# TMUA overseas candidates: the official numbers, and where they stop

> Official UAT-UK data on TMUA candidates outside the UK: a mean of 4.71 against 3.98, the full percentiles, the first-language split, and what it cannot show.

[Formatted version of this page](https://www.frontiervue.com/exams/tmua/guides/international-candidates)

The question a mainland applicant asks first is whether overseas candidates outscore UK ones on the TMUA. For the 2026 cycle the official technical report puts figures on it: of 16,092 TMUA candidates, 6,733 (42%) live outside the UK and the EU, and that group's mean scaled score is 4.71 against 3.98 for the UK and 3.94 for the EU. The first-language table points the same way. This page sets out both tables cell by cell, shows that overseas is counted three different ways across the official documents, and is just as careful about the conclusions these figures will not carry.

## Key takeaways
- Candidates resident outside the UK and the EU, the report's Other group, numbered 6,733 and made up 42% of the cohort. Their mean scaled score was 4.71 against 3.98 for the UK and 3.94 for the EU, a comparison the report states in its own prose.
- The gap is not only in the mean. The Other group's 90th, 75th and 50th percentiles are 7.5, 5.8 and 4.5; the UK group's are 6.0, 4.9 and 3.9. The Other group also carries the widest spread of the three, an SD of 1.89.
- First language runs the same way: the 6,837 candidates (42%) whose first language is not English averaged 4.60, against 4.05 for the 9,255 (58%) reporting English. The report does not leave that hanging. It calls the finding interesting and says it most likely reflects the low language load of a test whose questions are almost all mathematical.
- Overseas is counted three ways in the official documents: the technical report splits delivered candidates by residence into UK, EU and Other; the annual report splits bookings into UK and overseas by sitting; and its continent table pools all three UAT-UK tests. Mixing them produces figures that contradict each other.
- The Other group is not a proxy for candidates from China. It covers everywhere outside the UK and the EU, and the annual report's country-level bar chart prints no values in its text layer, so this page gives no candidate count or score for any single country.

### Quick answer: the overseas numbers

| Item | Value |
| --- | --- |
| Size of the overseas group | 6,733 candidates, 42% of the 16,092 who sat the TMUA (the report's Other group) |
| Means by residence | Other 4.71, UK 3.98, EU 3.94 |
| By first language | Not English 4.60 (6,837 candidates), English 4.05 (9,255) |
| Is there a threshold | No. No pass mark is published, and neither report prints a score threshold of any kind |

> Every figure here comes from a printed table or sentence in UAT-UK's 2025/26 TMUA technical report and its 2025-26 annual report. Nothing is read off a chart image and nothing is derived across tables.

## Contents
1. The official distribution, by area of residence
2. First language: which way the gap leans on a maths test
3. Three official ways to count an overseas candidate
4. What these numbers cannot tell you
5. How an overseas candidate should use these two tables

## 1. The official distribution, by area of residence

The TMUA is a non-modular test. Two papers, Applications of Mathematical Knowledge and Mathematical Reasoning, carry 20 items each with 75 minutes each: 40 items and 150 minutes in total, resolving into a single scaled score from 1.0 to 9.0 reported to one decimal place. So the question of how overseas candidates do has exactly one number to answer it here, with no module-by-module comparison to assemble. The table below puts the whole cohort and the three residence groups on the same ruler.

- **6,733** TMUA candidates resident outside the UK and the EU, 42% of the cohort
- **4.71** Mean for the Other group; the UK is 3.98 and the EU 3.94
- **7.5** Other group 90th percentile; the UK sits at 6.0
- **1.0 to 9.0** The observed range in every one of the three residence groups, the full scale

|  | All candidates | Resident in the UK | Resident in the EU | Resident elsewhere (Other) |
| --- | --- | --- | --- | --- |
| Candidates and share | 16,092 (100%) | 8,643 (54%) | 716 (4%) | 6,733 (42%) |
| Mean (SD) | 4.28 (1.74) | 3.98 (1.54) | 3.94 (1.74) | 4.71 (1.89) |
| 25th percentile | 3.0 | 2.8 | 2.7 | 3.4 |
| 50th percentile (median) | 4.1 | 3.9 | 3.7 | 4.5 |
| 75th percentile | 5.3 | 4.9 | 4.9 | 5.8 |
| 90th percentile | 6.7 | 6.0 | 6.4 | 7.5 |
| Observed score range | 1.0 to 9.0 | 1.0 to 9.0 | 1.0 to 9.0 | 1.0 to 9.0 |

> Two sentences carry this table in the report. One: candidates outside the UK and the EU performed more strongly, a mean scaled score of 4.71 against 3.94 for the EU and 3.98 for the UK. Two: the Other group's whole distribution sits shifted toward the higher scores. Both describe groups of people rather than sets of papers. Everyone in the cycle was scaled with the same constants, so form difficulty is already out of the number, and the report's own illustration is that a candidate scoring 6.5 in either January or October has higher ability than one scoring 4.2. This page repeats those two sentences and supplies no cause of its own.

## 2. First language: which way the gap leans on a maths test

- **English as first language:** Mean 4.05 (9,255 candidates, 58%; median 3.9, 75th 4.9, 90th 6.2)
- **Another first language:** Mean 4.60 (6,837 candidates, 42%; median 4.4, 75th 5.8, 90th 7.3)
- **The cause the report names:** Low language load (It most likely reflects the low language load of a mostly mathematical test, says the report, which in its summary traces the same gap to strong performance from candidates outside the UK)
- **Item-level bias check:** Nothing flagged (The report finds no significant differential item functioning by first language or UK school type)

Both tables cut the same 16,092 people, and they are not two independent findings. The report's summary states outright that candidates whose first language was not English outperformed native English speakers because of strong performance from candidates outside the UK, which makes the residence split and the language split largely one story told twice. English was the first language of 58% this cycle, down from 60% in the last one. At item level nothing was flagged for differential functioning by first language.

> Two cautions. The explanation above belongs to the report, not to us: low language load, plus a strong showing from candidates outside the UK. Quoted as the report's reading it holds; restated as your own conclusion it becomes speculation. And an explanation of a gap is still not a verdict on a population. Every group here spans the full 1.0 to 9.0, and one cohort's distribution cannot show that any country or curriculum teaches mathematics better.

## 3. Three official ways to count an overseas candidate

**Three official counts of overseas**: Different denominators, so the numbers will not line up

- Technical report, Table 7. Delivered TMUA candidates by residence: UK, EU, and the group the table calls Other while the prose calls it Rest of World
  - Other: 6,733 candidates, 42%
  - Comes with means and percentiles
- Annual report, Table 3: Bookings split into UK and overseas, event by event
  - October TMUA: UK 5,684, overseas 4,425
  - January TMUA: UK 3,897, overseas 2,882
- Annual report, Table 4: Overseas candidates by continent, all three tests pooled, no scores
  - Asia 8,745, Europe 1,789
  - North America 650, Africa 173

### Overseas candidates by continent

Annual report Table 4: Asia 8,745, Europe 1,789, North America 650, Africa 173, Oceania 135, South America 36. The count spans every UAT-UK test, ESAT, TARA and TMUA together, so none of it is a TMUA figure. This page neither splits it down to one test nor sets it against the technical report's 6,733. The two rest on different bases, and any ratio between them would be ours rather than anyone's official statistic.

### TMUA bookings, UK and overseas, by sitting

The TMUA row of annual report Table 3: October, 5,684 in the UK and 4,425 overseas; January, 3,897 in the UK and 2,882 overseas. These are bookings, while the technical report's 16,092 counts tests actually delivered, so the two were never going to reconcile. The annual report separately gives 16,888 TMUA bookings against 16,092 delivered, 48.9% of all UAT-UK tests. Check which basis a figure is on before you compare it with another.

### Where the test is actually sat

UAT-UK's stated reason for choosing Pearson is a network of over 5500 professional test centres across 180+ countries and territories. Elsewhere it reports that in 2025-26 tests were taken in 129 different countries. Those answer different questions, the size of the network and the places candidates actually turned up, and neither answers yours. Whether a seat exists near you is for the official booking system to say.

### The main markets outside the UK

The annual report names China, India, Singapore, Malaysia, Hong Kong, the United States, the UAE, Türkiye, Spain and France as key markets outside the UK. It also carries a bar chart broken down by country, with UK, China, India, Singapore, Malaysia and Hong Kong along the axis, but the bar values sit only inside the image and are not printed in the text. This page therefore gives no candidate count for any individual country.

The three counts do not substitute for one another. To judge which population your score sits inside, use the technical report's residence table. To gauge the size of a sitting, use the annual report's booking table. To read regional trend, use the continent table. A ratio built by mixing them belongs to no official document and cannot be checked by anyone.

## 4. What these numbers cannot tell you

### Not a verdict that overseas candidates are better at maths

What the report says under this table is the shape of it: the Other group's distribution is shifted toward the higher scores. It does not push that into a claim about which population is cleverer, and neither does this page. Nobody was assigned to a residence group; who books the test, and for which sitting, follows from where they are applying. Each group runs the full 1.0 to 9.0, so the spread inside a group swamps the distance between the group means. Reading 4.71 against 3.98 as a ranking of ability is a step the reader takes alone.

### The Other group is not shorthand for China

Other means everywhere outside the UK and the EU, from Asia to South America. No country-level score distribution is published anywhere in these documents, and the annual report's country chart keeps its values inside the image. So any claim about an average TMUA score for candidates from one particular country comes from somewhere other than these two documents, and this page will not derive one for them.

### A percentile is not a threshold

No pass mark is published and neither report prints a score threshold. The report says something closer to the opposite: a range of universities uses these tests and they select at different points along the scale, so a single line was never there to be found. The 75th and 90th percentiles are positions. Treat one as a safe zone and the safety is yours, not the data's. Use the distribution to locate yourself, and stop there.

### Absolute scores do not travel between cycles

The scale is anchored afresh every cycle. For this one the median October ability (-0.0639) was fixed at 4.5 and the 90th-percentile ability (1.4634) at 7.0, and the regression through those two points produced the constants 4.6046 and 1.6369, with the result capped at 1.0 and 9.0. Because the anchors come from this cycle's own data, this year's 4.71 cannot be laid alongside a figure from another year.

### A group mean is not a personal forecast

Scaled-score reliabilities run 0.82 to 0.87, with every form above 0.80, and the standard errors of measurement printed beside them fall between 0.62 and 0.64. Under classical test theory a true score lies within one SEM of the observed score about 68% of the time, and within two about 95%. The distance between the EU's 3.94 and the UK's 3.98 is far smaller than any of those SEMs. What survives is the distance between the Other group and the other two, and even that describes a cohort rather than a person.

> On sourcing: every figure on this page comes from a printed table or sentence in UAT-UK's 2025/26 TMUA technical report and its 2025-26 annual report. Nothing is read off a chart image, and nothing is added or derived across tables. On item quality the report's verdict is positive: 100% of the items used across the TMUA forms met the criteria, well above the 80% target typical for new items, so a low score cannot be blamed on a faulty paper.

## 5. How an overseas candidate should use these two tables

1. **Change your benchmark row** (Table 7, Other)
If you live outside the UK and the EU, the row to read yourself against is Other: median 4.5, 75th percentile 5.8, 90th percentile 7.5. The all-candidate row, 4.1, 5.3 and 6.7, will flatter your position, because it folds in the UK and EU groups whose distributions sit further left.

2. **Cross-check against first language** (Table 10)
The row for candidates whose first language is not English gives a median of 4.4, a 75th percentile of 5.8 and a 90th of 7.3. The two tables cut the same cohort, so if the position you read differs sharply between them, the likely cause is a mixed-up definition. Go back, confirm which group you fall into, and read again.

3. **Practise to the full clock** (150 minutes)
The standard limit is 02:30:00. Among the 15,481 candidates in the timing analysis the median was 02:29:37, and 91% of all candidates spent between 02:15:00 and 02:30:04. Separately, 2,517 candidates (16%) logged at least one item answered in 10 seconds or less, or not reached at all, which the report reads as random guessing. Running the clock out is the normal case here.

4. **Write the target as a position** (No threshold)
Since no published threshold exists, the only target you can act on is a position, for example moving from the Other group's median to its 75th percentile. That formulation holds in any year, whereas a specific number only means something inside the cycle it came from and has to be recalibrated the moment the cycle changes.

Turn it into paper-level work: the [Paper 1 guide](/exams/tmua/guides/paper-1) and the [Paper 2 guide](/exams/tmua/guides/paper-2) cover how the two papers differ, the [scoring guide](/exams/tmua/guides/scoring) covers how the scale is built, and full timed sets live in the [online question bank](/papers/tmua) and the [past papers overview](/exams/tmua/past-papers). For the whole-cohort and per-event distribution, see the [score statistics page](/exams/tmua/guides/score-statistics).

## Questions
**Do overseas candidates find it easier to get a high TMUA score?**

The official figures are a mean of 4.71 for the Other group against 3.98 for the UK, and a 90th percentile of 7.5 against 6.0. That describes how two populations scored, not how hard scoring is: everyone in the cycle shares one set of scaling constants and the scaled score already has form difficulty stripped out. All three groups span 1.0 to 9.0, a group mean does not predict an individual, and none of this is evidence that the test is easier for anybody.
**Does the report give an average TMUA score for candidates from China?**

No. The technical report splits residence only into UK, EU and Other, where Other covers everywhere outside the UK and the EU. The annual report does carry a country-level bar chart with China on the axis, but the bar values live inside the image and are not printed in the text. Any country-specific TMUA score figure therefore comes from somewhere other than these two documents, and this page does not derive one.
**Is a non-English first language a disadvantage on the TMUA?**

This cycle the numbers run the other way: candidates whose first language is not English averaged 4.60 with a median of 4.4, against 4.05 and 3.9 for English-first-language candidates, and the report's word for the latter is that they performed weaker overall. Unusually for a table like this, it also says why: most likely the low language load of a test whose questions are almost entirely mathematical, and, in the summary, the strong showing of candidates living outside the UK. No item was flagged by first language. Keep the explanation attributed to the report, and remember it describes two populations rather than you.
**Which row should I set my target against?**

The row that matches how you would be classified. If you live outside the UK and the EU, that is the Other row of Table 7: median 4.5, 75th percentile 5.8, 90th percentile 7.5. Cross-check it against the non-English first-language row of Table 10, at 4.4, 5.8 and 7.3. The all-candidate row, 4.1, 5.3 and 6.7, is useful context but is the wrong benchmark for an overseas candidate. State the target as a percentile rather than a score, because the scale is re-anchored every cycle.
**What should an overseas candidate know about cost and test centres?**

Two fee tiers, and an overseas candidate is always in the higher one: £75 in the UK and the Republic of Ireland, £130 everywhere else. Read the annual report's bursary section as a UK-cohort figure and nothing more, because it does not describe anything you can apply for from abroad. On centres, the reports cite Pearson's network of over 5500 test centres across 180+ countries and territories and record that tests were taken in 129 countries in 2025-26. Dates, seats and deadlines come from the official announcements.

## Official sources
- [UAT-UK TMUA Technical Report 2025-26, Table 7: scaled scores by area of residence](https://uat-wp.s3.eu-west-2.amazonaws.com/wp-content/uploads/2026/08/07161815/UAT-UK-TMUA-Technical-Report-2025-26.pdf)
- [Same report, Table 10 (first language) and Table 4 (whole cohort and both events)](https://uat-wp.s3.eu-west-2.amazonaws.com/wp-content/uploads/2026/08/07161815/UAT-UK-TMUA-Technical-Report-2025-26.pdf)
- [Same report: scale construction, reliability and SEM, test timing and item quality](https://uat-wp.s3.eu-west-2.amazonaws.com/wp-content/uploads/2026/08/07161815/UAT-UK-TMUA-Technical-Report-2025-26.pdf)
- [UAT-UK Annual Report 2025-26 Candidate Data, Tables 3 and 4: bookings by geography, overseas candidates by continent](https://uat-wp.s3.eu-west-2.amazonaws.com/wp-content/uploads/2026/08/07161742/Annual-Report-2025-26-Candidate-Data.pdf)
- [Same annual report: test fees, the Pearson delivery network and country coverage](https://uat-wp.s3.eu-west-2.amazonaws.com/wp-content/uploads/2026/08/07161742/Annual-Report-2025-26-Candidate-Data.pdf)
- [UAT-UK Annual Reports index (public entry point for the 2025-26 reports)](https://esat-tmua.ac.uk/about-uat-uk/annual-reports/)
