How to read a pick in 30 seconds

When the model finds a gap in your favour it doesn't say "back this". It hands you a card with five numbers, and each one answers a different question.

Learn to read it and you stop following picks blind.

The confidence tier

At the top sits the market and a confidence label. It comes from the size of the edge, and it isn't decoration:

• High — an edge of 7 % or more. Comes with a suggested stake.
• Medium — between 4 % and 7 %. Comes with a suggested stake.
• Low — between 3 % and 4 %. No stake. It's there for information.

Withholding a stake in the low tier is deliberate: when the edge is thin, we don't want to push money at you.

The edge

This is the heart of the card. It compares what the model believes against what the bookmaker pays, and expresses it as the expected return over the long run for every unit staked at that price.

An 8 % edge doesn't say "this match wins". It says: enter a hundred situations like this one and you come out ahead. The price beside it is exactly the one used in the calculation — which is why the next line matters.

Model probability vs market probability

Two percentages side by side. The first is what Delta believes; the second is what the bookmaker believes, with their margin stripped out.

When ours is higher, there's value. When they're nearly equal, there's nothing to do there, however much you like the match.

Minimum odds: your line of no return

This is the number that saves bankrolls and the one most people skim past.

It's the price below which the pick stops being worth taking. If your bookmaker offers less than that, the edge is gone: you're no longer making the bet the model found, you're making a worse one.

Prices move. The edge was measured against the price at that moment. Minimum odds tell you whether you're still in time.

The stake

The share of your bankroll that belongs on that bet. It comes from a classic growth formula, applied at a quarter of its strength and capped at 3 %.

Why so conservative? Because the full formula is brutally volatile and punishes any estimation error. A quarter captures nearly all the growth for a fraction of the risk of ruin. The 3 % cap exists because no single bet, however good it looks, deserves more than that.

When there's no card at all

Sometimes the analysis ends with no recommendation. That isn't laziness: a filter fired, and we always tell you which one.

• Not enough data — there was no real information for that match. See step one of the method: that's where the model goes quiet.
• No market — no prices to compare against, so there's nothing to measure.
• Low conviction — the model doesn't believe in that outcome enough. Each family of markets has a floor, and below it we don't bet however tempting the price.
• Gap too wide — if we disagree with the market by more than 20 points, the one who's wrong is probably us. Discarded.
• Edge under 3 % — it fits inside the error bar and the bookmaker's margin. That isn't value, it's noise.

An analysis with no picks is a result, not a failure. Most matches have nothing to offer, and saying so is part of the job.