants and runs them against the version in use — on matches neither of them has ever seen. Only a winner takes over.

tested on
held-out matches
promotion
on merit only
failed attempts
stay visible

└─ back to 03 · every day

The path is open. The recipe is not.

This drawing shows how data moves through the system and where it leaves it. It does not show which quantities enter the calculation, how they are weighted, or the threshold above which an analysis is published.

That is deliberate. The path can be checked without handing over months of work — and in the end what gets judged is not the formula but the rate.

A method you can count for yourself is one you need not believe.

Every figure on this page is counted afresh on each visit.