James Spragg

I hate the clocks going back more than ever…both my kids have woken up at 5:30 every day this week

a cartoon of homer simpson standing in a field ALT: a cartoon of homer simpson standing in a field

Stewards in Mexico…

Russel down to 7th as he got squeezed by Verstappen

Yup - I also never refer to athletes as ‘my athletes’ rather an athlete I work with

100% agree

One phrase that bugs me way more than it should is ‘my PhD student’ or ‘my athlete’

As a coach, supervisor or manager your job is to support people; not to own them or their achievements

Everywhere I have been today in Germany only has baby changing in the women’s toilets - an entire country stuck in the 1950s

It’s really time to start using video drones to capture bike races

Léon the pig

For anyone following the life of Little Léon the Tro Bro piglet, here he is at 5 months 😍

Really hope Valentin can win him a little brother next year 🥺

From instagram

Paris-Chauny has probably been a better watch for the last 50km!

hmm thinking about it... I guess it's not a 'tweet' on bsky?

niche knowledge tweet and I am all for it!

I think what we saw in the women’s race was a lot of politics being played out as tactics

This is basically every bike race to be fair

Bronson’s accent in ‘The Thursday Murder Club’ makes it unwatchable

The best thing about data is it doesn’t give care about your opinion

Consequently - this is also why people don’t like data

Elite sport support is basically taking resources and making them as effective as possible

haha no, it's super simple but works!

just a rolling average across 3s from 2s-15mins

For example, we averaged the power outputs associated with 123, 124, and 125s and plotted them against 124s

So there you have it. This was our proposed solution to interrogate real-world (messy) MMP data, identify maximal values in that data, and use (only) those to derive CP.

We used the smoothing as we found during preliminary analysis that with no smoothing, we identified too many (submaximal) points and with >3s we missed some crucial points.

for example, in this picture, the 'X' represent the shoulders identified by this method - its clear these happen when the curve takes a dive from flat/positive to negative

(nb a couple are missed due to the 3s smoothing we applied)

to identify a shoulder, we looked at changes in the gradient/slope of the PD curve - we applied a 3s smoothing to make things easier and then identified a shoulder as when the gradient goes from flat/positive to negative

therefore the hypothesis was that if we can identify these shoulders and only use them when calculating CP we should get more accurate CP estimates (compared with values from formal testing) - spoiler alert, this works!

When deriving CP from the PD realtionship we want to only use maximal values; otherwise, we include a load of submax values, which pull down our CP estimate

the shoulders represent maximum values; ie the longest duration a given power was sustained for

essentially, when you look at the PD relationship derived from real-world data you don't get a smooth curve; you get parts of the curve where it is flat (same MMP for multiple durations) and parts when it suddenly drops (the point between where its flat and where it drops suddenly are 'shoulders'

Olympic medallist Ben Proud becomes first British athlete to join Enhanced Games The Olympic swimming medallist Ben Proud has become the first British athlete to join the Enhanced Games, an event which allows athletes to take performance-enhancing drugs

Ban him for life from all professional and amateur sport. Make an example of him. This stuff is incredibly dangerous for sport and society. www.theguardian.com/sport/2025/s...

yeah not going to waste a single second more trying to help you out here mate

Academia is bathsit...

Journal editor: Can you review this paper?

Me: No, sorry. No capacity at the moment & I'm not really in academia, so reviewing less & less. Try x,y,z

Journal editor: You failed to properly respond to my email, as you didn't click the link & select a response