It Wasn’t Your Day

There are times when, despite the data collected and the depth of feedback, I am forced to conclude that the outcome of a race was just ‘one of those things’. It’s not a satisfactory moment, as conclusions go it denies any opportunity to learn from the experience, but race analysis is as much – if not more – art as it is science. A race is not a controlled experiment, the result, and the data, a sample of one while the potential variables are many. Judgements are just that, a weighing up of the evidence, conclusions drawn on the basis of most probable causes. Then, occasionally, there is no weight of evidence and for all the data to hand and the detail of the report, no firm conclusions can be reached.

Paul Deen raced Challenge Roth this year, finishing in a respectable 9:55, but like me he felt frustratingly flat and unable to perform at his intended power for the first few hours of the bike. Paul offered me the chance to review the race data and see if I could find a reason why he struggled early on. I’ve already hinted at the conclusions.

Paul Deen's Challenge Roth 2012 Power and Heart Rate Trends

It’s not that I don’t trust athletes, but the first step is to compare their perception against the data – did the subjective feeling of being powerless for the first three hours correspond to an actual drop in power? In Paul’s case the answer is yes, as the graph above shows his power output increased over the course of the race; much as I generally like an upward trend to power and a strong finish to the bike, this is too severe. Power is initially low, well below the 220W Paul tells me he targeted, but it builds to reach those levels for the final two hours of riding. The power distributions by quarter (below) simply confirm this: for the first half of the race, Paul spent 50% of the time below 68% of his FTP.

Paul Deen's Challenge Roth 2012 Power Distribution

Normally I deal with problems later in the race and look back to find potential causes. Run issues extend from bike issues, bike issues from earlier on the bike, or perhaps from the swim, or even from the taper. Notice how quickly the net widens, there are far too many variables to ever have certainty. If a problem occurs early in the bike and the strategy is sound we have to look to the swim or taper for answers, so my first questions to Paul focussed on these two areas. The swim, roughly 65 minutes, was perfectly within his abilities and his feedback on race week, nutrition, rest and all the other factors that go into a successful taper, revealed no obvious problems. Paul is coached and sufficiently careful in his preparations – and account of them – that I doubted he could have made an obvious error prior to the race.

Paul Deen's pre-Roth Performance Management Chart

We speculated that perhaps his taper was too severe, that he’d eased back too soon and was off form on race day. Possibly. It was certainly the case that my flatness was a product of an unmanaged taper, a race week of nothing after ten days of hard training is not appropriate preparation. The Performance Management Chart (PMC) above (apologies for the screen capture, TrainingPeaks does not allow export of raw data) does show quite a steep final week and a high TSB as a consequence – perhaps there is an explanation within the taper. Perhaps. If I look back at previous posts regarding the PMC and tapering I have raced well off such a high combined TSB; Paul’s pre-Roth chart doesn’t look wildly different to a number of mine.

Paul Deen's Heart Rate and Speed During Challenge Roth 2012

Despite the prolonged low patch this was never a bad race. The ride averages to a respectable 73% of FTP and the run that follows is well paced as the graph above shows. It’s worth considering that much as we might desire to eliminate the weaker start it may have been a contributory factor in the consistency and strength of the subsequent performance. Changing the pacing from the beginning – not that Paul felt he could at the time – would change the entire race with the potential it’s for the worse. I would still seek to eliminate low patches from future races, but bear in mind these lows may have had positive implications for later parts of the race; by removing them we need to be sure we won’t expose other areas of weakness.

Is it just that it wasn’t Paul’s day? Things happen. There are too many uncontrolled variables and their impact too subtle for me to pinpoint a single cause. Was it because of or despite the taper that he felt off form for the first three hours of the bike? Had he trained more during race week would he have performed better on the day, or might he have have started stronger and faded with fatigue?

I agree with Paul – he under performed early on, but the rest of the race is strong and consistent. Actually answering why remains elusive, something wasn’t right both in how he felt and the output he achieved. I just can’t say what. And with the structure of his preparations looking good it’s worth trying again. I would want to repeat the strategy and taper before considering any major changes, it’s too close to call on a single result, if Paul struggles again then he might consider increasing bike stress in race week the next time around. Refining you Ironman is a slow process of trial and error.


  • I can’t give you the quantitative answer because you and Paul are the guys with the race data file but I can add some speculation as to cause. It’s something I think is obvious for AG racing but I think you completely overlooked it.

    With a 65 minute swim, Paul comes out in the thick of a pretty congested pack. That’s the answer. Paul, despite likely being ‘ethically opposed’ to drafting and ‘trying’ not to draft, is drafting a lot for the first portion of the bike. He’s in the midst of a pack that is travelling quite quickly and surrounded by other athletes who are also drafting and they’re all getting down the road faster than they probably should be if they were all alone. The evidence to support this is a high power variability in the first 1/8th of the bike ride. When he’s moving around congestion he is doing appropriate work but he’s got frequent power readings down between 150 and 200 W. This is when he’s sat in and ‘stuck’ in the congestion. When he looks down at his power meter when moving past people he is observing the 220W target, when he is boxed in or congested he maintains a cadence and gear selection that has him moving at before gaining the draft. The resulting situation is that the pedal force drops off without any visual reference that you’re giving any less output.

    I wonder what paul had displayed for electronic feedback during the race? Was he looking at average power? I think it easily could be the case that his feeling “frustratingly flat and unable to perform at his intended power” was not related to any physical feelings, but rather a psychological feeling induced by electronic feedback that he did not understand at the moment. I’d guess that his emotions on raceday were goverened by electronic feedback… somehow he was patient enough to wait for the race to rise to meet him later on. That’s a rare quality in an athlete and his excellent consistency on the run is indicative of great patience on the bike. I would say that this bike-split is probably paced to perfection.

    Like I said, I’m not being critical of the ethical choices Paul made during the race, I’m not accusing him of doing anything that he shouldn’t have, the fact of the matter is though that he’s in the thick of some serious congestion on the road and he is gaining a benefit whether or not he realizes in the moment that he is. You can save 10+ watts riding single file at a legal distance. At the beginning of a race you’re riding two or three wide and typically nowhere near the 10m spacing. It’s reasonable to be saving 30 watts here when sucking wheels which looks to be about the amount that Paul is down early in the ride.

    Russ, if you email me the file I’ll do up a quick analysis that I believe will show that this is the case. I’d be happy for you to share the result on your blog.

  • Bear in mind this is a challenge race with wave starts so congestion behaves a little differently to a mass start Ironman. Paul started in a wave at 7:00am the wave before went off 15 minutes ahead of his and the next wave was 5 minutes after, wave size tended to be around 250, so he likely exited with a smaller set of people round him (all 1:05 swimmers from his wave and some faster than 1:00 swimmers from the next). So he might have had 150 people out in the 15-20 minutes around him. Which may be enough to allow drafting to be a factor in this race; I can certainly accept in some events it can play a role.

    With two laps if anything most athletes will report more congestion on the second lap where you hit a lot of slower athletes – that’s when I spent most of my time passing people, in the initial lap I spent very little time having to slow for any of the people ahead, most were fast enough or overtaking me if behind. Paul’s experience in his wave could have been different, but I know he too complained about congestion in the second lap. The start of the lap is fast mostly, with a general downhill trend and a number of corners – potentially variability there is a product of how you move through this area, it settles later as the roads lengthen and we are largely uninterrupted on country lanes.

    There’s a few graphs I could throw up here and I can check a number of figures over quartiles of the race. I’ll need to check with Paul if he minds you having a copy of his file, but happy for you to look at it.


  • I should add – I take your point on the ethical issue in the early part of many long distance races and the potential need to factor for this. Something I can add to the analysis sheet as a safe check. You are spot on in noting how athletes may not realise what they’re actually doing in terms of power and the impact of pacing changes.


  • Paul Deen

    Thanks for your feedback Joshua but I can honestly say that it doesn’t help explain my poor start on the bike. As Russ points out Roth is very different as regards congestion, It really wasn’t that busy on lap1 and by the time I started feeling good on lap 2, I was constantly overtaking. I probably gained more drafting benefit during my good patch in the last 2.5 hours as I legally came up behind and passed a constant stream of riders including the entire female field who started 15 mins before me, which made for a nice series of views 🙂

    The first 3 hours was horrible, my legs felt like jelly with no power and hip flexors were aching. My conservative 220 watt target felt incredibly difficult, so difficult that I couldn’t hit it without my PE feeling way too high. 220 watts was feeling like a 260-270 watt tempo ride felt in training. PE still felt uncomfortably high at the power I was managing (circa 200 watts). By comparison on an easy 1 hour spin the day before 220 watts felt like I was tickling the pedals & I was left thinking that I would start the next day at 220 but assumed it would feel too easy and I would gradually build it through the ride. I could understand a slow start and poor hour or so but 3 hours was a totally unexpected shock and made for a pretty miserable start to the race with plenty of negative feelings! Without doubt my issue on the bike helped me run a fairly decent neg split marathon but I can’t help but think that without the bad start on the bike I would have finished a good 10 minutes or so quicker.

    I have a Quarq with a Garmin 800 and was displaying Time, distance, speed, av speed, cadence, 3 sec power, lap normalised power. I pressed lap roughly every 20 miles (forgot to set auto lap up per race)

  • Well this discussion inspired me to make some further upgrades on my analysis spreadsheet. Rather than just splitting into quarters the latest version breaks down performance by 10 and 20 minute blocks allowing a far more detailed view of the stages of the race. I hoped this might highlight if there were periods of high variability early on that might relate to a crowded field. As Paul says I think in the case of Roth and most Challenge events you’re less likely to see this happening.

    Here’s a table of the data (excuse the image, it’s the easiest way to produce these now):

    More detailed analysis of Paul Deen's Challenge Roth 2012 Bike

    I looked at relative standard deviation of the power as well as consider VI to look at variability, there doesn’t seem to be much difference throughout the race, especially not with the full 10 minute block analysis. By also looking at percentage of time spent at the different race pacings I can also produce a more detailed chart of pacing over the race:

    Paul Deen's Detailed Power Distribution Relative to Targets During Challenge Roth 2012

    Happy for you to have a look at the file Joshua, and interested to see what you can add to the analysis. Will get Paul’s confirmation on that and send it over.