Training Log Archive: iansmith

BOS-KEF
Remembered passport.

C • d'oh 1

Prelim; a bit slow at the beginning. Painfully slow on the long run across the field to 10 for some reason, though I took the tunnel correctly. I made a 15s mistake on 13 overrunning it, then tripped over a rock and faceplanted hard running downhill out of 14. I was shaken up, banged my knee, and lost about 40s on the trivial last two controls.

Final at Franklin Park. I started out ok, though still painfully slow. My execution to 2 was sluggish, and I chose to suboptimally go straight to 3 rather than around on the trail. My route to 5 was also incorrectly straight instead of using the trail. I caught sight of Will, Giacomo, and Christian briefly while I was approaching 7 and they were leaving 8. I executed 9 poorly (5s hesitation, thicker vegetation than expected), epically botched 10 (20s trying to go straight before backing out and running the right trail route), and screwed up my attack on 11 - 10s. Isak was untouchable, but I think I could have run 14:30 with choices that were less idiotic. The fast kids all had yolo controls, so I squeaked into second.

Prelim B - Splits

Final - Splits

Basic skills morning at Magazine beach practicing (1) reading each leg completely, (2) looking up and seeing my exit and (3) smoothing my route. Ran faster than I should have given the day's workload.

C • Ran faster 4

Isak's diabolical one-way O.

Sprint O-tervals at Dana Park.

Warmup jaunt.

Sprint race at North Point Park. I was painfully slow, though I navigated comparatively cleanly. I finished about 75s back of Laraia. Sigh.

North Point sprint - Splits

2016 New England Champs
Candlepin game: 32- 8-- -3- 6-4 5-- --3 242 511 71- 513 = 66
Ten pin: 7- 5- 7/ 81 81 7/ 8/ X 8/ 81 = 133
FBP: 7.6
No 1-pin shots :'(

C • Really? 8

Run from the Springfield bus station to Spartan auto shop where das auto was waiting. I could have taken a local bus, but I figured I might as well get my workout in. I wore a long shirt and a hat to protect from the sun, and I must have been dehydrating. Near the end, I downed a liter of gatorade and barely noticed the dramatic influx of fluid.

C • Das auto 8

Event: Boston Sprint Camp 2016
 

Pine Banks Park-O. Extremely sloppy.

Control pickup.

I decided to polish off the day with a run home. My route was suboptimal.

Run from the T to Pine Banks.

While I prefer to let the WOC selection thread die quietly rather than resurrecting it into trolly flame again, I drafted a response to Eddie's last post, chiefly disambiguating the definition of bias:

Hi, Eddie - I find myself with an unplanned bit of time and a certain degree of masochism, so I'll dive back into this discussion. I apologize in advance to the entire community and you for propagating this discussion, but I think your points deserve a serious response. Also, I have no actual orienteering or administrative responsibilities to distract me, unlike the actual ESC.

0. I'm fighting very hard to avoid a pedantic exegesis of G.1.7.1, partly because a small part of me died on the inside when I wrote "I'd love for other people to weigh in on G.1.7.1." It has become the equal protection clause or maybe second amendment of the US team. I will restate that I think it's a spectacular leap to infer from "primarily based on the results of a team selection competition" that (1) the review panel can't exist (2) there is no provision for petitions (3) the selection process must therefore be objective or even (4) at least 50% of the team composition is identically taken from the sacrosanct (squashing function applied to) results from the trials races. The axiom underpinning your entire argument is that if one athlete finished ahead of another athlete in the list of (a squashing function applied to) trials results, that they therefore are more deserving of a place on the WOC team. I categorically reject that axiom. (Cute hypothetical counterexample: Matthias Kyburz becomes an American, mispunches in the long team trials race while I do not: therefore, I am more deserving of a WOC Long start.)

1. My first observation is that we're using bias in two different ways. Selection bias formally relates to choosing a subset of data points, people etc such that the population is skewed. Randy used this to mean that we're not correctly estimating the fraction of involved people who think the WUOC selection process had a conflict of interest was negatively affected by possible stigma.

I'm using bias in the statistical sense. The goal of any procedure for choosing a US team to WOC is to maximize our results at WOC (which is poorly defined - are we trying to say minimize the sum of our places or maximize the probability of a top k result?). We cannot a priori know the probability distribution of results for a given subset of athletes, so we have to estimate this function. There are numerous ways to do this. An objective selection method - ranking athletes by some non-stupid function on trials race results - is unbiased in that it converges in expectation to the true result. In the limit as the number of races goes to infinity, the results of selection races converges to WOC results. (Actually this is probably not true given differences in trials races compared to the WOC terrain etc, but we'll call this as small as is practical.) A subjective selection method - which basically means people using judgment to incorporate prior knowledge - is biased in that it does not converge in expectation to the true result. As an illustration, if someone trying out for team trials improved dramatically recently in such a way that is not reflected in the various data the selectors use, their prior performance will be less than their true value.

By the variance of the estimator, I mean some notion of the size of error between the true value (results at WOC) and our estimator. This could be mean squared error, say. I contend that a team selection process that uses some subjective assessment of the athletes will have less variance - i.e. better approximate the WOC results - than a purely objective system based on three races, and that this is desirable precisely because we want to choose the best team for WOC. Fairness, in this case, comes at the expense of a noisier measurement. Colloquially, a strictly objective trials race process is more accurate but less precise than an informed subjective process.

2. On the ranking score: my understanding is that various people argued that (1) the post-qualifier WOC rules make starts a scarcer resource, forcing us to explicitly assign people to races as part of the team selection and (2) the ranking score discards race type specific data. In particular, ranking scores for sprints are typically lower than those for middles and longs, which would bias a ranking score against the sprint. I'm paraphrasing, but I agree with this point.

I'd actually argue that any system incorporating a ranking score violates your interpretation of G.1.7.1 - which makes no mention of a ranking score or using results from non-trials events. Therefore, all US WOC teams selected using the ranking score are null and void, and I demand the IOF vacate all US WOC results until this blatant abuse of power can be investigated fully. But seriously, the ranking score is analogous to a prior on each athlete; it introduces bias in the statistical sense - trials rankings lists using the ranking score will not converge in expectation to the true result in the present.

3. "My two cents is it doesn't matter much how the last member (or two) of the team is selected; their value over a replacement runner is not high."
Sure, from a performance point of view the last few runners are all nearly equal.

This. When I ascribe "value" to a runner, I'm talking about performance, not anything personal. Kyburz's value as an orienteer at WOC vastly outstrips my own even though I'm super cool and fun and stuff precisely because his projected results at WOC will be far better than mine. It's pretty typical in the US for the last team spot to have several people of very similar skill; Boris mentioned that he made the team in 2003 by a fraction of a point. These tiny margins are not statistically significant. Put another way, I'd guess that if we could repeat the team trials races many times, the fifth and sixth runners (say) would be almost equally likely to finish fifth depending on who had a bad day or whatever. In that sense, I don't think it matters very much how that person is chosen. Were I on the ESC, I might start to ask things like which athlete has more potential for the future - i.e. trying to forecast future WOC results, or which athlete did a better job conforming to expectations like logging training, showing a commitment to the sport, etc. This "one is as good as another" thing is only a problem if you begin by assuming finishing higher implies selection preference, no matter how small the margin.

4. Regarding the RP makeup, yes it does currently satisfy the rules, but the reasoning I gave (an odd number is better than an even number, and more is better than less) still stands.

I agree, and I therefore suggest the fifth and open Review Panel spot be filled by a Paul the Octopus-esque selector. For instance, in case of ties, we could take Pavi's dog, offer him two dog treats, each labeled with an athlete's name, and give preference to whichever treat gets eaten first.

I would also note that I am aware of many people in the US community who are in favor of purely subjective systems, like the Canadian one where a set of selectors name people to starts - informed by races, but not constrained by any system. In fact, I think I am more in favor of some objective component to our process than most. But all those people are busy training and eating kebabs, so I am their proxy.

C • +1 4