Every danger ranking you have read counts casualties. Counting rewards long, busy roads. We measured the rate instead.
Last updated:
1,710
Roads ranked
4 of 20
Top count roads that are also top rate roads
0.86x
The A38, against the national average
2,170
GB rate, motorcyclist KSI per billion miles
384 roads with at least ten motorcyclist casualties, coloured by how their casualty rate per mile ridden compares with the national average. Zoom in to find your roads.
Loading map
To see one road on its own, with its bend count, crash clusters, live weather and fuel, open the Ride Planner.
Each dot is a Department for Transport traffic count point on that road in 2025. The colour is the whole road's rate across Great Britain, not the rate at that particular spot, so a dot on a quiet stretch of a bad road still shows the road's colour. Roads with fewer than ten casualties in five years are not shown, because a rate built on a handful of casualties is not meaningful.
Britain's road danger rankings, including the one we published on this site, count casualties. The A38 comes top because 183 motorcyclists were killed or seriously injured on it between 2021 and 2025. What that mostly measures is that the A38 is 296 miles long and busy. Riders cover about 99 million miles a year on it.
Divide the casualties by the miles ridden and the picture inverts. The A38 records 1,858 motorcyclist KSI casualties per billion vehicle miles, against a national average on major roads of 2,170. It is a below average road. It ranks 1,116th of the 1,710 roads we can measure.
Test each of the twenty roads that top the casualty count against the national average per mile ridden and only five of the top twenty by count are statistically above the national average. Eight are below it, and seven cannot be told apart from it. Only four of them are also in the top twenty by rate.
| Road | Motorcyclist KSI | Rate vs national average | 95% confidence interval | Verdict |
|---|---|---|---|---|
| A38 | 183 | 0.86x | 0.74 to 0.99 | Below |
| A259 | 181 | 2.16x | 1.86 to 2.50 | Above |
| A40 | 162 | 0.66x | 0.56 to 0.77 | Below |
| A6 | 159 | 1.51x | 1.28 to 1.76 | Above |
| A4 | 156 | 0.87x | 0.74 to 1.02 | Not distinguishable |
| A5 | 151 | 1.08x | 0.91 to 1.26 | Not distinguishable |
| A2 | 150 | 0.84x | 0.71 to 0.98 | Below |
| A23 | 148 | 1.32x | 1.12 to 1.55 | Above |
| A13 | 131 | 0.74x | 0.61 to 0.87 | Below |
| A27 | 130 | 1.11x | 0.93 to 1.32 | Not distinguishable |
| A3 | 128 | 0.61x | 0.51 to 0.72 | Below |
| A406 | 111 | 0.71x | 0.58 to 0.85 | Below |
| A41 | 110 | 0.89x | 0.73 to 1.07 | Not distinguishable |
| A30 | 105 | 0.85x | 0.69 to 1.03 | Not distinguishable |
| A35 | 101 | 1.62x | 1.32 to 1.97 | Above |
| A1 | 98 | 0.52x | 0.43 to 0.64 | Below |
| A10 | 91 | 0.92x | 0.74 to 1.13 | Not distinguishable |
| A24 | 89 | 1.08x | 0.87 to 1.33 | Not distinguishable |
| A12 | 88 | 0.51x | 0.41 to 0.63 | Below |
| A57 | 87 | 2.41x | 1.93 to 2.97 | Above |
The twenty roads with the most motorcyclist casualties, 2021 to 2025, each tested against the national average of 2,170 KSI per billion vehicle miles. Confidence intervals are exact Poisson intervals on the casualty count.
A road whose interval crosses the national average is not an average road, and not a safe one. It means five years of casualties on that road are too few to tell the difference either way.
The Department for Transport publishes two things separately. STATS19 records every road casualty reported to the police, including which road it happened on. The AADF traffic counts record how much traffic uses each stretch of road, broken out by vehicle type, including a column for two wheeled motor vehicles and one for pedal cycles.
Put them together and you get a rate: casualties per billion vehicle miles ridden on that road. Nobody publishes this for motorcyclists, so we have.
99.9% of motorcyclist casualties on major numbered roads matched to traffic data for the same road and year.
Britain already has a road risk map. The Road Safety Foundation publishes crash risk mapping for the motorway and A road network, banding each route both by crash density, meaning crashes per kilometre of road, and by crash rate, meaning crashes per vehicle-kilometre. That network is less than 15% of Britain's road length but carries around 60% of fatal crashes and over 40% of serious ones. Its most recent published results cover 2021 to 2023.
So exposure-adjusted road risk is not new, and we are not claiming it is. What is different here is the denominator. Crash rate per vehicle-kilometre measures every vehicle on the road. This page divides motorcyclist casualties by motorcycle traffic, and cyclist casualties by pedal cycle traffic, so each group is measured against its own exposure rather than against the traffic as a whole. A road can be unremarkable per vehicle-kilometre and still be high per mile ridden.
Two smaller differences. Risk mapping works in road sections; this works in whole numbered roads, 1,710 of them. And these figures run to 2025, the final release.
Click a column heading to sort. The table scrolls sideways on smaller screens.
| 1 | A29 | 42 | 0.006 | 7,446 | 5,367 to 10,065 | 3.43x | 54 |
| 2 | A57 | 87 | 0.017 | 5,229 | 4,188 to 6,450 | 2.41x | 20 |
| 3 | A259 | 181 | 0.039 | 4,691 | 4,032 to 5,426 | 2.16x | 2 |
| 4 | A85 | 34 | 0.006 | 5,684 | 3,936 to 7,942 | 2.62x | 69 |
| 5 | A62 | 30 | 0.006 | 5,267 | 3,554 to 7,519 | 2.43x | 87 |
| 6 | A272 | 59 | 0.014 | 4,356 | 3,316 to 5,619 | 2.01x | 34 |
| 7 | A44 | 70 | 0.017 | 4,077 | 3,178 to 5,151 | 1.88x | 26 |
| 8 | A18 | 36 | 0.009 | 4,198 | 2,940 to 5,812 | 1.93x | 66 |
| 9 | A32 | 46 | 0.012 | 3,984 | 2,917 to 5,314 | 1.84x | 45 |
| 10 | A35 | 101 | 0.029 | 3,522 | 2,869 to 4,280 | 1.62x | 15 |
| 11 | A6 | 159 | 0.049 | 3,278 | 2,788 to 3,829 | 1.51x | 4 |
| 12 | A26 | 40 | 0.010 | 3,885 | 2,775 to 5,290 | 1.79x | 59 |
| 13 | A16 | 39 | 0.010 | 3,876 | 2,756 to 5,298 | 1.79x | 60 |
| 14 | A25 | 44 | 0.012 | 3,788 | 2,752 to 5,085 | 1.75x | 50 |
| 15 | A149 | 40 | 0.011 | 3,789 | 2,707 to 5,159 | 1.75x | 58 |
| 16 | A215 | 44 | 0.012 | 3,653 | 2,654 to 4,903 | 1.68x | 48 |
| 17 | A339 | 30 | 0.008 | 3,930 | 2,652 to 5,610 | 1.81x | 90 |
| 18 | A124 | 30 | 0.008 | 3,915 | 2,641 to 5,589 | 1.80x | 86 |
| 19 | A58 | 41 | 0.011 | 3,672 | 2,635 to 4,981 | 1.69x | 55 |
| 20 | A270 | 30 | 0.008 | 3,839 | 2,590 to 5,480 | 1.77x | 85 |
| 21 | A458 | 32 | 0.009 | 3,728 | 2,550 to 5,263 | 1.72x | 76 |
| 22 | A22 | 48 | 0.014 | 3,442 | 2,538 to 4,564 | 1.59x | 43 |
| 23 | A487 | 41 | 0.012 | 3,519 | 2,526 to 4,775 | 1.62x | 56 |
| 24 | A23 | 148 | 0.052 | 2,864 | 2,422 to 3,365 | 1.32x | 8 |
| 25 | A228 | 30 | 0.008 | 3,582 | 2,417 to 5,113 | 1.65x | 89 |
| 26 | A21 | 68 | 0.023 | 2,911 | 2,261 to 3,690 | 1.34x | 27 |
| 27 | A350 | 42 | 0.013 | 3,131 | 2,256 to 4,232 | 1.44x | 53 |
| 28 | A505 | 35 | 0.011 | 3,186 | 2,219 to 4,431 | 1.47x | 67 |
| 29 | A165 | 32 | 0.010 | 3,233 | 2,212 to 4,565 | 1.49x | 77 |
| 30 | A61 | 62 | 0.022 | 2,797 | 2,145 to 3,586 | 1.29x | 31 |
| 31 | A28 | 33 | 0.011 | 3,110 | 2,141 to 4,368 | 1.43x | 75 |
| 32 | A104 | 36 | 0.012 | 3,049 | 2,136 to 4,222 | 1.41x | 63 |
| 33 | A370 | 32 | 0.010 | 3,113 | 2,129 to 4,394 | 1.43x | 81 |
| 34 | A329 | 34 | 0.011 | 3,002 | 2,079 to 4,195 | 1.38x | 68 |
| 35 | A361 | 57 | 0.021 | 2,720 | 2,060 to 3,524 | 1.25x | 35 |
| 36 | A27 | 130 | 0.054 | 2,419 | 2,021 to 2,872 | 1.11x | 10 |
| 37 | A414 | 46 | 0.017 | 2,745 | 2,010 to 3,662 | 1.26x | 46 |
| 38 | A59 | 50 | 0.018 | 2,704 | 2,007 to 3,564 | 1.25x | 42 |
| 39 | A112 | 31 | 0.011 | 2,927 | 1,989 to 4,155 | 1.35x | 83 |
| 40 | A36 | 36 | 0.013 | 2,838 | 1,988 to 3,929 | 1.31x | 65 |
Every road here carries at least 30 killed or seriously injured casualties, so these are not small number flukes, and the interval shown is an exact Poisson 95% interval on the casualty count. Roads are ranked on the bottom of that interval rather than the headline figure, so a road cannot reach the top on a lucky run of bad years.
Click a column heading to sort. The table scrolls sideways on smaller screens.
| 1 | A38 | 183 | 1,858 | 0.86x | 1,116 |
| 2 | A259 | 181 | 4,691 | 2.16x | 366 |
| 3 | A40 | 162 | 1,432 | 0.66x | 1,247 |
| 4 | A6 | 159 | 3,278 | 1.51x | 631 |
| 5 | A4 | 156 | 1,889 | 0.87x | 1,104 |
| 6 | A5 | 151 | 2,337 | 1.08x | 944 |
| 7 | A2 | 150 | 1,820 | 0.84x | 1,131 |
| 8 | A23 | 148 | 2,864 | 1.32x | 755 |
| 9 | A13 | 131 | 1,595 | 0.74x | 1,191 |
| 10 | A27 | 130 | 2,419 | 1.11x | 909 |
| 11 | A3 | 128 | 1,316 | 0.61x | 1,288 |
| 12 | A406 | 111 | 1,531 | 0.71x | 1,212 |
| 13 | A41 | 110 | 1,935 | 0.89x | 1,085 |
| 14 | A30 | 105 | 1,840 | 0.85x | 1,123 |
| 15 | A35 | 101 | 3,522 | 1.62x | 574 |
| 16 | A1 | 98 | 1,136 | 0.52x | 1,338 |
| 17 | A10 | 91 | 1,992 | 0.92x | 1,061 |
| 18 | A24 | 89 | 2,342 | 1.08x | 942 |
| 19 | A12 | 88 | 1,117 | 0.51x | 1,342 |
| 20 | A57 | 87 | 5,229 | 2.41x | 308 |
| 21 | A47 | 82 | 2,299 | 1.06x | 953 |
| 22 | A205 | 76 | 2,455 | 1.13x | 896 |
| 23 | A20 | 73 | 1,987 | 0.92x | 1,063 |
| 24 | A82 | 71 | 1,342 | 0.62x | 1,277 |
| 25 | A46 | 70 | 1,703 | 0.78x | 1,159 |
| 26 | A44 | 70 | 4,077 | 1.88x | 467 |
| 27 | A21 | 68 | 2,911 | 1.34x | 742 |
| 28 | M25 | 67 | 737 | 0.34x | 1,418 |
| 29 | A39 | 65 | 2,283 | 1.05x | 958 |
| 30 | A34 | 64 | 1,445 | 0.67x | 1,241 |
| 31 | A61 | 62 | 2,797 | 1.29x | 773 |
| 32 | A404 | 61 | 2,411 | 1.11x | 913 |
| 33 | A11 | 59 | 2,380 | 1.10x | 925 |
| 34 | A272 | 59 | 4,356 | 2.01x | 427 |
| 35 | A361 | 57 | 2,720 | 1.25x | 795 |
| 36 | A503 | 54 | 2,588 | 1.19x | 847 |
| 37 | A52 | 54 | 2,411 | 1.11x | 914 |
| 38 | A470 | 53 | 2,118 | 0.98x | 1,012 |
| 39 | A49 | 52 | 2,394 | 1.10x | 920 |
| 40 | A217 | 51 | 1,888 | 0.87x | 1,105 |
This is the same data seen the other way round. The A3, the A12 and the A1 all appear on danger lists every year. Per mile ridden they run at about half the national average.
This is the part that matters to anyone reading this on a bike.
The A57 over Snake Pass carries 5,229 motorcyclist KSI casualties per billion miles ridden, 2.41 times the national average, from 87 killed or seriously injured in five years. It is 308th of 1,710 by rate and 20th by count, which is unusual: most roads are high on one list or the other, not both.
The A686 over Hartside runs at 12,109 per billion miles, 5.58 times the national average, though on 24 casualties rather than 87, so the interval is wide and it sits outside our main table.
Not every famous road turns out to be dangerous, and the honest answer is often that we cannot tell. The A537 Cat and Fiddle records 6 riders killed or seriously injured in five years, which works out at 0.72 times the national average per mile ridden. But 6 casualties is a small number, and the 95% confidence interval runs from 0.26 to 1.57 times average. That interval includes the national average, so the fair conclusion is that the A537 is not distinguishable from an average road, not that it is safer than one. The roads people ride for pleasure are often quiet, and quiet roads produce small samples.
We measured it. For every point on a road's mapped centreline we took the circumradius of that point and its two neighbours, which gives the radius of the bend in metres. A straight gives a huge radius, a hairpin a small one. The score below sums the length of road divided by the tightness of its bends, so a long run of tight corners scores high and a single roundabout does not.
The measure checks out against roads you already know. The straightest roads in Britain come out as the M5, the M1 and the M6, at 0.15 to 0.19. The bendiest rural A roads run about 2.2.
| Rural roads by curviness | Bends per km | Median casualty rate per billion miles | Range |
|---|---|---|---|
| Straightest quarter | 0.15 to 0.86 | 796 | 346 to 1,703 |
| Second | 0.88 to 1.40 | 2,208 | 1,342 to 3,930 |
| Third | 1.40 to 1.63 | 3,551 | 1,813 to 7,446 |
| Bendiest quarter | 1.67 to 2.16 | 3,429 | 2,283 to 5,684 |
Split Britain's rural A roads into four groups by how bendy they are and the casualty rate climbs with every step, from 796 casualties killed or seriously injured per billion miles in the straightest quarter to 3,429 in the bendiest. The top two groups are close, so the effect flattens off rather than running away.
Across rural roads the relationship between curviness and casualty rate is a correlation of 0.73. On urban roads it is 0.31 and much less certain, which is what you would expect: a city A road looks bendy to this measure because it weaves around junctions, not because it has corners worth riding.
The rural correlation is 0.73 across 40 roads, with a 95% confidence interval of 0.54 to 0.85, so the relationship is well clear of zero. Ranking the roads rather than using their raw values gives 0.77, which means it is not being driven by one or two extreme roads. The urban figure is weaker than it looks: 0.31 across 50 roads, with a confidence interval running from 0.04 to 0.54. That very nearly touches zero, so treat it as barely distinguishable from no relationship at all rather than as a smaller version of the rural effect.
The group medians rest on ten roads each, and their ranges overlap: the third group runs from 1,813 to 7,446 and the bendiest from 2,283 to 5,684. The direction of the effect is clear. The size of any single step between groups is not.
Click a column heading to sort. The table scrolls sideways on smaller screens.
| A39 | 204 | 3,357 | 313 m | 2.16 | 2,283 | 1.05x |
| A25 | 52 | 590 | 340 m | 2.12 | 3,788 | 1.75x |
| A28 | 64 | 779 | 329 m | 1.90 | 3,110 | 1.43x |
| A350 | 74 | 1,292 | 339 m | 1.88 | 3,131 | 1.44x |
| A458 | 89 | 909 | 282 m | 1.83 | 3,728 | 1.72x |
| A370 | 34 | 411 | 301 m | 1.82 | 3,113 | 1.43x |
| A85 | 97 | 772 | 320 m | 1.80 | 5,684 | 2.62x |
| A44 | 149 | 1,107 | 373 m | 1.75 | 4,077 | 1.88x |
| A272 | 80 | 762 | 314 m | 1.72 | 4,356 | 2.01x |
| A414 | 87 | 864 | 367 m | 1.67 | 2,745 | 1.27x |
| A228 | 42 | 387 | 413 m | 1.63 | 3,582 | 1.65x |
| A29 | 36 | 442 | 266 m | 1.55 | 7,446 | 3.43x |
| A49 | 153 | 1,339 | 360 m | 1.53 | 2,394 | 1.10x |
| A487 | 158 | 1,012 | 381 m | 1.52 | 3,519 | 1.62x |
| A483 | 153 | 1,026 | 357 m | 1.48 | 2,626 | 1.21x |
| A18 | 68 | 536 | 364 m | 1.47 | 4,198 | 1.93x |
| A149 | 90 | 617 | 369 m | 1.46 | 3,789 | 1.75x |
| A48 | 157 | 1,023 | 431 m | 1.41 | 1,813 | 0.84x |
| A16 | 82 | 791 | 370 m | 1.41 | 3,876 | 1.79x |
| A82 | 185 | 1,137 | 390 m | 1.40 | 1,342 | 0.62x |
| A36 | 74 | 623 | 427 m | 1.40 | 2,838 | 1.31x |
| A361 | 184 | 1,689 | 378 m | 1.39 | 2,720 | 1.25x |
| A339 | 40 | 276 | 459 m | 1.39 | 3,930 | 1.81x |
| A15 | 116 | 1,026 | 332 m | 1.37 | 1,896 | 0.87x |
| A470 | 207 | 988 | 440 m | 1.30 | 2,118 | 0.98x |
These are the roads riders travel for, and most of them carry an above average casualty rate. That is the honest shape of the thing: the feature that makes a road worth riding is the same feature that shows up in the casualty record.
A correlation is not a cause. Bendier roads are also more rural, often faster, and attract riders who came for the road rather than to get to work. Any of those could be doing the work. What the number does say is that the roads riders pick out are not a random sample of Britain's roads, and the casualty record follows them.
Bend geometry comes from OpenStreetMap contributors under the Open Database Licence, restricted to Great Britain. It measures the road as mapped, so a road that has been realigned since the map was last edited will be measured as mapped rather than as built.
Click a column heading to sort. The table scrolls sideways on smaller screens.
| A23 | 102 | 0.042 | 2,438 | 1,988 to 2,960 | 1.01x |
| A5 | 60 | 0.023 | 2,590 | 1,976 to 3,334 | 1.07x |
| A10 | 105 | 0.040 | 2,625 | 2,147 to 3,178 | 1.08x |
| A3220 | 40 | 0.015 | 2,705 | 1,932 to 3,683 | 1.12x |
| A315 | 88 | 0.032 | 2,724 | 2,185 to 3,356 | 1.12x |
| A219 | 55 | 0.020 | 2,774 | 2,090 to 3,611 | 1.14x |
| A105 | 42 | 0.015 | 2,778 | 2,002 to 3,755 | 1.15x |
| A217 | 48 | 0.017 | 2,840 | 2,094 to 3,766 | 1.17x |
| A202 | 72 | 0.025 | 2,862 | 2,239 to 3,604 | 1.18x |
| A41 | 44 | 0.015 | 2,879 | 2,092 to 3,865 | 1.19x |
| A6 | 97 | 0.033 | 2,920 | 2,368 to 3,562 | 1.20x |
| A107 | 37 | 0.012 | 3,042 | 2,142 to 4,193 | 1.26x |
| A27 | 40 | 0.013 | 3,045 | 2,175 to 4,146 | 1.26x |
| A35 | 33 | 0.010 | 3,195 | 2,199 to 4,487 | 1.32x |
| A24 | 95 | 0.028 | 3,363 | 2,721 to 4,112 | 1.39x |
| A58 | 32 | 0.009 | 3,389 | 2,318 to 4,784 | 1.40x |
| A2 | 71 | 0.020 | 3,534 | 2,760 to 4,458 | 1.46x |
| A61 | 50 | 0.014 | 3,684 | 2,734 to 4,857 | 1.52x |
| A308 | 85 | 0.022 | 3,849 | 3,075 to 4,759 | 1.59x |
| A404 | 40 | 0.010 | 3,950 | 2,822 to 5,379 | 1.63x |
| A205 | 84 | 0.021 | 3,997 | 3,188 to 4,949 | 1.65x |
| A46 | 33 | 0.008 | 4,320 | 2,974 to 6,067 | 1.78x |
| A214 | 31 | 0.007 | 4,340 | 2,949 to 6,160 | 1.79x |
| A11 | 103 | 0.022 | 4,708 | 3,843 to 5,710 | 1.94x |
| A21 | 41 | 0.008 | 4,983 | 3,576 to 6,760 | 2.06x |
The national figure for cyclists on major roads is 2,423 KSI casualties per billion cycle miles. The worst roads for cyclists are a different set from the worst roads for motorcyclists, and they cluster around London and the commuter routes into it.
A rate is not a verdict on a road. Three things could each move these numbers and we cannot separate them: how the road is built, how fast people ride on it, and who rides there.
The traffic figures are annual average daily flows. They tell us how much riding happens on a road across a year, not what the traffic was on the day of any particular collision.
Casualties are assigned to the first road recorded for the collision, so a crash at a junction between two roads counts against one of them.
Only motorways and A roads can be rated. The Department for Transport publishes minor road count points as a rotating statistical sample without link lengths, so no mileage can be derived for B, C and unclassified roads. They are excluded here, and a lot of riding happens on them.
Police data undercounts injured riders, and undercounts injured cyclists more. Every count here is a floor.
Numerator: DfT STATS19 casualty and collision files for 2021 to 2025, casualties assigned to the first road of the collision. Motorcyclist means casualty_type 2, 3, 4, 5, 23 or 97, which covers riders and passengers of any engine size. Cyclist means casualty_type 1.
Denominator: DfT AADF traffic counts for the same five years. Vehicle miles for a road in a year is the sum, over every count point on that road, of the annual average daily flow multiplied by the link length in miles multiplied by 365. The two wheeled motor vehicles column is used for motorcyclists and the pedal cycles column for cyclists.
As a check on the denominator, the same method applied to all motor vehicles gives 223 billion vehicle miles on Britain's major roads in 2025, which is close to two thirds of the national all roads total, the split the Department for Transport reports. Summed link lengths also reproduce real road lengths: 296 miles for the A38, 118 for the A57, 274 for the A9.
Intervals are exact Poisson 95% intervals on the casualty count.
Published under the Open Government Licence v3.0. Contains public sector information licensed from the Department for Transport.
Cite as: Bikers.co.uk (2026), Motorcyclist and cyclist casualty rates per billion vehicle miles by road, Great Britain 2021 to 2025. Derived from DfT STATS19 and DfT AADF traffic counts.
It depends what you measure. By number of casualties it is the A38, with 183 motorcyclists killed or seriously injured between 2021 and 2025. Measured against how far people actually ride, the A38 is below average and the highest rate among roads with at least 30 casualties is the A29 in Sussex, at 7,446 KSI casualties per billion vehicle miles against a national average of 2,170.
It has the most motorcyclist casualties, because it is 296 miles long and carries about 99 million motorcycle miles a year. Per mile ridden it runs at 0.86 times the national average and ranks 1,116th of 1,710 roads.
Per mile ridden, yes. The A57 records 5,229 motorcyclist KSI casualties per billion vehicle miles, 2.41 times the national average, from 87 killed or seriously injured over five years.
Because a long, busy road gives more opportunities for something to go wrong. Counting casualties measures exposure as much as danger. Dividing by miles ridden removes that, and the ranking changes: only four of the top twenty roads by count are also in the top twenty by rate.
The Department for Transport, in two parts. STATS19 for the casualties and the AADF traffic counts for the traffic. Both are open data under the Open Government Licence.
Among rural A roads with at least 30 casualties, the A39 Atlantic Highway, with 3,357 bends tighter than a 175 metre radius across 204 miles. The straightest roads in Britain are the motorways: the M5 scores 0.15 bends per km against the A39's 2.16.
