Trajan, predictapp's soccer model
Trajan is the model behind every soccer probability in predictapp. It forecasts how many goals each team will score, turns that into the chance of every scoreline, and derives all its goal markets from that one distribution. It has been live since August 2026.
By Daniel, founder of predictapp. Updated 2026-11-01.
How it works, in four steps
- ›Team strength: it rates each team from its results and the quality of its performances, adjusted for the competition and the venue.
- ›Expected goals: a machine learning model turns that into how many goals each team should score in this match.
- ›Scorelines: those two numbers become a full grid of scoreline probabilities, 0-0, 1-0, 1-1 and so on. The grid accounts for the way two teams' goals depend on each other, which matters most for draws.
- ›Markets: goals, both teams to score, clean sheets, exact score and the rest are read from that same grid, so they cannot contradict each other.
Why one scoreline grid matters
Many prediction sites estimate each market on its own. That is how you end up with a home win and an under 2.5 goals that do not fit together. With Trajan, every goal market comes from one picture of the match. It also lets predictapp work out the real chance of two events in the same game landing together, instead of multiplying two numbers that are not independent.
Where bookmaker odds come in
Bookmaker odds are not an input to the model, with one exception. Trajan scores its own reliability on each match. When that score is weak and a sound market price exists, the published home, draw and away numbers are blended with bookmaker odds. That is about 6% of matches. Goal markets stay pure Trajan, and if there is no valid market price, the outcome numbers do too.
So Trajan is not fully independent of market data, and we would rather say so. We chose this because a model that knows when it is unsure is more useful than one that always sounds certain.
How it stays current
New finished matches update team state and the daily fit. The model's design stays frozen between deliberate refreshes, so a good week cannot quietly change the method and a bad one cannot be tuned away.
How we know it works
In the latest test, on 2,189 matches, Trajan's pick was right 54.1% of the time, and its confidence levels closely match real results.
What Trajan cannot do
- ›It gives probabilities, not certainties. A 70% pick loses about 3 times in 10.
- ›It forecasts before kickoff. It is not an in-play or live model.
- ›It does not model yellow cards as a market, because they are too volatile.
- ›It covers soccer. predictapp's NFL projections are a separate model.
Questions people ask
What is Trajan?
Trajan is predictapp's soccer prediction model. It turns each team's expected goals into a full scoreline distribution, and predictapp's soccer probabilities and fair prices come from it.
Who built Trajan?
Daniel, predictapp's founder, who has been building sports models for over 15 years and has a master's from KLU in Germany and data science specializations from the University of Michigan.
Does Trajan use bookmaker odds?
Only on about 6% of matches, where its own reliability score is weak, and only on the home, draw and away numbers. Goal markets never use them.
Can I use Trajan outside the app?
Yes. Its predictions are available on the web, in the iOS and Android apps, through the API and through MCP in AI assistants such as ChatGPT and Claude.
Read Trajan's numbers on a real match
One free reveal every day, on any match you choose.
Create free accountpredictapp is for adults 18 and over. No model can guarantee a result. Never bet more than you can afford to lose.