Author Topic: [HAMR] Visualizing the OYTT  (Read 221920 times)

hillbilly

Re: Visualizing the OYTT
« Reply #200 on: 14 February, 2015, 08:19:03 am »
That Kurt's 200km rest day looks like a tiny ride compared to the other points illustrates how heroic their daily averages are.  200km is still a bloomin long way.

red marley

Re: Visualizing the OYTT
« Reply #201 on: 15 February, 2015, 07:51:51 am »
Day 45: Kurt wins the Sebring 12hr race as a 'side project' with 253.2 miles before adding another 10 to the OYTT for good measure. Meanwhile Steve knocks out a cool 193 miles to keep his consistent pace. William records his highest daily total so far at 81 miles, but has a deficit of over 7,000 miles to make up.


Wowbagger

  • Former Sylph
    • Stuff mostly about weather
Re: Visualizing the OYTT
« Reply #202 on: 15 February, 2015, 09:51:59 am »
Steve's zig-zag is moving inexorably away from his planned higher schedule. In a few more weeks it will be approaching that magic horizontal line as he starts knocking out >206 mile days, day after day. Can't wait.
Quote from: Dez
It doesn’t matter where you start. Just start.

red marley

Re: Visualizing the OYTT
« Reply #203 on: 16 February, 2015, 06:28:11 am »
Day 46: The day after Kurt's 253 mile Sebring TT, he's back to 200 miles per day apparently without any ill effect from the weekend's exertions. The overall effect including the pre-race 'rest day' was to edge his average downward a little. Steve's trend is to continue to move slowly away from his upper schedule.


Re: Visualizing the OYTT
« Reply #204 on: 16 February, 2015, 11:22:30 am »
How soon can you plot a Polynomial to predict the final outcome?

red marley

Re: Visualizing the OYTT
« Reply #205 on: 16 February, 2015, 12:03:31 pm »
I guess that could have been done from day 3 onwards, but I'm not convinced it would be very useful. There is clearly a big seasonal influence on the pattern, so I think Steve's published schedule and Tommy's actual distances are the best predictors in the medium and long term.

Re: Visualizing the OYTT
« Reply #206 on: 16 February, 2015, 08:18:54 pm »
How soon can you plot a Polynomial to predict the final outcome?

"Predictions are very difficult, especially when they are about the future"

red marley

Re: Visualizing the OYTT
« Reply #207 on: 17 February, 2015, 06:30:26 am »
Day 47: A long day puts Kurt back on track with a cumulative average of 201 miles per day. Steve's 200 mile day remains above his February schedule. Kurt is now one-week's riding behind Steve after starting nine days behind.


Re: Visualizing the OYTT
« Reply #208 on: 17 February, 2015, 02:29:11 pm »
TrackLeaders reckons there’s only 30 miles between Steve and Kurt’s totals. Saturday, they may have the same total.  :-\

red marley

Re: Visualizing the OYTT
« Reply #209 on: 17 February, 2015, 02:45:19 pm »
If it does, then it is clearly wrong. Actual difference is currently 1,172 miles (8818 for Steve vs 7646 miles for Kurt). The SPOT is likely underestimating Steve's rides more than Kurts given that Steve tends to ride wigglier routes than Kurt. Although that wouldn't be enough to account for such a difference. Perhaps it is including Kurt's motorised stages?

Re: Visualizing the OYTT
« Reply #210 on: 17 February, 2015, 10:24:54 pm »
There also seems to be a discrepancy in average speeds on the SPOT site, with Steve's being higher than Kurt's, which is odd given the relative distances covered   ??? ???

Ahhh just noticed Jo's post on the 'tracker' thread  :facepalm:

red marley

Re: Visualizing the OYTT
« Reply #211 on: 18 February, 2015, 06:29:56 am »
Day 48: For four days in a row, Kurt has exceeded 200 miles with an average of 225 miles per day sending him towards that WR pace. Steve's daily average is also up at around 197 mpd. This is a different competition to that a month ago.


HTFB

  • The Monkey and the Plywood Violin
Re: Visualizing the OYTT
« Reply #212 on: 18 February, 2015, 06:41:38 am »
Is it only me who finds the click-and-drag resizing interface of the app utterly baffling?
Not especially helpful or mature

red marley

Re: Visualizing the OYTT
« Reply #213 on: 18 February, 2015, 06:51:57 am »
Is there anything that would make it easier? I'd like to be able to make rescaling as intuitive as possible, so suggestions welcome.

(The intention is that the point you drag stays fixed and the rest of the chart will expand if you drag up or right and contract if you drag down or left. The chart can be panned by dragging either of the axes).

Re: Visualizing the OYTT
« Reply #214 on: 18 February, 2015, 07:51:19 am »
Why do we get lovely smoothed cures one day and yukky zig zags the next?   

Oh, and for a numpty, what exactly does the y axis scale actually mean?   I think the point is that at some point in late October or November Steve's line will burst above x axis indicating that he's surpassed Tommy's annual mileage total but it's sort of an odd representation for somebody who isn't a statto by nature.

Re: Visualizing the OYTT
« Reply #215 on: 18 February, 2015, 08:06:28 am »
As a suggestion, could the 0 Y be marked as "205/day - T Godwin Av" and the top and the Y legend say "Annual miles above/below T Godwin"? (which may help the occasional wandering bear)

Re: Visualizing the OYTT
« Reply #216 on: 18 February, 2015, 08:11:38 am »
Speaking as someone who loves data and graphs, I think it's wonderful!

Looking at the lines you can clearly see how Steve had to scale back his mileage during the cold week that we had (days 31-38ish) but has really increased his mileage in the last couple of weeks, now that the days are getting longer quickly. 

For Kurt, those 3 big days have only pulled him back to where he was v Tommy G before his rest day.  It illustrates what they say about time trialling: that the secret is not to go fast but to not go slow.  Here, it's not the big days that move the line, it's the short days. 

For the mostpart, of course, both riders have been extremely consistent.  That is why little changes such as these stand out.

Re: Visualizing the OYTT
« Reply #217 on: 18 February, 2015, 08:51:22 am »
Why do we get lovely smoothed cures one day and yukky zig zags the next?   

Oh, and for a numpty, what exactly does the y axis scale actually mean?   I think the point is that at some point in late October or November Steve's line will burst above x axis indicating that he's surpassed Tommy's annual mileage total but it's sort of an odd representation for somebody who isn't a statto by nature.

I for one find the smoothed trend line more intuitive when visualising progress of TG and Tarzan.

Jo - on the OYTT and How To Read The OYTT Chart webpages on the Gicentre site the text is being rendered in a way that makes it very hard to read on my PC (using Firefox). Is this something other people have noticed?
9 miles SW of Marsh Gibbon

red marley

Re: Visualizing the OYTT
« Reply #218 on: 18 February, 2015, 09:29:15 am »
Oh, and for a numpty, what exactly does the y axis scale actually mean?   I think the point is that at some point in late October or November Steve's line will burst above x axis indicating that he's surpassed Tommy's annual mileage total but it's sort of an odd representation for somebody who isn't a statto by nature.

Tommy rode an average of 205 miles per day in 1939. That's equivalent to riding continuously at a little over 8 mph. Imagine that Tommy, Steve, Kurt and William all set off together on Jan 1st. The y-axis is simply many miles ahead or behind each rider is compared to 'average Tommy' riding at 8mph. This is why the lines tend to by zig-zagged - while they are moving, riders tend be going faster than 8mph so begin to catch up, shrinking the distance behind average Tommy (move upwards on the chart). When riders stop for food or sleep, average Tommy keeps going at his 8mph, so begins to pull in front (riders' lines move downwards on the chart).

Re: Visualizing the OYTT
« Reply #219 on: 18 February, 2015, 01:39:29 pm »
Is there anything that would make it easier? I'd like to be able to make rescaling as intuitive as possible, so suggestions welcome.
I keep getting the functions the wrong way round - i.e. trying to rescale using the axes, and pan by dragging the chart

HTFB

  • The Monkey and the Plywood Violin
Re: Visualizing the OYTT
« Reply #220 on: 19 February, 2015, 03:22:02 am »
Is there anything that would make it easier? I'd like to be able to make rescaling as intuitive as possible, so suggestions welcome.

(The intention is that the point you drag stays fixed and the rest of the chart will expand if you drag up or right and contract if you drag down or left. The chart can be panned by dragging either of the axes).

I wrote a long screed in reply to this, and then lost the whole lot with a misclick, so maybe I'm not the ideally competent user. But since data visualisation could also involve the interface with the visualisation, here's tedious detail.

The paradigmbehaviour I expected at first was that click-and-drag would select a box to expand into the whole plot, with the Full Year and Recent Progress buttons providing the undo function.

Current behaviour, where undo is also provided by a drag, has the fundamental problem that it's not a homomorphism1: after a click-and-drag that didn't do what I expected, clicking again and dragging the pointer back to where I started doesn't get me back to the same place (nor do two drags releasing the button midway have the same effect as one continuous drag between the same endpoints). I try to undo and find myself at -4000 distance or so with only IronOx's progress displayed.

The other real difficulty is that the scale of response doesn't seem natural or consistent. A long drag right  sometimes expands the time axis enormously; sometimes an attempt to contract the axis leaves the graph at the same scale but changes the grid lines to a fine mesh.

If after overexpanding the time axis I manage to contract again, I often find the distance axis has contracted too. Possibly the manoeuvre has made the app oversensitive to a small wobble up or down as I drag left? The X and Y scales need to be completely independent. Of course when the distance axis is overcontracted it's very hard, and quite unnatural, to find the right place to click and reexpand: particularly if the zero line is off the screen.

Perhaps I'm overconditioned to expect the axis to be on the zero line, but the location of the time axis scroll bar is unintuitive. It took until a couple of days ago to discover either scroll bar, and when I found them the natural action was to click on the pointed ends, rather than dragging, but this had no effect.

Of course it would be much easier to learn how it worked if the graph could respond continuously during a click-and-drag; if this is impossible maybe a rectangular that appeared around the pointer could do?

Quite unrelatedly, the font displays really badly on this Windows XP machine. In Firefox horizontal text is illegible, in Chrome the vertical axis label is.
Not especially helpful or mature

red marley

Re: Visualizing the OYTT
« Reply #221 on: 19 February, 2015, 05:10:40 am »
Thanks andrew_s and HTFB for those comments. Swapping the zoom and pan regions (axis dragging vs main chart area dragging) makes sense. I'll have a go changing that round along with moving the legend by dragging it.

I'll have a think about the homomorphic zooming, although to me I wouldn't expect it to be so (increasing a value by 50% then decreasing the resulting value by 50% doesn't take you back to where you started: 100 -> 150; 150 -> 75). Dragging a point to some arbitrary position and then back to its start point without releasing does take you back though.

Of course it would be much easier to learn how it worked if the graph could respond continuously during a click-and-drag; if this is impossible maybe a rectangular that appeared around the pointer could do?

This is a bit odd as on my machine at least (Mac OS 10.10.2 on Safari, Firefox and Chrome), it does respond during the drag as intended. Not sure why it might not on your setup but agree that if it is not doing that for you then it would certainly be much harder to navigate and may be the reason for the apparently unintuitive behaviour.

Quite unrelatedly, the font displays really badly on this Windows XP machine. In Firefox horizontal text is illegible, in Chrome the vertical axis label is.

OK, that's a bit of a pain. Could you send me a screenshot via PM and details of your browser version? My daily charts on this thread are how it appears on my setup with very minor variations in the font rendering between browsers, but all appear legible to me.

red marley

Re: Visualizing the OYTT
« Reply #222 on: 19 February, 2015, 05:25:27 am »
Day 49: Kurt posts one of his higher daily rides at over 220 miles and heads steadily towards Godwin's WR pace. Steve bangs out yet another 200 mile day, ahead of his own schedule. Kurt has lengthened his lead over Steve by 153 miles in the last two weeks.


mattc

  • n.b. have grown beard since photo taken
    • Didcot Audaxes
Re: Visualizing the OYTT
« Reply #223 on: 19 February, 2015, 06:54:52 am »
Apologies if I've missed this, but:
What period is the smoothing done over, to get the trend line?
Has never ridden RAAM
---------
No.11  Because of the great host of those who dislike the least appearance of "swank " when they travel the roads and lanes. - From Kuklos' 39 Articles

red marley

Re: Visualizing the OYTT
« Reply #224 on: 19 February, 2015, 07:07:22 am »
The trend is simply a line joining the daily totals and then with some minor (spline) smoothing that rounds off any sharp changes between consecutive days.