+00 Reykjavik
Profile
T
FC Orenburg 0:0 Akron · 8'Dynamo Kyiv 0:0 LNZ Cherkasy · 48'Groningen 2:2 Twente · 45'Leça 0:1 Fafe · 46'Amora 0:0 Real · 48'
Menu
Statistics sections06.09.2026
Matches
Schedule, results, lineups
1209matches today
TodayTomorrow
Teams
Form, streaks, home and away
24775teams
OverviewSeason
Leagues and Cups
Tables, tournament trends
1239tournaments
Top-5All
Referee
Cards, penalties, strictness
19296referees
RatingBy league
1209
today's matches in the database
1239
tournaments in the database
24775
teams in the database
Analytics tools06.09.2026
Dropping odds
Line movement and volumes
892194signals today
TodayTomorrow
Day Analysis
Summary of the day's matches
1209matches in the overview
OverviewSeason
AI-match predictions
ML model probabilities
predictions with value
ValueAll
Tools
Referees, calendar, load, scenarios
4tools
Referee modelBuilder
Matchlists
Export data to Excel
XLSXexport format
DownloadHistory
Home/Articles/Glossary

Author consensus and agreement with the model: how to read them

Consensus shows which outcome most authors picked and how unanimous they are; agreement shows whether the model’s estimate matches them. Unanimity is not correctness, and disagreement is not an error.

Published September 7, 2026TheAnalytic

In the verifier’s match table every match has three columns you will not find on other sites: how many picks have been published on it, where the majority leans, and whether the model agrees with the majority. This is not a recommendation to bet, but a way to see where opinions diverge.

Consensus

Consensus is the most frequent selection among the published picks on a match, and its share. “Over 2.5 · 71%” means that of all picks on the match, 71% took “more than 2.5 goals”. The number of picks is shown next to it: 71% of seven and 71% of two are different things.

Only open picks count towards the consensus. Bets from closed paid feeds are revealed after the match starts and do not enter the aggregate before that: otherwise a paid pick could be inferred from the consensus figure.

What consensus does not mean

A high share is unanimity, not correctness. Authors read the same news and look at the same lines, so they often converge on the obvious, and the obvious has already been priced by the market: the consensus outcome usually carries low odds. From the history of settled picks, the service shows next to each row of the distribution the historical 30-day ROI of bets of that kind — that is the answer to “do bets like this actually pay off”.

Agreement with the model

The TheAnalytic models estimate outcome probabilities on their markets independently of the authors. Agreement shows whether the model considers the same outcome most likely as the majority did. “Agrees” means they think alike; “disagrees” means the model sees it differently.

Disagreement is not anyone’s mistake but the most interesting spot in the table. It is precisely on disagreements that you can see who adds value: if authors are systematically right where the model disagrees, they have information that is not in the data; if the model is right, the authors are following the crowd. Which markets the models cover and how to verify them is described in the article on the models in the verifier.

The model’s pick

The “Model” column shows not the outcome with the highest probability but a meaningful pick: “over 0.5 goals” at 94% is trivial and says nothing. The service chooses an outcome from a confidence band where the estimate means something, and always labels the market: “1” on corners and “1” on goals are different picks.

How to use this

  • Look at the consensus together with the number of picks and the row’s historical ROI, not on its own.
  • Disagreement with the model is a reason to open the match page and see exactly what the authors took and what their CLV and ROI are on that market.
  • None of the three columns is a prediction by the service. The service shows who said what and how it holds up — the decision stays with the reader.