There have been countless discussions about whether and when a recumbent bike is faster than standard diamond frame. Conclusions vary, but I have never seen anyone to compare comparable bikes ridden by the same rider. So I decided to try it myself.
I used what was at hand: Favorit road bike and Python lowracer, both equipped with similar components and SPD clipless pedals. The Favorit is slightly handicapped by a hub dynamo (lights turned off during the tests) which may suck something around 1 or 2 watts. Python loses around 1.5 % of power in pedal steering because legs can't push directly forward (that percent corresponds to cos 10° which is my guess of an angle between pedals, steering axis and bottom bracket centre point), classic 'bents don't have this disadvantage. Let's say mechanical losses are more or less the same.
By the time of the experiment (summer 2014) I had clocked over 3000 km on the Python, so I'm pretty well used to the horizontal pedaling and leg steering. I had even larger mileage on various upright bikes I ride almost daily. I used the same pedaling style on both bikes: high cadence circular spinning, no standing up. Length of the test track was way beyond the limits of anaerobic sprint, so the maximum power was limited quite precisely by the lungs.
Empty Favorit weighs 15 kg, Python weighs 21. To avoid comparing apples with oranges, I added three two-litre bottles of water to Favorit's rack. With my 69 kilograms, emergency tool set, half litre of drinking water, notebook and pencil, both rigs ended up at the same 92 kg.
A three-kilometre section of main road between Dobroměřice and Most was used as the calibration hill: up, down. Straight road covered with smooth asphalt without any potholes or debris, the only bumps are bridge dilation gaps at the very beginning and end. The shoulder is comfortably wide, so it is possible to ignore cars and fully concentrate on pedaling. Total altitude gain is somewhere between 40 and 60 metres depending on which map you believe more. Steepness of the hill varies, but it never reaches extreme "granny gear" values.
It would be ideal to ride in windless conditions, but that's hard to achieve, so at least I rode there and back again to make wind direction more easily observable in the records. The downhill results are not very sound: first, the road bike doesn't have enough gears to pedal at 50 km/h; second, I was usually quite exhausted after the uphill ride and the power was everything but constant and sustainable on the way back. But approximate wind direction should be clearly visible. I checked it by a wind sock on the nearest hilltop, the strength is estimated.
I performed ten rides, the procedure was the same every time: several kilometres to warm up, stopping and resetting the speedometer under the hill, full-power uphill sprint, catching breath and recording measured values, crossing the road, resetting the speedo, remaining-full-power downhill sprint, catching breath and recording values again, going home.
Source data here (XLS)
The test rides are plotted in chronologic order on the horizontal axis. Rides 1 and 2 were just a light exercise, without having to open my mouth to breathe. All others were done at maximum power, with several minutes of panting at the top. Ride 6 came the very next day after ride 5 (bad idea, my legs kept aching for a week after that), all others were done after at least one day pause. I was switching the bikes throughout the whole experiment.
Average speeds of each ride are plotted on the vertical axis. To eliminate speedometer errors (they measured the track length with a 60 m difference), I calculated the speed by dividing elapsed time (measured with equal precision by both speedos) by average track length of exactly 3 km. There were no instrument failures, measured distance was the same every time. Average uphill and downhill speeds are plotted in solid lines, average of these two values is dashed.
Black arrows indicate wind direction (arrow down = headwind when riding uphill). Bold arrows mean stronger wind than usual. Mostly the wind was blowing obliquely from the left. As you can see on the chart, I got most headwinds laying down and the only stronger tailwind bumped the lowracer's speed up by some 5 km/h. I met no strong headwind standing up and the differences in tailwind strength only affected the downhills. Several more rides would be needed to make the results more statistically solid, but I don't have time and mood for that right now.
Without wattmeter, I can't tell exactly how the riding position affects rider's power output. Subjectively I like the laid back position better - breathing is easier and my head doesn't lack blood flow. But different people may have different preferences.
On really steep hills where the ride is not much faster than a walk, aerodynamic drag becomes negligible and I'm slower on the Python than on any upright bike with similar gears. Most probably because slow speed requires large steering inputs. Front half of a Python including rider's legs weighs dozens of kilograms and throwing such a mass left and right irregularly is a wasted hard work that doesn't propel you forward. I haven't yet tried how hand-steered recumbents behave in this mode - they shouldn't have this problem. Anyway, slow speeds are not a topic of this article.
If power, wind and weight are the same, aerodynamics make all the difference. At speeds from cca 25 km/h up, a lowracer is always faster, regardless of how steep the hill is.