Skip to content

LFM, the large football model

rendonarango.com
Football. Decoded.

The world’s first league-agnostic, multi-task AI that learns football — not stats.

Every pass, every duel, every heartbeat — understood by one neural model.

An intelligence engine to predict, simulate, and understand football in real time.

LFM — The Large Football Model.
→ Join the waitlist
Scroll to discover

One model. Every league.Every task.

Football’s foundation model — a shared neural core that understands structure, flow & intent across competitions.

LFM unifies global football knowledge — Premier League to Copa Libertadores. Trained on millions of matches, it predicts outcomes, surfaces hidden dynamics, and powers a new era of tactical intelligence.

Predicts · Simulates · Explains · Learns

Shared encoderCross-league generalizationContinuous learningMulti-task heads
Recalibrating in 00:15…

Real-time football intelligence.

Every 15 seconds, LFM recalibrates using live events, odds, and context.

It doesn’t just forecast — it plays along.

Live ingestion · Event-driven updates · Odds fusion · Neural calibration

Momentum Index
0.62
Event Delta
+1
Win Probability
64.0%
Calibration Error
0.028

The hidden geometry of the game.

LFM embeds every player, team, and event into a shared latent space.

Styles cluster together — discovered by AI, not predefined tags.

The game’s DNA, visualised.

Rewind. Rerun. Reimagine.

Simulate injuries, transfers, tactics, and momentum shifts to see how reality might unfold.

What if Arteta switched shape at halftime?
LFM rolls the timelines.
Is Martinelli the right late-game option?
LFM shows the possible impact.
Where should Raya target his goal kicks?
LFM finds the highest-EV target.
How would a 4-3-3 reshape our build-up?
LFM runs the scenario.

Neural Monte Carlo · Multi-agent dynamics · Predictive tactics

Build-up
  • Swap LB ↔ inverted CB
    build-up
    → +3.2% possession retained, +1.1% xG delta
  • Double pivot in build-up
    structure
    → press resistance ↑, final-third entries ↑
Formation & Tactics
  • Shift to 3-5-2 at 70'
    formation
    → +4.1% possession, +1.3% xG delta
  • Overload left with inverted winger
    overload
    → entries to zone 14 ↑, weak-side switch threat
Press & Defense
  • Trigger press on left center back
    pressing
    → more turnovers, higher counter risk
  • Man-mark #10 out of possession
    marking
    → disrupt passing lanes, fouls ↑ slightly
  • Hold line 5m higher
    rest-defense
    → offside traps ↑, counter risk ↑
Transitions
  • Counter-press 5s rule
    transitions
    → regained possessions ↑, transition shots ↑
Set-pieces
  • Short corners routine
    set-piece
    → xThreat +0.07 / match, more shots from zone 14
Tempo & Load
  • Bench fatigued CM
    load mgmt
    → variance narrows, late-game stability improves
  • Lower tempo after 60'
    tempo
    → variance narrows, recovery improves
Distribution
  • Keeper long to right half-space
    distribution
    → field tilt +2.1%, late variance ↑
Substitutions
  • Substitute forward at 65'
    substitution
    → +2.4% possession, +0.8% xG delta

AI meets reality

Prediction meets precision.

LFM blends neural certainty with market signals to deliver calibrated probabilities and interpretable EV.

Built for analysts, clubs, and investors seeking truth — not noise.

From probability to profitability.

Model probabilityMarket odds
Illustrative · synthetic data
ECE
0.092 → 0.054
Brier
0.186 → 0.142
Log-loss
0.624 → 0.571

Built for builders, analysts, coaches, investors, engineers, fans

LFM is a platform: APIs, SDKs, embeddings, and simulations for research, products, and club workflows.

The heart of a system

Native and web apps that display probabilities, xG flow, scenario outcomes, and much more in real time. Powered by LFM.

API playground

Experiment with endpoints, inspect payloads, and share queries.

Embedding explorer

Search, cluster, and compare match embeddings across leagues.

bash · lfm api
curl $LFM_API_URL/v1/match/predict   -H "Authorization: Bearer $LFM_KEY"   -H "Content-Type: application/json"   -d '{
        "home":"ARS",
        "away":"MCI",
        "league":"EPL",
        "features":{"odds":true,"recent_form":true,"injuries":true}
      }'

Preview. The API is not public yet.

LFM — Coming soon. The AI that understands football.
Join the waitlist

Join the waitlist

Be first to know when LFM opens up.

Join the waitlist
Tell me a bit about you and your project.
About you

I will only use this to contact you about LFM and the updates you opt into.

Organisation
Your project
Interests
Links optional

By submitting, you consent to the processing of your personal data for the purpose of responding to your request and, if you opt in, to receive product updates. You can withdraw consent at any time. I do not sell personal data.

Disclaimer

All data on this page is synthetic and for illustration only. Real model results will differ.