Tuesday, October 30, 2012

November Election: President

As much to document to myself my positions going into the election as anything else (this being a bicycling blog, primarily), I'll stake out my position on the presidential election. Really, this should be no surprise.

I'm voting Obama.

A sufficient if not necessary grounds for this position is energy policy, and with it, environmental policy. If you read transcripts of the debates, you may not have seen much of a difference between the two major candidates. However, listening revealed a clear difference. When describing his position, Romney would say "Coal. Oil. Nuclear. andrenewables (one syllable)". On the other hand, Obama would say "Of course oil and clean coal (sic) (rushed), but also a focus on the energy of the future... (emphasis)." Is this significant? I say it is.

In the Republican National Convention, Romney ridiculed Obama for trying to "pick winners" in the energy sphere by supporting Solyndra, a solar energy company which went under not so long after. Additionally, he ridiculed Obama for trying to "stop the rising oceans" (not exact quote, but close) with his policies. Laughter erupted from the crowd.

First the prior issue. The government, Romney argued, shouldn't pick winners. That sounds all nice and all, but Romney has no issue promoting the benefits to the economy from increasing military spending, which explicitly involves picking who and who will not receive military contracts. He has no problem dumping billions into agricultural subsidies, which directly involves picking winners, since some crops (like corn) receive more subsidies, others less. He has no problem picking winners in transportation, as the federal government supports massive transportation projects which have a huge influence on which win (like roads) and which lose (like rail). So Romney is a hypocrite, plain and simple. He has zero reluctance about picking winners. Indeed, in energy he's picked his winner. He has stated his plan to open federal land to more oil drilling. Federal land is a shared resource, owned by every citizen. Romney is taking it upon himself to sell the value of that land off in the quest to sustain cheap oil. If that's not picking winners, then nothing is. Romney is willing to talk the conservative talk when and only when it serves the interests of his donors. There is nothing about him which is actually conservative in the true sense. The only conservative in the election was Ron Paul, and he got blown out of the water early on.

Obama has been at least consistent on the matter. His view is te federal government plays a roll in the success of the nation. Consider that the nation's manufacturing dominance came about primarily as a result of our success in WWII, and nobody would deny war is a "federal program". Our success in the car industry came about only due to the massive investment in the interstate highways, a federal military program. And our success in information technology came about largely following our investment in internet infrastructure, which was also a federal military program. And who is the dominent nation in manufacturing right now? That champion of small government, the champion of individualism, that champion of low taxes: China. So the whole "we built it" line the Republicans have been toting is a line of deception and denial. Without a supporting infrastructure, we'd have at best a feudal economy.

Now the second item: climate change. Well, a lot has been said on the manner and I'm not going to summarize it in a paragraph. But my view, and it's my view, is if you ignore the risks of continued reliance on carbon-emitting fuels, you are a reckless fool or a self-serving fraud. I'm willing to give Romney the benefit of the doubt and give him credit for both. Indeed, how is it that the Republican Party promotes such enormous skepticism about scientific projects of the effect of massive carbon injection into the atmosphere on the climate, but is so willing to accept selected warnings about the effect of debt on the economy? A "conservative" view would be to take the cautious position in both. But in reality, Republican presidents have increased the deficit way more than Democratic presidents have, and there's no indication Romney's going to reduce the present deficit any faster than Obama will. So the whole climate change "skepticism" is nothing more than a weak front for well-paid influence.

That climate change has become such a religious issue in the United States, and perhaps only in the United States, is a testament to the efficacy of marketing and the gullibility induced by neglect of science in our school systems. After all, ignorance and religion go hand-in-hand. This is not an area for Obama enforcement, as his position here has been gross neglect, but the sad state of affairs is even gross neglect is better than what the Republicans offer.

So in summary, It's Obama all the way.

I've been following the polls, and in particular the excellent Five Thirty Eight blog on the New York Times, for months, and my confidence Obama has the support to win this election has been unwavering. However, you don't win on support, you win on votes, and if you claim to support Obama and do not vote in the election, I consider that far worse than voting for Romney. Please vote, not because the vote is likely to swingy the election, but because an educated vote is a personal responsibility in a democracy. A vote is as strong a statement of support for checks and balances, of the strength of our system as it is support for a particular candidate. It's especially tragic when entire blocks of the population under-represent themselves through laziness, allowing fringe fanatics their disproportionate influence.

Sunday, October 28, 2012

The elasticity of rest

One concept which comes up over and over is what I call "the elasticity of rest". When I drop a rubber ball from a height h and it bounces off the ground, it will typically bounce up to a height αh, where α < 1. α in this context is typically called the "elasticity" of the bounce. For a fully elastic bounce, α = 1, and all kinetic energy is retained (the momentum changes from downward to upward). For α = 0, the bounce is fully inelastic, and the ball sticks to the floor (I assume the floor is at rest).

I apply (misapply?) this concept to recovery. Suppose when cycling I can average no more than 300 watts for 10 minutes. If I ride at 300 watts, I collapse with exhaustion at exactly 10 minutes. But instead of riding 300 watts for 10 minutes, half-way through I slow to 290 watts for 10 seconds. Then I do my best possible effort for the remaining 4:50. How much less work can I do during that 10 minutes than if I'd stuck to 300 watts?

Using the elasticity concept, an elasticity of zero would imply that I do 100 joules less work: all I can do is average 300 watts for the rest of the ten minutes, so the 10 less watts during the 10 seconds I was at 290 watts is lost, unrecoverable. On the other hand, an elasticity of 1 suggests that I can boost the power the rest of the 10 minutes by 100 joules / 290 seconds (0.34 watts), recovering all of the lost work (this wouldn't produce the same time in a time trial, it would be microscopically slower, but that's another matter).

You might argue a bit of rest might do me good. After all, there's no such thing as constant power. Every pedal stroke the power varies as the foot moves around the crank circle. In this case, you'd claim, elasticity > 1. The rest pays off with interest. But I'm assuming from the start there's an optimal power schedule and the rider is able to follow it. So any change in power from this schedule cannot improve on my work done. At best I can achieve the same result.

The same problem can come in the opposite order: if I go 310 watts for 10 seconds, and thus need to reduce my power for what remains of the ten minute effort, can I rest efficiently enough to still maintain a 300 watt average for the ten minutes? Or must I rest even more, losing more than the 100 joules I just gained in extra 10 watts for 10 seconds?

The first quantitative approach I've seen to addressing this question is Andrew Coggan's Normalized Power constraint. Using normalized power, you can come up with definitive answers to questions such as the one I proposed, and the result is an elasticity greater than 0 but less than 1.

The most recent instance where I've been thinking about this is in the Low-Key Hillclimbs where we've had back-to-back weeks of individual time trials. We follow triathlon passing rules for passing, although we don't have the resources to enforce them. As a result, it's basically honor system that riders pass definitively and that riders being passed allow the passing rider to do so. But it's "low-key", so honor system should be good enough. The question becomes in letting a rider pass, how much does that hurt my result? If I slow by 10% for 10 seconds, I'll cover 10% less distance during that 10 seconds. But I'll recover a bit during this time, allowing higher power the rest of the way. So how much of this do I get back?

Similarly, triathlon rules require a rider passing to close 3 bike lengths within 15 seconds. This probably requires going harder than optimal for a brief time ( this issue is complicatedly the small drafting advantage experienced by the passing rider ). The question is then can the rider recover from that effort without hurting overall power much?

Here's an example of solving the problem with the normalized power model. Neglecting the 30 second smoothing, which is unfortunately critical to the result, if power is reduced from 300 to 290 watts for 10 seconds, that reduces the 4th power of power from 8.1×109 watts4 to 7.07281×109 watts4. So for the remaining 290 seconds I can increase the 4th power of power proportionately: by 2.0963×107 watts4. Power is reduced to 300.194 watts. This increases work done during this time by 95.0 joules. But I lost 100 joules during the recovery period. So the normalized power model, neglecting smoothing, says the elasticity of this rest is 95%. When I add in the 30 second smoothing it will get considerably better.

The essence of my question in this context is: "what is the cost of courtesy?"

Now obviously I didn't invent this concept and I'm sure it already has another name. But this is just the term I've used to myself when thinking about it.

Saturday, October 27, 2012

Numerical Simulation of Low-Key 2012 week 4 Passing

Last time I made an off-the-cuff prediction each rider in today's Boulder Creek to Saratoga Gap Low-Key time trial would be passed, or be passed, by around 2 riders on average. That was based on the assumption riders varied by around 0.5% each in the sequence and that riders varied week-to-week by 2%. These were low estimates, I concluded, after examining the data.

For this post I actually did a numerical simulation, using numbers derived from Low-Key historical data.

speed distribution

  1. For each starter, I generated a normally distributed "true speed", with sigma = 14.4% about a mean (derived from historical data).
  2. For each rider, I then ran a "qualifying climb" in which speed varied normally with sigma = 4.2% from true speed (derived from my score history). Note this was based on the old scoring system, % of median, which includes an artifact that some weeks tend to yield different amounts of score spread than others. So it's probably an over-estimate.
  3. I assigned riders a start time based on the qualifying week, at one-minute intervals.
  4. I again assumed speeds vary normally with sigma = 4.2% on the time trial climb, which for the average rider takes 1 hour.
  5. I calculated the number of passes for each rider based on start order, and the number of times passed for the rider.
  6. I did the experiment 100 thousand times, averaging the results.


speed distribution

On average each riders is passed, and passes, 2.29 times during the time trial. Riders starting first or last tend to be statistical outliers, and are passed, or pass more than this average. This result is around twice what I'd ball-parked last time, primarily because I basically doubled my estimate of week-to-week variability. The distribution of passes and being passed versus start position is as follows:

So the first rider off was passed an average of 20 times in around 90 minutes of riding. Of course, he/she doesn't pass anyone. The final rider passes on average around 3.5 riders, although the rider passing the most riders was the ninth, who passes approximately 6 riders. This is because the slower riders have more time to make up a one-minute gap than the fastest riders.

In 10 thousand separate repetitions, the extreme cases were 28 passes by a rider, and by some statistical fluke one rider was passed 104 times (different riders, different years). Unfortunately the model doesn't handle extreme outliers well: it allows for near-zero or even negative-speed riders if the random numbers come out that way. Even so, if this was the worst thing to happen in a 10 millenia of W Hwy 9, that wouldn't be so bad.

We'll see how it goes today...

Thursday, October 25, 2012

Low-Key Hillclimbs week 4 passing prediction

The 2012 Low-Key Hillclimbs have reached their fourth week. After two weeks of mass-start, and the third week being a small group start, week four will be an individual time trial due to the relative narrowness and relative flatness of the west side of Highway 9 as it leaves Boulder Creek. We're a hill climb series, not a road race series.

There's three ways to do the start: fast riders first, fast riders last, or intermixing the fast and slow riders. We've done all three of these at different points in the past, and this week coordinator Rich Brown has chosen to pre-assign start times going from slower to faster riders.

A few riders expressed concern over this, expecting the extended mass of riders to begin to converge, the faster riders gaining on the slower, until a singularity is reached similar to that of a black hole's mass converging on a singular point causing a cataclysm in the very fabric of space-time. Okay, I'm overstating the case, but at least there have been worries about excessive pass events.

As I've said, we've done slow-to-fast before, because the opposite takes too much time for volunteers who then have to wait for the slowest riders to not only complete the course but additionally to wait for their late starts. So it's worth calculating how many pass events each rider can expect to experience.

Looking at the week 1 results, there was around a 1/2% per rider difference in speed. That means if you line up riders in order of score, and each rider is perfectly repeatable, then if there's a 1/2% difference between riders and a typical rider takes around 1 hour (3600 seconds), then each rider would be 18 seconds faster than the rider ahead, 18 seconds slower than the rider behind. The finish line would thus have riders finishing with gaps 18 second less than they started with.

If riders started with 18 second gaps between riders, we'd thus have issues. But Rich is a smart guy, and realized such short gaps would be a problem, so instead set the start gap to 60 seconds. Over the course of the hour, on average, riders will therefore reduce the gap to the rider ahead by 18 seconds, down to 42 seconds. Additionally, the rider behind will reduce the gap by 18 seconds, to 42 seconds.

Now of course this is terribly idealistic. On top of this systematic gap reduction, there will be essentially a random perturbation. There's multiple reasons for this. One is that riders aren't uniformly distributed in ability. There might be a 0.5% gap, or the gap might be 2%, or it might be 0.01%. If it's 2%, the rider behind over the course of an hour would gain 72 seconds, catching the rider ahead.

But additionally, riders vary week-to-week in how they ride. Some riders do better on flatter roads, some on steeper, some on shorter, some on longer, some when it's hotter, some when it's cooler. Things are never the same. Even on the same course, rider performance chances.

I'm going to assume on average rider performance varies by about 2%. That's around 72 seconds, plus or minus.

So what we have is the initial gap of 60 seconds is systematically reduced to 48 seconds, but on top of that systematic gap reduction we add a random perturbation. Imagine I assign each rider a time 48 seconds different from the riders adjacent to them, then I randomize their times with a normally distributed random number of sigma 48 seconds.

The result is there's a good chance the rider will catch the rider ahead or be caught by the rider behind. Further, the gap to the rider two ahead is 96 seconds: that rider might be caught, or similarly the rider 2 behind might catch the rider. By three ahead, the gap is 144 seconds, so perturbing positions by around 48 seconds is going to have a hard time closing that gap.

Without doing a numerical simulation, I seems clear riders will be passed or will pass around two total riders during the ride. Some won't experience any passes. Others might experience four. But on average it will be around two.

This is a super-crude analysis, of course, better to do a simulation. But it's clear that the concern of some sort of collapsing singularity of riders is unfounded. A few passings over the course of an hour ride is really no big deal. The far greater concern, as it always is, is vehicular traffic. Yet that's something we deal with every time we go onto the roads.

Sunday, October 21, 2012

Low-Key Hillclimbs: test of 2012 scoring code

The Low-Key Hillclimbs are now three weeks into their 2012 eight-week schedule and it's a good time to check on how the new scoring code is working.

In previous years I scored riders based on the ratio of the median rider time to their time, multiplying by 100. Women were given a boost to their scores to balance historical men's and women's scores.

But there were two scenarios which caused problems with this. One was rider turn-out. Montebello, the first climb in the series, historically attracts a broader range of riders than following weeks. Perhaps riders scoring lower in the first week become discouraged, or they ride less and prefer the convenience and more favorable date of Montebello (less off-season), I'm not sure. It would be interesting to research this. But for whatever reason, for riders doing Montebello in addition to other weeks, there's a good chance Montebello was their best score of the year. On the other hand, especially steep and challenging climbs tend to attract stronger riders, pushing the strength of the rider(s) generating the median score higher, lowering everyone's score. The other issue was that steep climbs tend to fractionally spread riders out more: less draft benefit, plus riding on flatter roads, even slower, yields less fractional speed difference for the same fractional power difference.

To compensate for these, I put in a new system, where I calculated a "reference time" and a "slope factor" for each week, self-consistently with other weeks, such that scores should have a fairly equivalent spread for any pair of weeks, considering solo riders who did both, not counting those who had an identified mechanical problem.

So I can check if that's worked so far by plotting scores of riders for week 1 versus week 2 and week 3. What I want to see is basically a straight line of slope 1, with intercept 0: y = x. Riders should scatter about that line, some doing relatively better one week, others relatively better the other. But the goal is to eliminate the motivation for one type of climb to have more effect on net results than another. At Low-Key, we view all climbs as having equal merit: short and long, shallow and steep. They all hurt, just in different ways.

So here are the plots. I think it's safe to say things are working well.

plot

plot

The one curious aspect is that glaring outlier in week 1. At first I thought it was a mechanical problem, but reports are that Holger was just having a very bad day on Quimby. I know how it goes: that happened to me at Mt Hamilton in 1996.

Thursday, October 18, 2012

Proposed additional matching criterion for Strava segments

Strava!

Strava is brilliant at quickly matching typical point-to-point segments. The speed with which it accomplishes this is simply incredible, with long rides in the segment-saturated San Francisco Bay area getting checked in seconds. It's truly amazing.

Part of their approach to this speed is keeping the segment matching as simple as possible. You need to hit the start and end points, but for points in between, there's some wiggle room.

This is important because GPS errors mean that a perfect match of a segment is impossible. Additionally, riders may make minor deviations from an optimal route. On wide roads, there's of course different lane choices. But consider a climbing segment where a rider turned back briefly during the climb to fetch a dropped water bottle, or took a brief wrong turn before retracing steps and continuing on the specified route. Most people would agree these trajectories should match the segment. The result of this is the first and final points of the segment are granted special status: they must be closely matched, but with intermediate points the match criterion is looser, and not all points in the reference data need to be closely passed.

This usually works find. For example, many climbs are up roads where they represent the quickest, shortest route between the two points. If someone wants to climb, say, 80% of the climb on the road then bush-wack the rest, that's their problem. They're only slowing themselves.

But where it fails miserably is on out-and-back segments. For example, a segment I designed and a favorite part of many of my runs is the Gashouse Cove Pier out-and-back. I run out the pier, around the small structure at the end, then return.

But the problem with such segments is Strava fails to recognize the importance of the turn-around point. To Strava, it's just another point, one which can remain unmatched if a sufficient fraction of the other points are matched. But using that 80% number, it's clear running 80% of the way out an then back will, in the case of minimal GPS error, match 80% of the segment points. If Strava's segment is no more than 80%, credit is given for the segment.

Indeed, that's what happened with the Gashouse Cove Pier segment. Here's the present KOM.

I propose a solution. The key is to introduce a "special point" which must be matched on an equal basis with the start and finish points. The question is how should this point be chosen? Manually by the user at the time of segment creation? That is a possibility, but there's already a LOT of segments out there which could use this treatment, and I doubt Strava wants to add complexity to the segment-creation process. For example users might be tempted to add too many "special points", yielding a large number of false negatives.

In cases like this it's a good policy to not try for perfection. Instead recognize that cases like the Gashouse Cove Pier segment tend mostly to consist of point A, a route to point B maximally distant from point A, then a return path back to point A which is maximally distant from point B. You could have a spiral path to a turn-around then spiral out, but this would be an exceptional case.

So this suggests a simple approach: take the two points which are furthest apart in the defining segment, and declare these as "special points". So then not only must one go from point A to point B, one must also visit the two points which are furthest apart. In our simple out-and-back case, these will likely be the start/finish and the turn-around.

The primary concern with this is if the reference data are of poor quality (but then poor quality data shouldn't be used to define segments). If there is a lot of GPS noise in intermediate points, it may be that a rogue point is more likely to be one of these implicit special points. It would be good to check if the extreme point appears to be a rogue point, for example if it is separated from its neighbors by an extreme relative speed. But this is in the details.

A nice thing about this approach is except in pathological cases it is likely to be purely restrictive: data which would have matched a segment would no longer do so when applying these criteria. Thus the test can be applied after the normal matching algorithm has accepted a match. This makes it relatively efficient: the check would only need to be done after the vast majority of segment candidates in the region had already been eliminated.

Finding the points of maximal separation is non-trivial computationally, but it would need be done only once per segment, then cached. From then on these points could be referenced directly, at least until the segment was possibly edited.

Anyway, easy to propose algorithms like this, but more impressive if I could test them. Unfortunately time for that sort of thing is perpetually limited. And it does nothing for the multi-lap problem, which could nonetheless be solved using an extension of the idea.

Wednesday, October 17, 2012

Presidential Debates

Some random peeves about the U.S. Presidential (and Vice Presidential) debates so far:

  • The obsession with short-term statistics: jobs since Obama took office, price of gas since Obama took office, etc: what's needed to promote the long-term interests of the U.S. isn't to chase short-term statistics. Investments, for example primary education, typically hurt short-term statistics in favor of a large long-term payback. Want to promote short-term jobs and GDP? Blow out the deficit, cut all education spending, and abandon all infrastructure maintenance. Hmmm... seems like the Bush administration. I don't claim Obama has been inspiring on long-term versus short-term focus, but the repeated attacks from Romney/Ryan using short-term statistics undermines their credibility towards long-range thinking.
  • The obsession with fuel prices. It is a myth that cheap fuel is good. It increases auto use which increases pollution, increases carbon emission, increases road congestion, encourages medium-term investment in fuel-inefficient vehicles, renders sustainable energy research and investment uncompetitive, increases long-term investment in sprawled exurbs which cannot be adequately served by public transportation, increases demand for roads which carry a perpetual and high maintenance cost. All of these increase time spent in traffic, reduce public health, and squander land in unsustainable development. I could go on. To the contrary, there is an optimal price for fuel, as there is an optimal price for any valuable resource. The Republicans claim to be proponents of the free-market and should acknowledge this basic fact. Obama doesn't score much higher: his claim owners of federal drilling rights are "hoarding" them, keeping the resources in check: hey, that's what we call "the free market promoting conservation". It's long-term investment. But Romney's focus on the short-term drilling rate shows he's far worse.
  • Apparently the environment is taboo. Score one for big oil and those in their employ: Fox News, Rush Limbaugh, and the Drudge Report. Not a single word was said about the devastating environmental impacts of coal mining, for example. Obama at least hinted that alternatives were better. But "clean coal technology" is oxymoronic.
  • Romney's flip-flopping: I'm a big fan of flip-flopping: reasonable people listen to other views and revise their own. Indeed, someone who never "flip-flops" shouldn't be in office: it suggests their brains congealed at age 21 and they've become drones incapable of independent thought. But Romney's cynical attempt to recast his positions towards moderation since the Republican National Convention reeks of cynicism. His present hints he's going to cap tax credits, which will preferentially tax high earners, is news to many of his supporters. And make no mistake: those supporters funded his campaign and did so with the reliable expectation that favor will be returned. His present promises lack any credibility.
  • Military spending: Obama called Romney on this one, so it's not so much of a pet peeve. But to think that the US needs to increase its bloated, pork-infested military budget by another trillion dollars is beyond cynical. It makes Dwight Eisenhower look like a radical liberal. It's ironic those who never served in the military tend to be those who speak most assertively for increased military spending. This issue alone is sufficient to vote against Romney, for Obama. There's not a chance that Romney's going to cut the deficit while increasing military spending without draconian measures he's not yet described.
  • The assertion that taxes on upper incomes tax small businesses. The reason small businesses may file as individuals is because the individual rates are so favorable. The remedy is to make it more attractive for small businesses to file as businesses, as I think any rational individual would recognize they should.
  • "... since Obama took office". Obama took office when statistics were in a free-fall. It's generally accepted it takes up to a year for administration policies to grab hold. Certainly statistics from the first budget year, rather than the first calendar year, of an administration are more applicable. Yet Romney and Ryan repetitively use this phrase in the debates and in campaigns. It's a classic Fox News tactic, at least from what I've read. Correct me if I'm wrong: I never watch Fox News.
  • "Get tough on China": Romney, like George W. Bush, claimed he was going to get tough on the Chinese, to make them inflate their currency. Sure, I support this, as it makes "shit people don't need" more expensive. Americans clearly consume too much stuff: we've got thousands of acres of long-term storage facilities crammed with it, garages stuffed with it leaving their intended contents parked on crowded streets. I'm a big fan of increasing the cost of imported goods, making U.S. goods, manufactured to higher work and environmental standards, more competitive. But that's going to result in a substantial inflation hit, as conventionally measured (fixed basket of goods). Is Romney willing to accept that? Snort. In any case, manufacturing doesn't magically spring forth from the ether. It requires technology and management which we have virtually abandoned. With the Republican Party obsessed only with eliminating credible science from the curriculum, I don't see them leading the way in changing this.
  • Party-compliant ties. Funny. It makes me want to never wear red ties again.

But this is a bike blog. Who's the most bike friendly candidate? Obama is clearly fitter, and Romney's obsession with lower oil prices implicitly promotes bigger, more powerful vehicles driven further and more often. So Obama wins.

I look forward to seeing how the polls respond, if at all. There tends to be a late surge by the incumbent as people on the fence default to the one in office. Romney has been getting close to Obama in polls but that's not good enough: he needs to be ahead right now. I'm quite confident Obama will win this one if the voters who would tend to support him: young, female, and black, bother to show.

Sunday, October 14, 2012

Low-Key Hillclimb: Quimby Road

Yesterday was the second Low-Key Hillclimb this year, up Quimby Road. It was the only one of the first three which I'm not coordinating, so I got to participate.

It was a gorgeous day in San Jose, perfect weather for the climb. I'd decided to run instead of ride. Low-Key Hillclimbs are generally about cycling but we've allowed runners in the past, most notably Gary Gellin, then Brian Lucido, and me on a few occasions. Runners add to the eclectic nature of the event. I still remember Gary chugging past me while I rode on the steep lower slopes of Hicks Road a few years ago. It provided added motivation to pick up the pace when the slope relented a bit, and I was able to repass him, but I think it added a lot to the "Low-Key" aspect.

There were a few issues. One volunteer was worried that riders were failing to hand in release forms. These are important because even if a rider is a close friend, riding the series for years, it's important to have a properly signed waiver because if something tragic happens (always a possibility whenever doing anything on the roads, riding, walking, or driving) we may trust the rider but we can't trust the rider's family. The Kim Flint lawsuit against Strava demonstrated the importance of that lesson: Kim Flint surely would have never held Strava responsible for the risks he was taking, yet Kim wasn't around to exercise reason when his family selfishly sued.

Another small issue was cups: I forgot to bring them, so riders were deprived of pre-ride juice which coordinator Lane had hoped to provide. I've gotta be more organized next week!

But these small hiccups aside, it was a great day. I'd encouraged Matt Allie, whom I knew from the Palo Alto Noon Ride and who is an excellent runner, to join and he did so. He started near the front, me further back, so he got probably a 10 second head-start on me as I waited for the lead riders to get moving, but this proved justified as the Strava record shows he steadily expanded that gap every bit of the way to the finish.

A nice difference running versus cycling is runners have effectively an infinite range of low gears: shorten stride, even go from a formal run to a fast shuffle, and until traction becomes a problem, no grade is too steep. It allows running tangents, including the inside line of Quimby's intimidating switchbacks, and of staying out of the red zone where many riders are forced to over-extend just to keep the bike moving forward. The steeper the hill, the better runners can do versus cyclists, and so since the initial portion of the climb is quite gradual, I found myself quickly off the back. As the grade increased to the 9% average it holds for most of its first half, however, I was able to regain the pace of the rearward riders, then close the gap.

I watched my Garmin Edge 500, which I held in my hand, initially for distance but then later for altitude gained. Really it's a combination of the two which predicts remaining effort, since I can gain altitude faster, despite running slower, when the road is steeper. But altitude is the more conservative of the two when the road's getting steeper towards the end as it does on Quimby.

I passed a few runners on this less-steep portion when I reached the brief break in the steep climbing at about half-way elevation-wise. I lost a bit of ground versus the rider near me at this point, but I knew the really steep stuff followed, so I bided my time.

The final half of the elevation gain is 13% average, but the switch-backs are much steeper. Here I tried to maximize my rate of progress, as a full run stride wasn't sustainable at my speed, so I adopted more of a shuffle-walk. I was doing well, though: breathing hard, gaining on riders on the steep slopes.

The finish is visible from a good way off, but then disappears from view before reappearing when the grade levels out slightly for the final few hundred meters. I was afraid of losing places here but I did okay, finishing well ahead of my projected time, with a VAM just on the happy side of 1 km/hour. I often climb at that VAM on my bike, so I was really pleased with that.

Low-Key finishes are always wonderfully social occasions for the participants. I often help with results after finishing, but since I was too slow to reach the top in time for crunch-time the super-excellent volunteers had things under control. The results were challenging since the riders were grouped on the run-in to the finish line, creating noise which made it hard for the volunteers to hear rider numbers. If we do Quimby again we'll need to think about how to handle this.

Anyway, I still need to finish results, which is a slight detective game since a few numbers were missed. Despite the challenges, however, the finish line volunteers did an overall excellent job, so I'll work it out. We have a video so we can check that, as well as photos of most riders crossing the line.

Friday, October 12, 2012

November Election Pt 2: California State Propositions

Now it's time for the state propositions. These will probably be tricker than the city ones.

  • Proposition 30 - income tax increase of 0.25% for incomes over $250k to fund schools: Reluctant yes. I oppose targeted revenue. I believe all revenue should go into the general fund. In the end, targeted revenue is generally a deception, anyway, since unless the individual targeted component is more than the entire budget, then funding from the general fund can be reduced to bring the funding for that target back to or even lower than it originally was. Classic case: San Francisco voters decided to send more funding to MUNI so the police increased their fees they charge MUNI essentially taking all of that increased money for themselves. But I think the revenue is needed, especially in an age when the federal government is looking to push more fiscal responsibilities onto states.
  • Proposition 31 - Budgeting rule changes: No. Are you kidding me? I stared at this one for several minutes and I still had no idea what it actually does. This is way to complex for each citizen in the state to fully analyze and comprehend. This sort of complex legislation is why we hire people to represent us in Sacramento.
  • Proposition 32 - Political contributions by payroll deduction: Yes. This would prohibit unions from using payroll-deducted funds for political contributions. I am a full supporter of voluntary contributions from workers, but unions have a particular position of power over their workers, and to use dues deducted from worker pay for the purpose of lobbying is an abuse of that power. I do not feel the influence of unions on state government is net positive. There is an issue with the imbalance created by unlimited corporate donations, but I feel that this problem should and will be dealt with separately.
  • Proposition 33 - Auto insurance rates: No, no, no, absolutely no! This allows insurance companies to set rates based on a past history of auto insurance. This is grossly unfair to people who go "off the grid" and use alternate transportation, seeking to reduce their expenses by declining to purchase expensive auto insurance. This proposition is being claimed to facilitate switching insurance companies but it's a fraud.
  • Proposition 34 - Death Penalty: Absolutely yes. The death penalty is corrosive, expensive, and unproductive. This would essentially end it in California. Victims families will need to find "closure" in other ways, similar to the vast majority of the civilized world which long ago banned this barbaric institution.
  • Proposition 35 - Human Trafficking: No. I almost always oppose manditory sentencing. This isn't needed: there's already laws on the books against kidnapping and sexual exploitation. The remedy is to enforce the laws we already have.
  • Proposition 36 - Three Strikes Reform: Absolutely yes. I have always been against three strikes on many levels. This restricts the felonies to which the three stike law is applied to violent crimes or those involving firearms. This comes closer to the intent of the original law.
  • Proposition 37 - Labeling Genetrically Engineered Food: Yes. There is no conceivable reason to oppose this unless you are against all food labeling laws. There is considerable concern about the effects of genetically enginnered foods and the people should be able to decide for themselves whether the lower prices are worth what they perceive to be the risk.
  • Proposition 38 - Tax to fund education and early childhood programs: Yes. I need to read more of the fine print, but this seems decent. 30% of the revenue is targeted for debt reduction, which is good (but I'm not sure how it is enforcable). 30 and 38 are mutually exclusive but if you are unwilling to vote for both then vote-slipping condemns both to failure.
  • Proposition 39 - Yes. Miltistate businesses and "clean energy": I like the proposal to simplify the corporate tax code to remove the incentive for corporations to move their offices outside the state. I do not like the earmarking of revenue received for "clean energy upgrades." I fear these programs tend to be boondoggles of unchecked spending on installations which may or may not accomplish anything. The best approach to "cleaner energy" is to make the dirty energy more expensive, then the free market takes over. But I'll vote for this one.
  • Proposition 40 - Retain state districts designed by the redistricting committee: Yes. This one is no-brainer. There's not even an argument against in the voter guide.

Wow -- that's 8/11 yesses. I'm clearly slipping.

Election time: San Francisco propositions

Early voting has already begun in San Francisco. Time to crack open the voter information guide and make decisions on the too-many-to-count-on-my-fingers propositions facing the fine voters of San Francisco this year. To honor the futility of it all, I am today wearing my "David Chiu for Mayor" volunteer T-shirt.

My vote on propositions tends to be like Governor Brown's view on signing bills: I need a compelling reason to vote yes, otherwise I vote no. So with that, first, the city measures.

  • A - City College Parcel Tax: No. Parcel taxes are regressive, since they tax property independent of its size.
  • B - Clean and Safe Neighborhood Parks Bond: No. Park maintenance should be spent out of annual funds, not bond debt. At least with last year's roads bond, Proposition B, which I also opposed, there was the argument that there was a permanent cost to delaying repairs. I'm not going to buy a similar argument for Parks. Don't let the city cop out of its fiscal obligation to fund parks responsibly.
  • C - Housing Trust Fund: No. Efforts to "create affordable housing" invariably increase, rather than decrease, the cost of housing. The reason is by creating below-market-rate housing that is restricted to a limited number of people, the lucky few who have these discounted spots never want to leave, the available supply of at-rate housing is reduced, and the price of that housing increases: simple supply-and-demand. As a result companies who wish to grow in San Francisco are unable to find housing for workers, and so go elsewhere. Jobs leave the city, the economy tanks, and there's no resources left for these "affordable housing" programs. The best way to encourage affordable housing is to provide free market incentives, such as removing burdensome parking requirements which fundamentally limit density. Increased supply yields lower prices, not convoluted government programs.
  • D - Consolidating Odd-Year Miltiple Elections: Yes, since the controller says it will save money. Downside is it may result in more efficient voter fraud, since you get more office holders for the buck by consolidating elections into off-year elections. But I'm not sure about this, so I'll stick with the claimed cost savings.
  • E - Gross Receipts Tax: non-trivial Yes. This measure would increase the number of businesses which pay tax to the city from 7500 to 15000, increasing city revenue by $28.5M/year (around $35/person/year). As such, it's an increast cost to business, which discourages businesses from being here. On the other hand, it reduces the marginal cost of hiring, since it shifts revenue from a payroll basis to a gross receipts basis. Gross receipts preferentially taxes low-margin businesses (like supermarkets) versus high-margin businesses, since for the same net income, gross receipts are higher if margin is lower. However, I think this is a net win for the city, since a payroll tax is clearly counter-rational: we want to encourage, rather than punish, hiring.
  • F - Water and Environment Plan: No. This proposition supports draining the Hetch-Hetchy reservoir, which at this point is incredibly stupid. There's plenty of conservation battles to be fought but this lateral move simply makes no sense.
  • G - Policy Opposing Corporate Personhood: Yes. This is a nonbinding "city policy" statement which I fully support. Corporations do not have fundamental free speech rights on the same level as individuals. The individuals within a corporation have full right, of course, but to grant additional rights to the corporation is double-counting (as we say in the semiconductor device modeling world). That was never the intention nor is it the content of the U.S. Constutution. Conservatives, who nominally support a restricted reading of the constitution rather than broad, liberal interpretation, should be huge supporters of this one.

Wow -- that was fast. Only one toughie: the gross receipts tax. But I think the payroll tax is sufficiently damaging I feel okay with my position on that one.

Wednesday, October 10, 2012

Low-Key "SuperTeams"


When I designed the team scoring for the Low-Key Hillclimbs I decided to keep it relatively simple: the top 3 scorers from a given team on a given week contribute to the team's score.

A small dose of complexity: if a member of a team volunteers for a given week, his median score from other weeks can rank as one of the counting top 3 scores for the purposes of the important overall team ranking, although it will not qualify for the team's score for that particular week's ranking.

To get that overall score, the top half of a team's scores (rounded up) are summed. This gives teams plenty of throw-aways, for example when they aren't able to get a full complement of three to the climb.

If you're the 4th best rider for your team in a given week your score just got flushed down a big Low-Key toilet. You contributed nothing to a team result.

This was intentionally designed to be forgiving to smaller teams. They need to produce only three good scores, and to do so for only half the weeks. Encouraging smaller teams encourages diversity and competition, and that creates interest and excitement.

It's not totally fair to say the fourth-best rider didn't contribute, though: even if your score didn't count, it was available as a back-up in case one of the top 3 in some way faltered. Maybe it didn't count this week, but in some week maybe it would count. Stuff happens.

So this applies to the fourth-best rider on a team, and to a substantially lesser degree to the fifth, but by the time you dig down to the sixth-rank rider in a team, it's pretty much guaranteed that rider won't contribute at all to the team's standings. Even if team turnout is so low that the sixth-ranked rider is among the top three scorers for the team, that score will probably end up being discarded in the end since the week probably wasn't in the top half of the team's scores.

So really, the sixth best climber on a team is wasting his or her effort. Better than sticking with the super-team to whom he or she is contributing absolutely nothing is to instead join a smaller team not in contention for the overall win but very much in contention, for example, for top 10. It would be far more satisfying to have the scores go to good use then to get ground into the compost heap of super-team discards.

Yet the super-teams persist, recruiting rider after rider, crushing the loyalties of former, slower recruits into mush. But these former recruits keep showing up, week after week, perhaps thinking they are helping the cause. But they are not. They've been dismissed, ignored, left behind.

So my recommendation? Unless you're up there in the top three or maybe four of your present super team, find one of the smaller teams and make a contribution.

Interbike 2012: Regrets

A problem with my Interbike experience is I didn't adequately prepare. In retrospect, I should have made a list of must-see booths and made sure I'd tagged all of those. I tried to do a comprehensive scan of the floor, but while this was possible on the ground floor where the booths were laid out in a fairly regular fashion, on the upper floor it was more of a maze, and attempts to systematically search the floor tended to quickly degrade into a random walk.

I wanted to see Parlee and they weren't listed. Their new Z0 looks really nice. It turns out they were there, but piggy-backing on the Enve booth, which I also would have visited but simply missed.

I really wanted to meet Jason of Fairwheel Bikes. They used to be a regular, indeed a highlight, of Interbike with their project bikes proving that the little guy can nevertheless impress. In cooperation with Brent/Bre Ruegamer they produced some truly impressive weight-weenie specials. But this year Freewheel didn't have a booth. It turns out Jason was there, walking the floor, but the likelihood I'd both encounter him and either recognize him or happen to read his badge was very low. It turns out Fairwheel also had a presence at that Enve booth.... oh, the pain. CyclingNews coverage is here.

Other than this there wasn't much weightweenieism this year. Cycling, like any other technology area, doesn't progress uniformly but does so in steps: rapid progress followed by plateus. This is clearly a plateu time. For example, back in 2008 Brent Ruegamer was making amazingly light custom carbon frames. Now there's a bunch of custom builders getting close to what Brent did then, perhaps giving themselves a bit more of a safety margin. But showing a sub-850 gram custom carbon frame isn't as big a deal as it used to be. The limits of the material have pretty much been reached, but more people are now reaching those limits.

I was at the Ritchey booth, but can you believe I overlooked the classic 1977 650B hardtail? That would have been a highlight of the show for me. I would have loved to know the specs on that bike, and how they compared to the latest surge in 650B frames.

I also missed some big names. Mario Cipollini was there, promoting his bike brand. I stopped by their booth several times and never saw him. I know he at least ventured off to the De Feet booth to get some custom gloves sized, but it's also possible my focus was so on the bikes that I missed him, amazing as that seems. I was hoping to see their new Bond frame, but they didn't have it there. Their existing frames were a lot thicker than I'd expected based on photos I've seen.

I missed Miguel Indurain. Miguel's a super-classy guy, interesting in light of what has been revealed about the era in which he dominated the Tour. It's interesting he's avoided serious controversy so far.

If I return, I'll definitely download their app which allowed searching for specific vendor booths. They had a map with numbered booths, then a long alphabetized list of venders with associated booth numbers. It was cumbersome to use and a phone app would have been much faster. At some point convention halls like this one will provide indoor location services which would help even more.

Tuesday, October 9, 2012

Lance Armstrong blood values from 2009 Tour de France

There's recently been a big story about how Lance Armstrong's biological passport blood values from 2009 were "consistent with doping". My response was "this is news?"

It was news back in 2009, when Lance was boldly publishing his blood values on-line to prove his cleanliness, perhaps thinking people would pay attention only to that his hematocrit was well under the 50% limit, until it was pointed out by Jakob Mørkeberg, a Danish researcher, that the blood values he was posting appeared consistent with doping. His response: "What do you call a guy who graduates last in the class in medical school? Doctor." Funny, dude, but it did nothing to address the analysis.

You don't need to be a medical expert to see something was amiss. Here's a plot I made at the time:

Lance came out of the second rest day that Tour raging. A year later, out of contention, he came out of the second rest day fatigued and uncompetitive. Lance skipped his supplements, was the obvious interpretation at the time. I don't think there was anything subtle about it.

Here's what Michael Ashenden says in this recent article in VeloNation:

His reticulocyte levels were below the average of the rest of his reported results...consistent with the use of blood transfusions.

It doesn't take a degree in statistics to analyze the data in that plot.

There's a nice article on this matter on The Sports Scientists Blog.

Monday, October 8, 2012

Low-Key Hillclimbs begin!

As we approach October, my anxiety level always rises. The Low-Key Hillclimbs loom.

But somehow it always comes together, and this past Saturday was no different. We had perfect conditions and essentially no conflict as our group of merry climbers tackled the traditional series-opener, Montebello Road.

It was good to see a lot of returning faces, and a good number of them were looking noticably leaner. It's great to see when people come in, lighter and fitter, and take minutes off their previous best times. It's also great seeing how much the junior riders improve, which they always do.

I mostly milled around during registration, answering questions while a fantastic group of volunteers signed in riders. Then we organized a neutral roll-out to the base of the climb, started the timers with a honk of the car horn, and drove up the hill to the summit for timing.

Timing went exceptionally well. Howard is a pro with his stop-watch. Others recorded rider numbers and the all-important rider-time combos. Riders arrive too quickly to get numbers and times for every one, but by having a dedicated list of times, a dedicated list of numbers, and a list of occasional time-number pairs, it's fairly easy to construct a virtually complete set of results even if a few numbers or times are missing. Without the combos, if the times and numbers differ in length by one, where does the gap go to bring them into alignment?

But that's not enough. My goal for the day was to get rider numbers a second time, after they'd rolled past the finish. If necessary, I checked the sticker we'd attached to their stems to which they were to have referred prior to shouting it out at the finish. or finishes of small packs this role is very useful, because at the finish line it can be difficult to discrimate the near-simultaneous shouts. And sometimes riders simply forget all about the number shouting thing, even though we verbally request it as they approach. When the muscles have sapped all avaible oxygen, leaving none left for the brain, cognition isn't at its snappy best.

With all of this duplication we add even more: a photographer photographs each rider crossing the finish. Photographing the stop watch synchronizes EXIF metadata in the photo with ride time. We could do full results this way, but it would be very time consuming. But it's critical for untangling very close finishes, or for resolving issues when riders misread their number stickers and I don't pick up the error in my role as sweeper.

Then when I get home with all of these reg and results sheets, I make a spreadsheet with a list of numbers and times and run it through my Perl-based scoring code. That handles all of the bookkeeping. If I have duplicate numbers, or have numbers associated with riders who hadn't registered, it lets me know.

If there's still issues, I check the photos posted on-line by the finish line volunteer and/or check the rider jersey and bike description written next to the rider's name at check-in. I've also got rider mass to help identify photos if the user voluntarily provided it for my "fun" ranking on mass-adjusted climbing rate (VAM multiplied by mass, a crude measure of raw watts).

After results are done, I make a list of web sites of photos volunteer photographers have uploaded (typically there's several) with credits, I select a photo to go at the top of the results page for the week and fill in a caption and link in my data file, then finally I write a little race story. Then I need to reset the "RSVP" page to allow riders to register for the following week's climb and manually shut off the "registration full" flag if I'd set it the previous week. I then run the code which presses this all into HTML and uploads it to my site.

Tipically there's errors: riders with the wrong team, spelling errors, even incorrect numbers somehow entered despite the checks in my code. But I can make these changes extremely rapidly, typically typing a few characters in an appropriate file then re-running my script.

Next week: Quimby Road. The fun begins again. This week I have an excellent volunteer coordinator to run the main show, so I will probably run it as part of my preparation for the CIM in December.

Friday, October 5, 2012

Interbike 2012: Services

In addition to equipment, there were service providers at Interbike. Here's a few which caught my attention.

Guru
The Guru bike fit rig.

First, bike fitting. The news was that Specialized bought Retul, then in return Cannondale bought Guru. I stopped by Guru's booth to see if they were dropping their bike line, including the Guru Photon, which at one point was my dream bike (until I started reading reports of them breaking). Guru had some slick software which took contact points and referenced them to geometry from a range of commercially available stock frames. This seemed nice because a rider who was considering custom geometry could select from a stock frame which came close to "ideal", whatever that is. Different brands follow different trajectories through fit space. For example, considering just stack and reach, at a given stack different bike brands will have different reach values. And when you consider seat tube angle, head tube angle, bottom bracket drop, and chainstay length there's a lot of room for differentiation. The irony is, however, that since they've been bought by Cannondale I wonder if they're going to be willing to provide geometry for competitor bikes.

They weren't the only fitter there: I even saw "Fit Kit", which was founded by Bill Boston decades ago. I didn't even realize they were still in the game.

There was ChipABike, which seemed to somehow involve putting RFID tags into bikes to help them get recovered. It seems like a good idea, but if someone steels my bike and sells it at the Oakland Flea Market, nobody's breaking the cycle to swap RFID tags. The old school equivalent is to slide a sheet with your name into the handlebar, so you can prove the bike was yours. I didn't quite have ChipABike figured out, but I'll check their web site after posting this.

Risk Placement Services seemed to be a source of event insurance. I wasn't totally sure: I took a card. I've gotten insurance from USA Cycling and League of American Bicyclists. It's good to know options.

MyLaps provides event timing services, including hardware and web-posting of results. I was a bit confused because I saw the booth featured ChampionChip, which has been around for at least ten years, but what I didn't pick up was that MyLaps bought ChampionChip. In the Low-Key Hillclimbs we go old school and just write down numbers and times, with a backup of EXIF data encoded in JPEG files taken of riders crossing the finish. It's labor-intensive but it's the most flexible system. It would be fun to try chip timing, though. MyLaps quoted a fixed cost of around $5000 for event timing. At Low-Key we could eventually amortize that but until then it would seriously bite into our charity contribution. A cheaper option would be bar-code reading but it would be an interesting exercise trying to read bar-codes of a pack crossing the finish. You'd need to bin cyclists and sequentially scan them but by that point you may as well just ask them for their number.

There was a lot more, of course. It's a huge show and doing justice to everything is impossible, especially in one day.

Wednesday, October 3, 2012

Bridge-to-Bridge Run 2012: pacing analysis

A nice thing about running with GPS is the ability to go back afterwards and try to analyze what happened. Did I go out too hard or too easy? Was there a particular section of the course where I had special difficulty?

First, results. I finished in 41:46, 33rd overall out of 1653, 32rd overall out of 727 males. I can't complain about that. But I need to remember a large number of the runners were just out there to enjoy the day and the views, not shaving seconds.

In cycling a power meter is really nice for pace analysis, but in running we only have pace and road grade. These are usually fairly good because running pace isn't as dependent on external conditions such as wind and surface quality as is cycling, but even a few % difference in speed-at-a-given-effort can confound trends in effort due to fatigue or loss of focus.

On Bridge to Bridge of the more than 1600 runners, 20 others are listed on Strava as having done the run with me. To get a match, they need to have run a sufficiently similar pace that Strava recognizes us has having been together. Additionally, of course, the runner would have needed to record their run on a phone or GPS unit, and to have uploaded it to Strava. Their market share around here is less in running then in cycling, but it's growing.

I stumbled upon one runner in particular who seemed to be suitable for comparison: Philippe Le Rohellec of San Francisco. A quick glance at his Strava record shows he mixes cycling and running. He finished the Bridge to Bridge in 39:56, right at my reach target.

First, temperature: my Garmin reported between 18 and 22 C during the run. However, it was mounted to my wrist and obviously there it's exposed to body heat. So this places an upper bound on the ambient temperature.

Next, I'll look at speed. My target was 10 kph, as I've noted. In the following, I smooth the speed with respect to distance with a characteristic length of 50 meters, so peaks and valleys will be underestimated, but "noise" due to GPS error will be reduced.

speed

You can see I start out above pace, then settle into my 15 kph target, then there's the hiccup from Fort Mason, then I sag a bit before partially recovering in the final kilometer. In contrast, Phillippe had a very strong middle race, actually picking up his initial pace.

Of particular interest was Fort Mason. I zoom into that here:

speed

I curiously had two sags in speed leading into the climb. I think the first may have been distraction from the aid station which I skipped, or perhaps the view of the upcoming hill. I'm not sure. Phillippe had no such issues.

Then we hit the main climb. Phillippe was substantially faster on the early portion. I recovered quicker over the top, perhaps having paced myself a bit more conservatively on the steep climb. On the descent I was actually okay in comparison to him, but then once we hit the flats he kept his speed, while I soon dropped below my 15 kph target, never again to recover it.

gap

I can also compare our relative positions. Here I take, for every Phillippe position, how far back I was. The most interesting aspect of this plot is its simplicity: I held him close for 2.5 km, then he started to pull away at a steady pace for the rest of the way.

So what held me back from my 40-minute reach goal? The answer, it seems, is I ran too slowly. You could say I started out too quickly, but I started out at almost exactly the same pace Phillippe ran. Rather, I simply couldn't make it happen after the first 2.5 km. I didn't really crash: once I slowed relative to Phillippe I was able to hold that speed differential.

I shouldn't give up hope. My friend from the start line, who I now see was Tim McMenomey, finished the Avenue of the Giants 10 km race (results) in 38:33 on May 6th, yet he finished only two slots ahead of me here, in 41:22. So differences race-to-race of several minutes may not be unusual, although I've been remarkably more consistent in my spotty but long-in-time 10 km running career.

Tuesday, October 2, 2012

Race Report: Bridge to Bridge 10 km run

I've been training for the Califoronia International Marathon (CIM) in Sacramento, California on 02 Dec, focusing primarily to this point on endurance. When two weeks ago I was able to do a 19.6 mile run then follow it up the next day with a 7.5 mile run, I decided my endurance was on track, and I could indulge in a race to check my speed. I saw the Bridge to Bridge had changed its traditional 12 km long course to a "certified" 10 km this year due to construction. I couldn't resist, as I've got a goal of breaking 40 min for 10 km (I guess 48 min for 12 km would also qualify, but that's less likely to happen): I registered.

The race started at the Ferry Building in San Francisco, around 4 km from where I live, so I decided to run to the start as warm-up. This is the most warm-up I've ever given myself in a running race: in the past I've taken the approach that miles running take more out of the tank then the advantage gained in priming the engine, so a "warm-up" run is counter-productive. But I've been feeling better in the middle of my long runs than near the beginning, so I decided to take a lesson from my training. In any case, running was by far the most convenient way there.

So I did a nice easy run out my door. It was nice running without the slightest pressure for speed. Indeed, the pressure was against speed. I felt good when I got to the Ferry building, an hour before the start.

I'd worn a sweat suit over my running clothes, so despite my slow pace I was feeling nicely sweaty on the uncharacteristically warm and clear San Francisco morning. The rest of my time I spent ironing out sort/tight spots in my legs with a massage stick I'd taken along in my small backpack, doing some light stretches and balancing poses, and a few very short running drills to work on my foot strike. I felt fairly ready.

I moved to the start line with 12 minutes to go. In the start zone there were anticipated pace signs, based on expected mile times. The first was 5 minutes, then they were in one-minute intervals. 4 minute kilometers are 6:26/mile, so I stood halfway between 6 minutes and 7 minutes. I was surprised only a few runners were ahead of me, the 5-6 minute zone almost empty. With five minutes to go the announcer asked us to move forward to fill the space. I ended up in the second row. This seemed too far forward for me given my running credentials, but I asked people around me what their target times were. A fast-looking guy next to me, registed 50-55 and wearing a Dolphin Southend Runner's club shirt, said 39 minutes, and some women said 6:40 miles, so it seemed good.

The start of a race is always a bit surreal, knowing the pain is about to begin, but a 10 km race is so short that I knew I'd be able to handle it without much worry. Indeed 40 minutes of pain is very simular to what I get in a typical Low-Key Hillclimb, so I have a lot of experience with efforts of this duration, even if I'd done only four 10 km races before.

KNBR's Gary Radnich finally counted us down from 5 seconds and we were off. The anticipated surge happened, and I was getting passed on both sides in the stampede. I know what a 4 min/km pace feels like and I knew I was running at close to that target, if not faster. A glance at my Garmin confirmed it: I was ahead of my 15 km/hr target. I thus knew many if not most of those passing me wouldn't be able to hold their early pace.

After a few minutes things settled out. I was following one of the woman, then my 39-minute friend, both of whom had been near me on the start line, I wasn't closing the gap, but they were no longer pulling away. I'd lose track of the woman but the guy would be a rabbit in front of me the rest of the way.

During this early, post-surge portion of the run I had three considerations. One was drafting: my cyclist's appreciation from drafting compels me to follow people if possible. The draft benefit in running isn't as big as in cycling, of course, but a back-of-the-envelope estimation yields the conclusion it could be 2%. 2% of 40 minutes is 48 seconds: that's huge. So I prefer following people.

But there's competing considerations. One is to avoid following people who are fading. With my relative lack of start-line sprint, if I'm with someone, there's a good chance I'm running faster than they are, having closed a gap to catch them. It's important to not let myself drift out of the pain zone on a run this short. Another consideration in tangents. Runners at this level seem remarkably inattentive to the old adage that a shortest distance between two points is a straight line. That means you want to, if possible, run towards the furthest tangent (corner apex) to which a straight line is on the course. If someone I was following ran a constant lane in the road, as is common, I had to balance the cost of losing the draft versus the advantage of running the shorter distance. I typically take the shorter distance unless there's an obvious headwind.

My goal was to keep my speed on the Garmin over 15: a 40 minute pace. As a fellow SF2G'er, Scott Crosby once said, "average speed is a cruel mistress". When trying to reach a given average, it's important to target being over that average, since delays invariably occur along the way. A few times I found myself following a runner, realizing I was daydreaming, glancing down at my Edge 500 and seeing a BAD NUMBER. It was time to kick myself out of my lethargy and pass. The body doesn't like a 10 km pace so it takes a degree of mental focus to keep it there.

At mile two was the first aid station. I saw only one runner of those ahead get water, and indeed it seemed way too early to be worrying about drinking, so passed it by. I almost immediately questioned this decision since I could use practice grabbing water at race pace, and resolved to take advantage of the second stop.

Despite my occasional speed transgressions I was feeling pretty good when I came to the Fort Mason climb which marked the principal challenge on the route. I run this climb a lot in training, and while it's not long enough to be a concern, it does come with a time penalty, a penalty for which my target time of a 41 minute PR hadn't provided. I knew it was important to hit the hill hard, since the grade, while well below the organizer-claimed "30 degrees" or even the likely intended 30%, was still enough that different muscles were involved. So it was good to take advantage of the opportunity to use fresher muscles, muscles which wouldn't be needed as much later in the run, by keeping a good tempo here.

My time on the Strava segment for the climb, 53 seconds, was only one second off my PR. Despite this others near me were much faster. I don't quite understand how people are able to so quickly power up short climbs. But my real issue with these climbs is what follows: the descent. Here the real runners are able to open their stride and make up some of the time lost. I can't do this: my speed only slightly increased on the descent. In all I lost 28 seconds on the kilometer including Fort Mason relative to the kilometers preceding and following.

After Fort Mason the course turned due west. First we headed out on a paved trail, then two sharp rights returning east, then a left U-turn back west, then a zig-zag left-right back west again which led to a dirt path. On this turn I accidentlly bumped a guy who, apparently drunk, lunged across my path, but this was only a brief part of my reduced pace here as I slowed from the 4-minutes-per-mile I had been able to hold on the flats. I'm not completely sure why. Tight corners are slow, the prevailing wind here is from the ocean toward which we ran, the dirt is likely slower than pavement, and since we were off closed roads at this point we were forced to dodge occasionally pig-headeded pedestrians who insisted on walking up to 4-abreast when they were obviously blocking an active race course. But it's possible the over-riding reason was simply that I was tired. Another runner with whom I had been closely matched through Fort Mason did not slow down nearly as much, as revealed by his Strava track.

At mile 4 was the second water station. They were lined up on the left side. I figured I could skip it, but as I noted I wanted practice. I grabbed the cup fine, but when I went to drink it, it went up my nose instead of down my throat. Not wanting to lose any more time fumbling, I dumped the rest down my sweaty back.

I saw Cara at mile 5, which was nice. She'd stopped during her ride around the city. I was not feeling good at this point: trying to constantly push my speed higher, ever higher, realizing the end was too close to worry about withholding anything. At this point the Golden Gate Bridge was looming large: we were close to the turn-around at Fort Point, which featured prominently in a scene in Hitchcock's great film, Vertigo. It's a popular destination on my long runs.

I only slightly flubbed my line to the turnaround due to confusion about which way we were supposed to go, and thus began the sprint back to the finish. Cara caught me again here before she continued on her way, cheering as before.

I was following a guy ahead of me, trying desparately to close the gap, able at best to hold the gap steady. My friend from the start line was just ahead of him. We kept that order through the sharp right turn leading to the finishing sprint, where I tried to get more from my legs when there was simply no more there. I crossed the line, tried to regain my senses, and slowly began walking towards the expo, where there were samples of bars and drinks and music. My time: 41:46, off my nominal goal of 41 minutes and my "reach" goal of 40-flat.

The music was quite good: a local band called "The Novelists". I had been handed a bottle of water at the finish, and I irrationally let my objection of the environmental impact of bottled water get in my way of drinking it. I gave it to an emergency response guy instead. I fetched my backpack from the start where I had a bit of water, got a bit more from one of the booths, but I really should have hydrated better.

After the band's first set, I decided I wasn't going to run the 10 km or so home from here, and moved to the long line for the shuttle buses back to the start. Eventually I got to the front of the line where I got on a bus for the trip back.

From the ferry building it was no longer a decision whether to run back or not. I didn't have a plan B. So I set off first in a trot, then in a legitimate run, and with my now-loaded backpack (it was way to hot for my sweats) I was able to do a credible run home. But this exascerbated my dehydration, and it took quite awhile after getting home before I started to feel normal again.

Overall it was a fun race. On one hand nothing I've done this year has really approximated the sustained pace of the day. But I had hoped for more. This was not a fast course: the idea that I would get my 10 km PR today was optimistic, while thinking about sub-40 was obviously ridiculous. But I still think a 40 minute time is within my reach. My speed has been improving in recent training, and not counting Fort Mason I was on target for this for the first 6 km. I'm not sure when I'll get my next chance, since my goals for November and December loom large.

Monday, October 1, 2012

Interbike 2012: virtual reality trainers

There were plenty of trainers to catch the attention. So many, in fact, that I found myself who buys these things. Riding an indoor trainer is something I have to drag myself kicking and screaming to. I admit part of this may be that I don't have an optimized set-up, with preferably a heavy-duty fan and direct access to fresh, cool outdoor air. But a big part of it is I'm so spoiled: living in the San Francisco Bay area means access to world-class cycling routes. When the weather isn't to my liking, I switch to running.

But for many people, a trainer is the only way they're going to get the experience of climbing 1st category climbs, or riding without the interruption of traffic signals or heavy traffic. For these poor, wretched souls, being able to escape the drudgery of riding in a small room without stimulation may be worth the considerable expense of some of the latest virtual-reality units.

The first such trainer in my experience was Computrainer. They're still alive and kicking, and while I didn't take any photos of their booth, their latest release is certainly impressive. They were the commercial originators of "spin-scan analysis", which claimed to analyze the L-R balance of your pedal stroke. Without any sensors in either the left or right leg specific power transmission path, however, spin-scan had to rely on the pattern of force in one full pedal revolution. There will be two "power peaks" in the full revolution, and it's generally safe to assume one is from each leg. So comparing these peaks provides a clue about how much work each leg is doing. The issue is both legs are always doing something, so there's some assumptions and approximations to be made in this analysis.

The next trainer which I found particularly notable was TacX. They recorded videos of famous climbs and produced a trainer, the Fortius, which controlled the video play rate based on how fast the cyclist moved the trainer roller, the resistance of the roller controlled based on the grade of the road. There were some issues when the video happened to catch moving objects like cars or other cyclists. Then these would also move at a speed depending on how the cyclist pedaled. But supposedly it worked fairly well.

Computrainer now has a similar system, but TacX is still there producing these videos. I think they sound fun: I'd love to climb L'Alpe d'Huez during my indoor training sessions (if I still did indoor training). But the real attraction at the TacX booth was the virtual reality and the head-to-head competition modes.

TacX
TacX trainer hooked up to its amazing virtual reality video game. Not surprisingly, the graphics didn't come out well in the photo.

The virtual reality was simply amazing. It looked like something from EA Sports. The amount of solid modeling clearly went way beyond what I'd expect could be amortized over the indoor training market, a market which I may well substantially underestimate. I tried to take a photo but it didn't turn out well.

Then there was Wahoo. Wahoo had what appeared to be physically the best trainer unit. It followed the example of (defunct?) LeMond Fitness in having the user connect the chain from the bike directly to a cassette mounted on the trainer. This eliminates tire wear. Unlike the LeMond, the Wahoo uses electronically controlled magnetic resistance, which substantially reduces the noise (the major problem of the LeMond). The LeMond uses wind resistance because that yields the most realistic power-speed curve of the most common passive resistance methods, but with electronic control, magnetic resistance can produce whatever power-speed relationship it's programmed to produce.

They also had several apps to control the trainer, but the one which most caught my attention was one which allowed the rider to ride the Strava segment of their choosing. I thought this was fantastic: virtual reality with any climb on the huge Strava database, competing against real times produced by real riders. I spoke with the developer and was amazed he had produced this "in his spare time". His biggest complaint: that when you download segment data with the Strava API you end up with the profile data from the ride which defined the segment, even if that profile data was of relatively poor quality, for example from an iPhone app using topographical data lookup instead of a barometric altimeter in conjunction with GPS. There were work-arounds: he could have scanned the activities at the top of the leaderboard for rides tagged with Garmin Edge computers, preference to Garmin Edge 800's. But then there'd be the risk of a ride with a false positive match.

But these details aside, I really liked the idea of being able to ride Strava segments on a trainer, and this alone would have been compelling to me if I was in the market. Combined with the attraction of the direct attachment to the chain and I think this trainer was my favorite of the three, no matter how impressed I was with the virtual world in the TacX software.

To seal the deal, Wahoo's trainer is it has an API, so users could write their own apps for it on iPad. The possibilities are endless.

Here's a fantastic review by DCRainmaker.