The Speed of CW

Hi all,
while I was playing with some analytical tools, I looked into the raw data provided by the Reverse Beacon Network.

As a first result, please find a breakdown of CW speed by continents. You can see that only 25 % of all CQ calls spotted around the globe are using less than 22 - 25 wpm. The median is 26 - 28 wpm. So if you master 26 … 28 wpm, you can work at least have of the stations calling CQ.

The data is the full set of RBN spots from January 1 - March 12, 2018. Note that the distribution properties are based on spots, not stations, so there is a slight bias towards stations that are spotted by more skimmers than others.


73 de Martin, DK3IT


Hi Martin,
Did you gather spots data from all amateur bands or just one single band?

Contest season, hence the higher speeds?

Hi Guru,
this is across all bands for the last three months.

Here is a breakdown by bands:

73 de Martin, DK3IT

1 Like

Exactly, the higher speeds correspond to the big contest HF bands 10, 15, 20, 40, 80, 160.
The others have slower mean values.

73 and GN (off to sleep now)


I am not so sure that the contest season has that much impact. Note that the median (50%) is much more telling than the mean - the mean is heavily biased by the extremes.

Attached, please find diagrams that show the CW speeds on 40m vs. 30m as a combination of a kernel density estimation and a histogram.

73 es gn de Martin, DK3IT



1 Like

And here are a few boxplots that likely better illustrate the CW speeds found in spots, grouped by DX continent and by band.

73 es gn de Martin, DK3IT

Good analysis and graphs! I was wondering why no data for NA continent? Did I misinterpret?

I have to investigate that. It is either a problem in the raw RBN data, or in the definition of the continent codes.


I did a similar analysis in August 2015, using the RBN data from two days, one mid-week and one on a contest weekend day. The pie-charts weren’t the best of illustrations, but the chart at the end shows the main difference:

This confirms why I no longer have any interest in CW contesting. My usual sending speed on summits is 18 to 20 wpm and that works fine for me. I would rather send the information once at this speed than several times at a faster speed. Of course, sending is one thing, receiving is another. I often find it difficult to copy faster speeds while battling the wind and rain on a summit. For that reason alone it would be an act of desperation if I were to chase contesters in order to qualify the summit or indeed (should vanity prevail) increase the number of contacts in my log.

Thanks for the data Martin. It makes interesting reading.

73, Gerald G4OIG

1 Like

I dial the keyer speed down when I’m on a summit, at least partly in the hope that chasers won’t answer quite so fast… :wink:

if you on a summit sometimes in a noisy enviroument from wind also from QRM its hard to read DX stations.
I also speed down, most to abt 18. I assume that a chaser will - should meet more or less the speed.

In contests same with QRM. the big guns with speeds over 30 - not to be to understand,
Sometimes I listen to one with high speed - he calls and calls several minutes - no answers. If he speeds down to 20-22 then suddently a lot answers. They must be learned :wink:
Not all are profis - we make amateur radio.

Keyer? Oh yes, I wind my wrist down, probably because I can’t feel my fingers and have to resort to straight key “bashing” to produce intelligible morse. Those who work me regularly will know what I mean! Thankfully 99% of chasers have the patience and common sense to match my sending speed, whereas in a contest I have actually been “rejected” because of sending around 20wpm. Thankfully SOTA is not a contest.

1 Like

For very weak signals and/or QRN and/or QRM and/or QSB, QRS (slower morse) is more successful than QRQ (faster). This is a fact and has been clearly demonstrated several times to me in my +33 years working CW.



Hi all,

there was a small bug in the first analysis - “NA” for North America was taken for “NaN” - “not a number”, so all NA spots were discarded.

I also updated the visualization of the WPM to pure histograms with a bin size of 1 wpm.

Here is the definitive data.

73 de Martin, DK3IT




I am not a number! I am a free man!

Sorry I wasn’t able to participate in this, it looks like it was great fun.


1 Like

Interesting data. It might be worth doing a test for Normality to see if std is valid.

On this reflector???



I would say that the histograms above hint that the WPM is actually normally distributed.

No time for a Shapiro-Wilk test right now :wink:

73 de Martin, DK3IT