How long does it take to learn riding?

Deze pagina in het Nederlands


Every unicyclist probably gets asked many times "How long does it take to learn riding?" The shortest correct answer is "It depends". Age is often being thought of as a prime discriminating factor. But now we know more...

In March 2003 someone posting in by the handle of "wobbling bear" suggested that we should study how the learning time depended on age. I took up this suggestion and posted a poll. Initially I asked people to supply:
A. age they started learning;
B. number of hours spent before they could ride 50 metres;
C. male or female

There was a massive response: in about a week I collected data from 66 people. Mostly via the newsgroup, some by e-mail, private message or IRL communication. Many people volunteered additional data, notably:
D. how many calender days it took to learn (57 entries);
E. what sort of instruction they had, if any (21 entries);
F. wheel size they learned on (35 entries).
These additional data were quite welcome. The reason I asked for only the bare essentials (A, B and C) was that I didn't want to shy people away from responding.

I processed in this first batch all of the data received before 6 April 2003 09:00 am CET. I plan to process data that may be coming in at a later stage in subsequent analysis, so keep them coming.

The raw data are presented in the table below, in the order in which they appear in my news reading software (Free Agent 1.21/32). I have assumed that people who responded in implicitly agreed to republication on this website, from the other people in the table I have requested permission to publish - all agreed. 

Some responses were ambiguous in one way or the other, but you can check in the table how I reduced your data to simple, 'hard' facts. Data that are too 'vague' to be used are printed in a red font - those individuals unfortunately had to be excluded from the analysis. In addition to raw data, I also included in the table (if available):
* calculated hours/day;
* "Quality of instruction" which I estimated from your words;
* other relevant remarks.

Note that some of the numbers in the table appear rounded. However, in the analysis I used all available decimal places.

name, handle or other ID age when started hours to learn spread over days hours/ day male/ female instruction? Q of instr wheel size relevant remarks
Cokerhead 11 30.00 120 0.25 male        
nikkifrog 13 20.00 90 0.22 female on my own 0.0 20  
harper 11 12.00 3 4.00 male        
JJuggle 15.5 30.00 14 2.14 male no instructor 0.0 24  
Rockey 16 8.50 14 0.61 male        
friend of harper 12 8.00 14 0.57 male        
another friend of harper 20 12.00     male        
son of harper 11 18.00 21 0.86 male        
colleague of harper 28 no good estimate 90   female        
Mattvl 20 8.00 1 8.00 male     24  
Sofa 25 6.00 17 0.35 male     20 only practiced freemounting
unimoron 12 10.00 5 2.00 male     20  
leo (Leo Vandewoestijne) 19 20.00 7 2.86 male     24  
Mikefule 24 5.00 7 0.71 male        
dustin 10 21.00 14 1.50 male self-taught with book 0.5 20  
m_extreme_uni 14 2.00     male        
jerryg 8 150.00 90 1.67 male        
Erin 47 15.00 14 1.07 female newsgroup to give tips 0.5 24  
Jaxer103 11 unclear     male        
Jayjae13 14 14.00 14 1.00 male        
JonM 22 8.00 17 0.47 male        
wobbling bear 54 60.00 120 0.50 male        
djm 27 0.17     male     20  
Krashin'Kenny 49 9.00 9 1.00 male     20  
ache too complex to quantify too complex to quantify too complex to quantify   male        
U-Turn 43 22.00 21 1.05 male        
Jason (KJP) 30 2.92 7 0.42 male        
Jason from Alaska 35 9.00 9 1.00 male     24  
daughter of Jason from Alaska 11 11.00 11 1.00 female     20  
son of Jason from Alaska aka King MUni-Man 5 16.00 12 1.33 male     16  
bbraf 28 21.00 21 1.00 male     20  
bbraf's older son 6 10.50 7 1.50 male     16 age was stated as 6.5
bbraf's daughter 9 10.71 75 0.14 female     20  
bbraf's wife 38 10.71 75 0.14 female     20  
lozza 13 12.00 3 4.00 female        
manic_mark 15 6.00 7 0.86 male        
iunicycle 41 10.00     male few tips from daughter 0.5    
Zook 53 42.00 42 1.00 male self-taught 0.0 20  
Doug Massey 11 30.00 30 1.00 male no instructor 0.0 26  
Carley Massey 12 15.00 90 0.17 female instructor 1.0 20  
Andrew Carter 16 2.50     male no instructor 0.0 20  
zod 29 9.00 14 0.64 male no instructor 0.0 20  
UniBrier 16 14.00 14 1.00 male        
eldest daughter of UniBrier 9 11.00 12 0.92 female        
neighbourhood friend of eldest daughter of UniBrier 8 9.00 12 0.75 female        
youngest daughter of UniBrier 5 5.00 45 0.11 female        
Byron, college friend of UniBrier 20 10.50 14 0.75 male        
fme 42 6.00 14 0.43 male no instructor but the WEB 0.5 20  
uniikki 38 45.00 30 1.50 male        
gizmoduck 22 17.50     male no instructor 0.0 20  
Ground-rider 33 20.00 7 2.86 male     20 included freemount
greyhairedguy 38 23.00 45 0.51 male        
hubble702 11 30.00 14 2.14 male taught myself 0.0 16 bad uni; modern uni would learn quicker
Daniel Wolfe 16 140.00 60 2.33 male        
Mika Holkeri 39 10.00 10 1.00 male     26  
swim learning 42 30.00 90 0.33 male        
sarah 19 10.00     female     24 grass playing field
Janne Himanka 35 7.00 3 2.33 male        
Liam 16 4.50 31.5 0.14 male instructor 1.0    
spider 38 5.00 10 0.50 male Self taught   20  
Klaas Bil 47 20.00 35 0.57 male instructor 1.0 24  
eldest daughter of Klaas Bil 11 25.00 67 0.37 female non-unirider as instructor 0.5 20  
youngest daughter of Klaas Bil 9 15.00 45 0.33 female instructor 1.0 20  
colleague of Klaas Bil 36 3.50 21 0.17 male instructor 1.0 24  
friend of colleague of Klaas Bil 38 10.00 70 0.14 male instructor 1.0 24  
Uniwitold 64 21.00 42 0.50 male self educated 0.0 16  


In the figure below, I have plotted the learning time versus age for all respondents except the few that were unusable. For some people, both age and hours fully coincided. For those, I have shifted some points in the plot over a distance of 0.3 years (+ or -) from their true age. (I know that this is unconventional practice but I wanted all datapoints to be visible. Note that for all calculations, true ages were used.) 

The blue line in the picture is a 'typical' fit, based on another bit of unconventionality involving weighted median values of logaritmically distributed bins. (I have favoured this approach over e.g. a least squares fit of some sort, to lessen the influence of the few people who required very long or short times to learn. I think such cases could have appeared at any age.) 

The main trend is that older people take longer to learn, which confirms our common assumption. 
Other observations: 
1. there is no evidence of a 'minimum age' to learn. I would love to have data for children up to five to see at what age that statement breaks down.
2. Somewhat surprising is the 'hump' around 10 years followed by an 'optimum' learning age up to at least 40. 

Just for fun I include the age distribution histogram of the people who responded (or who was responded for). The 'hump' around 10-15 years of age is remarkable now that we know that this is a slightly difficult age to learn. I guess unicyclists are challenge seekers after all. Also remarkable is the concentration of learners around 35-40 years. Maybe these are the parents of the 10-15 year olds?

OK, so now we know that age is an important determinant of learning time even though the effect may be different from what most expected. But what else matters? Well, simply everything. In order of importance:


  • The 10 people that mentioned they'd had no instruction at all, took on average 36% more time to learn than average (corrected for age only).
  • The 5 people who mentioned a limited form of instruction (such as the web or a book) took 2% less time than average.
  • The 6 people who stated they had an instructor took 19% less time than average. 

Moral: find an instructor.

Wheel size

The people in the database learned on wheels ranging from 16" to 26". Adults (defined as age 16 and up) on a 16" or 20" (13 in total) typically learned 30% quicker than adults on a 24" or 26" (9 in total). (Based on learning times corrected for age and quality of instruction.) 

Moral: pick a small wheel to learn on.

Note that there were no small children in the dataset who learned on an unreasonably large wheel (or so I think). The youngest learner on a 20" was 9 years old, the youngest on a 24" or 26" was 11 years old. C'mon parents or club leaders, who's gonna put his learning child on a Coker from day 1, for statistics' sake? And did any adults learn on 28" or larger wheels, besides Scott Kurland (of whom I have no data)?


There was an appreciable difference in expected learning time for males (50) and females (13). The (median) expected learning time of females was 18% quicker than that of males. (Based on learning times corrected for age, quality of instruction and wheel size.) 

I'm hesitant to give you a moral for this one...

Hours spent per day

Time spent per day on learning varied from about 7 minutes to 8 hours. One might expect an effect of that on learning but the data (note that both scales are logarithmic) look like a amorphous cloud. Yet I had the nerve to throw a linear fit through the data (in the log-log domain) and use that as a basis for correction. 

I hate to say it but the more you practice per day, the more hours you will spend in total. If you double your daily practice, you will spend about 13% more total hours. Bad news huh? Well, the good news is that the data on this particular issue are so noisy that the effect is probably not significant anyway. 

And moral? The more time you spend per day, the more fun you have!

OK, this was all boring but now the interesting stuff. Find out how long YOU will need to practice!

Based on the described relationships I've built a numerical model that can predict the typical learning time for any individual. It is reasonably accurate: for about 60% of the people, their actual learning time will be less than 50% different from the prediction, while only 14% will need more than twice the practice predicted by the model. The uncertainty stems from things like previous experience, determination, physical condition, talent and attitude; such factors are difficult to quantify and have not been included in the model.

To use the model, download the spreadsheet hour_est.xls. Don't left-click to open in your browser, but right-click, Save As and then open the saved file. You need Excel 4.0 or higher to open it. Answer five simple questions and hey presto! The spreadsheet returns your estimated time to learn and ride 50 metres on a unicycle. You will also get a probability histogram so that you can estimate how much uncertainty there is in the predicted number of hours.

Oh and if you think that your estimated time is discouragingly long, you could have a look at this discussion in, where the experiences of very fast learners are discussed. I will not include such data in this analysis though, because they are biased. 

Get your talent score!

Can you ride already? Then download the Unicycling Talent-o-meter from this page and get your personal unicycling talent score! 

(Bonus: that spreadsheet makes it also very convenient to e-mail me your personal learning data for inclusion in a future update of this page, if you desire to contribute to science!) 

Back to Unicycling statistics.