If you have read anything about EPSO/AD/427/26 — including the earlier guides on this blog — you already know the headline: verbal reasoning is the test that sets your ranking. That is true. But it has an awkward side effect. Candidates read “verbal is what counts” and quietly downgrade numerical and abstract reasoning to background noise. In the autumn 2026 sitting, that is exactly how well-prepared people get eliminated before their verbal score is ever weighed.
The rule almost everyone gets half-right
Part 1 is three tests: verbal reasoning (20 questions, 35 minutes), numerical reasoning (10 questions, 20 minutes) and abstract reasoning (10 questions, 10 minutes). To go through to Part 2 you have to clear two thresholds at once — at least 10 out of 20 on verbal, and a combined 10 out of 20 across numerical and abstract together. Miss either one and the competition is over, whatever you scored on the rest.
The part people remember is the second half of the story: numerical and abstract do not feed the ranking. The official weighting table is blunt about it — those two tests count for nothing in the preliminary ranking and nothing in the final one. Only verbal, and later EU Knowledge, Digital Skills and the essay, move your position on the list. What gets lost is that “no ranking points” is not the same as “doesn't matter.”
Why “counts for nothing” is a trap, not a discount
A mark on the numerical test is worth nothing and everything at the same time. Nothing, because a perfect 10 earns you the same ranking position as a bare pass — the points simply do not accumulate. Everything, because dropping one mark under the combined threshold does not push you a few places down the list; it takes you off the list entirely.
That is the shift in how to think about them. Numerical and abstract are not scored exams in the way verbal is. They behave more like a locked door in front of Part 2. You either open it or you go home — and once it is open, it does not matter how elegantly you opened it.
The combined threshold changes how you train
Because the pass mark is combined, numerical and abstract are really one hurdle, not two. Ten correct answers out of twenty, in any split. Seven on abstract and three on numerical clears it. So does four and six. You do not have to pass each test on its own — you have to make sure the pair, together, reaches ten.
That has a practical consequence for the final weeks. If one of the two is clearly your weaker test, you do not need to drag it up to a standalone pass. You need to protect your stronger test so it actually delivers on the day, then squeeze whatever you can from the weaker one. What you cannot do is write either test off completely: a genuine collapse on one — a two-out-of-ten — leaves too much for the other to make up in the time available.
Numerical reasoning in twenty minutes
Ten questions, twenty minutes — two minutes each. The task is reading tables and graphs, pulling the right figures out of them, and doing basic arithmetic with the on-screen calculator. It is not a maths test in the school sense. There is no algebra to derive and no formula to recall. The whole difficulty is speed and accuracy under a clock, on data that is deliberately a little messy.
Where marks leak away is predictable:
- Unit mismatches — the table is in thousands, or in millions of euro, and the answer is calculated as if it were plain units.
- Percentage versus absolute — the question asks for a share and you compute a total, or the other way round.
- Reading the wrong row or column under time pressure, especially in a crowded table.
- Growth-rate confusion — the same absolute rise means a very different percentage from a small base than from a large one.
The fix is not more maths. It is repetition on the exact format until the reading becomes automatic: know where to look, decide what the question is really asking, and check the answer against the options before moving on. Practise timed from the start, and review every wrong answer to see which of the traps above caught you.
Abstract reasoning in ten minutes
Ten questions, ten minutes — sixty seconds each, the tightest budget in Part 1. You are shown a sequence of figures, work out the rule that connects them, and pick the figure that comes next. The rules recur: rotation, reflection, shapes added or removed, movement around the frame, changes in shading or in number. There is a fairly small library of them, and it repeats.
Abstract eliminates people not because the patterns are hard, but because a minute is barely long enough to spot one when the format is unfamiliar. Cold, most people are slow. After a few hundred items, the common rules start to announce themselves, and the same question that took ninety seconds in week one takes thirty in week four. That change is entirely a product of practice.
On the day, work in a fixed order: scan the sequence for what changes between figures, name the rule to yourself, predict the next figure, then find it among the options rather than reasoning about each option in turn. If a question will not yield after your minute, mark an answer and move — which brings up the point most candidates forget.
There is no penalty for a wrong answer — use it
Neither test penalises a wrong answer. A blank and a wrong tick score the same zero, so with the clock running down there is never a reason to leave a question empty. On abstract especially, where the time is brutal, a quick guess on the last question or two is simply free expected value. Answer everything.
The calendar is the real pressure
Part 1 is expected in October or November 2026. You will receive an invitation roughly three weeks beforehand, and the exact date and time appear on the testing platform at least two weeks out. Rescheduling is not possible, so the sensible plan is to be ready before the invitation lands rather than after.
Work backwards from that window. Short, timed sets most days beat occasional long sessions — the pattern recognition these tests reward is built by frequency, not by marathon cramming. And split your attention honestly: verbal deserves the larger share because it sets your ranking, but numerical and abstract deserve enough regular, timed practice that neither becomes the reason a strong verbal score is never counted.
The bottom line
Verbal decides where you land. Numerical and abstract decide whether you get to land at all. Give the two pass/fail tests the respect a locked door deserves — not the bulk of your hours, but steady, timed practice in the weeks before Part 1, so the gate is open when it counts.
You can drill both right now. EPSODrill opens ten questions per test with no sign-up, so you can try the exact format today — numerical reasoning and abstract reasoning — and see where your minute actually goes.