Saturday, February 25, 2012

A couple of quick notes on Jr Iditarod tracking


  1. IonEarth has switched to Google as their map source this year.  This may or may not be a win -- Bing has better place name (rivers, hills, etc.) labeling in remote areas.  I tend to like to have a topo map included as one of the layers and actually spend more time with that showing than with other formats, when it's available
  2. The individual locations all show '0 seconds' since the last update.  We know that can't be true all the time, right?  Understanding where someone really is is usually a function of knowing where they were at a given point in time and knowing how long ago it was.  For example, if a tracker point is about 10 minutes old it's possible that someone's traveled about a mile since then, more or less.  There's a young man who's way, way behind the rest of the pack according to the tracker and I can't tell if the issue is that his location hasn't updated or that that's really where his tracker is right now.
  3. Related: there's no history for individual mushers.  
  4. This made me laugh:

I don't know what really happened here but the best guess is that it looks like a landowner right-of-way issue.  We get those, even in Alaska.

Iditarod, and on thinking about how to think about thinking about location data

Iditarod is coming up in just a few weeks.  I think that they're the first race I saw use actual GPS trackers (as opposed to the Can-Am's map projections).  Their trackers show you where the teams are on a map, the speed at which they're traveling (and this may actually be calculated by their GPS, but I tend to think not - more on that in a subsequent post), the temperature (always entertaining to know in wintertime Alaska, but it appears from the trackers on the Junior Iditarod that they're no longer displaying that data), etc.  IonEarth is an outfit with some very smart engineers who've worked out how to build a gizmo that takes periodic GPS readings, uplinks data, and has a battery that can survive doing this in subzero temperatures for two weeks while not being overly large and heavy.  It's pretty clear that they've figured out how to do a very good job of showing you who's where.

I've been writing quite a bit this winter about Trackleaders.  That's mostly because they're what's being used by most races, but it's given me an opportunity to think about what they're doing and how it differs from what IonEarth is doing, besides just the obvious use of inexpensive commodity hardware and a commercial tracking service.

I'm going to digress for a moment:  When I was in graduate school I TA'ed classes in introductory statistics and in research methods for library school students.  Most of the students had humanities backgrounds and were somewhere between disinterested and terrified when it came to quantitative tools.  Their understanding of the task at hand was that they had to take a chunk of numbers and calculate certain values from them.  I found this somewhat disturbing, since my understanding of statistics is that it's just a set of tools to help you summarize the data in a way that can help you figure out how things are going, answer important questions about whether or not you ought to be surprised, and help you develop intuitions about what's really going on.  It's unfortunate that the way the material was presented at that time tended to underscore that what matters is the arithmetic and tended to gloss over the more intuitive, visual aspects of it, because I think when you put people off from material like that they basically lose access to a very handy set of tools.  Obviously, for researchers and actual statisticians it's a lot more complicated than that, but for most of us it's just a way of grokking the bigger picture.

And this brings us back around to Trackleaders.  I have no affiliation with the company -- I just really like what they're doing.  And from my perspective what they're doing is a lot more than just showing you where teams are on a map.  Scott and Matthew, the company's founders, are bicycle ultraracers themselves, and so they've got a natural interest in representing the race and how it's unfolding, rather than just solving the technical problem of plotting individual locations.  That's come out in some very handy tools like the Race Flow Chart, and the speed charts on individual pages (if you haven't spent time with those, you might go back and take a look).  I think Trackleaders strength is in giving us tools to understand how a race is really going, while IonEarth is doing a top-notch job just giving us reliable location data (and as we've seen from other races, that's not easy and not everybody can be anywhere near as reliable).

I'm hopeful that IonEarth will start to provide more interesting tools.  This year it looks like the major advance over previous years is the ability to receive personalized updates on up to five mushers.  That's nice, but I'm not sure it's going to help me understand the bigger picture better, or tell me how the race is really unfolding.  Obviously it's all a work in process and things will be changing over time, and IonEarth has some top-notch people and outstanding technology.  I'm really, really interested to see what kinds of directions they decide to take the technology, and whether or not they adapt their business model in response to challenges created by lower-cost competitors.

Sunday, February 12, 2012

Are Quest mushers tampering with their trackers?

On the Quest Facebook page there have been several sideways allusions to the possibility that Quest mushers have been tampering with their Spot trackers, and the allusions have tended to have a "wink-wink, nudge-nudge" tone to them.  Today the comments became more explicit, more annoyed, and less "ha ha ha"-ish, when someone raised questions about the reliability of the trackers and said that they'd be willing to pay for more reliable ones (the Iditarod uses heavy-duty, purpose-built GPS trackers and charges fans who want to follow them during the race).  

Before going into a more detailed discussion, I'd like to make one thing clear: it's pretty obviously the case that mushers are tampering with their trackers, whether it's covering them up so they can't transmit, pulling batteries, turning them off, whatever - I don't know specifically what they're doing but the extremely high "failure" rate during this race is not matched by any other races going on at the moment.  It may be the case that they're using different hardware for the Quest than they are for the other races, but it seems unlikely that they're using different Spot Messengers for the Quest than they've used for any other race this season, or that it's just the result of random chance that there are more outages in the front of the pack than in the back.

To be clear, not every outage is the result of monkey business - there have been some honest tracker failures during the race.  But this post is about the ones that were the result of tampering.

The trackers are required equipment during the race.  A team cannot check into a checkpoint without theirs, just as they can't check in without a sleeping bag, snow shoes, axe, etc.  The rules don't say that the tracker can't be turned off, as far as I know.  But let's say that there's going to be a new rule that the trackers must not be interfered with.  How would you determine whether or not there was a real failure, or if someone hid their tracker in a cooker for most of their trip?  How can you tell whether or not a tracker was covered deliberately?  You can tell if batteries really are dead but you can't tell if someone's chosen to swap in some dead batteries (at least not without marking and identifying "official" batteries, a logistical headache) to make it look like the battery died on the trail.  The possibility of developing a mount that would hold the tracker in a place that couldn't be covered by accident has been raised but the first time I mentioned it to someone who's done that trail his immediate answer was "It would snap off very soon."

Instituting rules you can't enforce is a pretty terrible idea, and I don't think the Quest should penalize teams for tampering with trackers unless they can provide very strong evidence that there actually was tampering.  On the other hand I do think that over the longer term this is not good for the race.  I believe that GPS tracking is largely responsible for transforming distance mushing into a spectator sport, and that that's resulted in increased donations and merchandise revenues, not to mention less tangible things like community-building.  I hope that someone can figure out a way to deal with this.

Thursday, February 9, 2012

What do the horizontal squigglies in the race flow chart mean?


This is not the first time on this blog that we point out how much we like the Trackleaders' race flow chart. But it has its own surprising details. For example, if you look very closely, you see little horizontal segments that make me scratch my head a bit. Here is one on Allen Moore's chart, just a few status updates out of the Central checkpoint:


Remember, horizontal lines should mean that the team did not move, as the y-axis is location (and horizontal means location doesn't change, so they stay in the same place). Coming from the left, we see Allen's long horizontal line that signifies the end of his stay in Central. Then he takes off - the first check-in is from still very close to the checkpoint, and then he seems to fall into his rhythm. Except that between the second and the third track point, the curve is flat. For how long? Well, we can eyeball this as about 0.2 h plus/minus - that's close enough to 10 min to make me believe the problem is exactly 1 check-in interval long.

Now looking at Allen's individual track, he didn't stop for about 10 minutes (the time between two check-ins). Here is the relevant segment (blue arrow):


There are a few interesting things we can see by looking at this section in detail. The first is that the segment doesn't actually correspond to the usual 10 min check-in interval. It already looks too long for that naively. The endpoint was at 10:29:58 pm (this is shown on the screen shot), and the start point at 10:09:05 pm (you have to believe me here or check it yourself). Which means that right in the spot where the race flow chart has a 1-interval long horizontal squiggle the actual track is missing a point! This is what makes me think that the way the software draws the race flow curve, a missing data point gets replaced with the same y-value at the next point, thereby creating an artificial flat section. Well, that's my current opinion, and I'm sticking to it for a while, but Melinda is corresponding with Matthew Lee from Trackleaders, and he seems to think this explanation is wrong -- I'd be curious for his take on it!

The other interesting thing is that looking at the windy trail (on the tightly meandering Birch Creek) after Central, the track length gets cut across pretty badly across the check-on intervals. It would be nice to look at a speed chart across this section of trail.

Plug-it-in calculators and improvised data

A number of Facebook fans have noticed that Allen Moore's speed between Forty Mile and Dawson was fairly remarkable:



Here's why:  the Quest staff have said that not all teams were checked in and out of Forty Mile (which is not a checkpoint, but rather a hospitality stop), but that their software required that a time be entered, so they made something up when real data were missing.  In Allen's case they entered a value that has to be considerably later than when he actually left.  This means that he supposedly covered this distance in less time, which in turn means that he's calculated as traveling faster.  The Quest website gives the distance as 48 miles between Forty Mile and Dawson.  If he covered it in the roughly 3 hours that the Forty Mile time out allows, he indeed was moving over 16mph.

On the other hand, given the wacky time it's almost certainly the case that if we did the arithmetic to find out how fast he was traveling between Eagle and Forty Mile we might see that he was moving disturbingly slowly, so let's do just that.  The Quest website gives the distance as 99 miles, which they say he covered in nearly 27 hours.  That would put him traveling in the ballpark of 3.6mph to Forty Mile, which is considerably slower than his overall average speed and extremely unlikely.

I guess the good news is that we've found out that they're using software to calculate speeds rather than using error-prone humans.  The less good news is that error-prone humans can always find ways to introduce odd things.

I think the bottom line is that the software needs to be smarter about dealing with missing data, but that's so often the case.  Here's hoping that they fix this, as well as the web development company allowing Quest staff more control over what's on their website -- the Maren Bradley "scratch" at Two Rivers could have been taken down long before it was if the staff had more editorial control.

Anyway, missing data: blech.

Wednesday, February 8, 2012

Arriving vs checking in

Marcelle Fressineau has a blog and one of her handlers recently posted an update on how her race has gone so far.  In it she describes arriving at a checkpoint only to find that she does not have all of her mandatory gear:
Au point de contrôle à Central, Marcelle découvre qu’elle a perdu sa hache. Impossible de passer la ligne d’arrivée. Chaque musheur doit avoir en sa possession tout au long de la course un sac de couchage, une hache, un couteau, un cuiseur, des booties de rechange et le dossier vétérinaire des chiens. Nous lui apportons la hache qu’on a dans le camion, mais elle n’est pas réglementaire, elle est trop petite. Après quelques recherches, les handlers de Brian Wilmshurst (dossard no. 10) nous prêtent la leur. Ouf! Marcelle peut signer et passer le point de contrôle, mais récolte 30 minutes de pénalité. Sans cette hache, sa course s’arrêtait à Central.
Here's what the rules say:

If a driver loses a required article of gear between checkpoints, he/she cannot check in at the checkpoint until he/she has acquired and replaced the lost item. In the event of accidental and unavoidable loss along the trail, the driver will be allowed to replace the missing item(s) from a public source at the next checkpoint before checking in. The driver may also obtain items from a private source with the approval of the Race Marshal or Race Judge and a time penalty of thirty (30) minutes at Dawson or the last designated mandatory stop.
Arriving at a checkpoint doesn't necessarily mean that someone can check in.  While they don't say how much time elapsed between the time she arrived and the time she checked in, if it was, on average, more than 5 minutes it would likely have shown up on the tracker and if it was more than 10 it would have (barring the possibility that the tracker wasn't able to transmit data successfully).

Also, "well done" to Brian Wilmshurst handlers for their sportsmanship.

Monday, February 6, 2012

Traveling together

It's still really early in the Yukon Quest.  My guess is that there's not a lot of racing going on right now but that there's a lot of positioning so that they'll be set up to make a move later in the race.  There's always the possibility of someone making a surprise move, like Lance Mackey's famous, heroic run from Nulato to Unalakleet in the 2010 Iditarod, but it's the nature of surprises to be surprising (you can quote me on that).  Competitive teams want to make sure that other contenders can be reeled in when the time comes to make a move, so you sometimes see them traveling together and keeping an eye on one another.  That appears to be what's going on at the moment.  Here's what it looks like on the race flow plot:



We're seeing six people traveling in three pairs, plus Sonny Lindner running alone.  At the time this screenshot was grabbed Hugh and Brent had been traveling together for 3.5 hours.  Abbie and Allen left at about the same time.  Allen had a brief, er, "diversion" leaving Circle, but he caught up and they're now traveling together.  The final pair is Kristy Berington and Lance Mackey.  With each of these duos there are occasional passes, but the overall slopes of the line (i.e. speed) is remaining more-or-mess constant while they run down the Yukon towards Eagle.  What we're seeing is strategy, not flat-out racing.

Sunday, February 5, 2012

Hi, Darren

For some reason Blogger is eating comments, but I did receive email containing one you posted.  You're right!  That's really interesting, and I have not one single clue what's going on.  I've asked Matthew at Trackleaders if he knows, and while I can guarantee he's ridiculously busy with this and other races I'm hopeful he'll have some insights.

For everybody else, this is what Darren noticed:
Here is a question. At race mile 144 on the race flow chart there are several mushers that have little plateaus - mostly two points. Now this might be a short stop, but these occur 4 miles past Central where teams getting drop bags stopped a short time that look just like these little plateaus too. So I thought, okay maybe they go a few miles out the checkpoint and repack or snack or something so the dogs don't get the idea they are staying at the checkpoint. After a few more teams did the exact same thing in the exact same spot I wondered about the reality? Could all these teams be doing the same thing in the same spot?
I looked on the map to see were mile 144-5 was and it's on the road to the hotsprings - okay? I then checked a few mushers individual tracks to see where they had stopped, should be two points together right? Well I did not find them, just a string of points 10-minutes apart, seemed normal.
So are these short plateaus a glitch? We might have to be careful not to interpret too much from these if they cannot be proved otherwise?
He's right, and I'm really curious about what's going on there.

Let's do a little arithmetic

We're at about hour 32 of the Yukon Quest, a race that you can figure taking a little over 9 days in this direction.  That is to say, we're about 14% of the way through timewise (and about 16% in terms of distance), which means there's a lot of room for things to change, and to change drastically.  There's been a lot of speculation about who's leading the Yukon Quest and excitement about changes in position, and I thought it might be worthwhile to do a little arithmetic to figure out just how drastic a change would be needed to cause a major shift in outcomes.

So, right now Allen Moore is at mile 161.1 on the tracker, and is leading.  Gus Guenther is further back in the standings -- the Quest "current standings" page has him in 13th place.  He's currently at the Central checkpoint, or mile 140.5.  So, a little less than 21 miles separates Allen and Gus.  I'm not sure what the exact trail mileage is and the trackleaders site doesn't really say, but let's go with 991 miles, since that's what the Quest website says.

So, let's start with some assumptions.  Figure that it takes roughly 220 hours total for the finisher to arrive in Whitehorse, based on past results.  Now, let's assume that the current leader, Allen, holds his position and is the guy that arrives at hour 220.  Let's subtract the 48* hours of mandatory layover yet to come, to understand what kind of trail speed he'd need.  With 188 hours to go, minus 48 hours for layover, we'd get 140 trail hours.  To cover that in time to arrive at hour 220, he'd need to travel at an average speed of 5.9mph.

What about Gus?  Let's say that he's going to beat Allen by an hour -- how much faster would he have to travel?  He'd have to cover 851 miles (he's at Central) in 139 trail hours.  That's an average speed of 6.1mph, or 2/10 of a mile per hour.  Or it may be clearer to say that he'd need to be a little over 3% faster. And that's for arriving an hour earlier.  He could arrive a minute earlier and win.

If it were just that simple this small a difference with so much trail left, current positions could probably be considered to have virtually no predictive value.  But it's a dog race with huge human factors, major weather and terrain considerations, and so on.  Someone might have a strategy of holding back early and keeping their dogs rested, calculating just how much trail they've got left to reel in the leaders.  The weather or terrain may favor one team over another.  If it gets warm, teams from interior Alaska will have a harder time, while if the temperature plummets the dogs (and mushers) from down south are going to have more difficulty holding their race together.  You've got rookies who've never seen the trail running against people who've run the race many times and know this trail well.  There are so many different factors that affect outcomes.  Current trail position is just one, and arguably not much of one -- at least not yet.

[* A discerning Quest fan pointed out that while everybody knows that there is a mandatory 36 in Dawson City, YT, but may be less aware that there's also a mandatory 4-hour layover in Eagle, AK and a mandatory 8 in Braeburn, YT (the last checkpoint before the finish).  See this page at the Quest site for details.]

What it looks like when something ain't right

Tamra Reynolds was pretty persistent on the Quest Facebook page about asking why Hugh Neff's data looked weird.  She made a good catch that a lot of us didn't see - if Hugh really did park on Eagle Summit for 90 minutes, it was very unlikely he could have traveled fast enough to come into Central so close behind Brent Sass.  Something was wrong, but what?

If we'd taken a good look at the race flow chart it would have been clear immediately what the nature of the problem was:



but we didn't, so we didn't see that according to the data coming out of Trackleaders, Hugh traveled the section of trail between shortly before mile 120 to shortly after mile 140 at over 20mph.  He's got a fast team, but not that fast.  So, the possibilities were:

  1. he covered that distance in more time than it shows,
  2. he covered a shorter distance in the time period shown, 
  3. Trackleaders's software is buggy, or
  4. he actually does have an insanely fast team
The fourth option seems impossible.  I tend to have a lot of faith in the Spot devices's reports of time and location so I tended to rule out the second option.  The third option seemed possible but unlikely, given their history of delivering just this sort of information very reliably, and to be honest the first option seemed unlikely as well, given the reliability of the Spot trackers.  

Well, as usual it turned out to be a very real-world, physical error.  Hugh apparently had a wild ride down the back side of Eagle and dropped his tracker, only to be reunited with it later.  Mystery solved.

This does point up once again how good data visualization tools allow us to take in a lot of information with just one glance.

The race flow chart and terrain challenges

As I've mentioned in earlier posts, I'm fascinated by the Trackleaders race flow chart and think it might be one of the handiest tools for understanding what's going on on the trail.  So, I thought I'd take a look at what happened around a challenging bit of terrain.

Everybody knows that Eagle Summit can be extremely difficult.  Rosebud gets less press, but it's very difficult, as well.  At the mushers' meeting the mushers were warned about it, and veterans warned rookies outside the meeting.  Rosebud is early in the race, at about mile 90 according to the trackleaders track.  So, here's what the race flow chart showed:


To provide a little more background, the x-axis is the time on the race clock -- how many hours we are into the race.  The y-axis is the miles traveled.   What this means is that when a team is stopped the line is perfectly horizontal (flat), and the steeper the line is the faster the team is traveling.  Also, because the plot shows us where people were at what time on the race clock we can understand where they really were in relation to one another.  When lines cross, that's a pass.  If the lines look like they've merged, the teams are traveling together.  When the line goes perfectly vertical a team is traveling infinitely fast (and from what I saw in the staging area yesterday, we could possibly see Jake Berkowitz's line do this).  In this screen grab you can see everybody's lines go flat at about mile 113 - that's the mile 101 checkpoint.  As I said, fantastic tool.


So, what happened at Rosebud?  Well, one of the things we see is that Mike Ellis's and Kristy Berington's lines go flat at about race mile 89, and they stay flat for about an hour.  Mike said he was going to camp for a bit before going over so it's a little surprising it's not a longer stop, but not that surprising.  Then at about 15:15 on the race clock (the "time" axis shows decimal fractions, so "15.5" is 15 hours and 30 minutes) they start moving, with Mike leading.  Again, not a surprise -- Mike is a veteran and knows the trail.  Then they stop again for about 20 minutes on the other side before moving along.  It looks like Brent also took a break at about mile 89 before climbing Rosebud, without stopping on the other side.


Another interesting thing that was going on was that Kyla Durham was stopping often, and for short periods.

The front of the pack is going over Eagle Summit as I write this, and I am sure that the race flow chart is going to be fascinating.  Some mushers stop at the top to rest their dogs or to put rough locks on their runners, while others feel that the best way to deal with the precipitous drop on the downhill side is to just go for it and not give the dogs a chance to get perky again.  Here's a chance to find out who's in which corner.

Saturday, February 4, 2012

The Yukon Quest start



I write about data, but at core this whole thing is really just about the dogs and the mushers and the trail.  Here's video I recorded in the staging area as teams prepared to enter the start chute.

Friday, February 3, 2012

Blogs

I said that I'd get around to posting links to good Quest-related blogs, so tonight I took a look at what's in my RSS reader and just groaned.  There's a *lot* out there.  Most of it tends to be mushers keeping fans and sponsors up-to-date on how training is going, what day-to-day life is like, and so on.  A lot of them are great, some are not, and I'm a little short on time.  So, instead I'm going to provide a link to a new blog from KUAC reporter Emily Schwing.  It's KUAC's Yukon Quest Coverage.

This one's a little different because it's by someone who's writing about the race and about the mushers, but not as a musher.  She's a journalist and I think provides the best reporting and "color" coverage out there, and it's all original content.  While the KUAC blog is new she covered the race last year at her own blog, "Current Resonance."  With the race starting in 12 hours (ack!) this might be a great time to take a look at her posts from last year for a reminder of how things went.  It might jigger your expectations for this year a little.

Basically, I think that if there's one blog to follow during the race (on top of the Quest website, the Trackleaders website, the Facebook Quest page, the mushers' Facebook pages ... ), this would definitely be the one.

Testing 1, 2, 3

Fairbanks is definitely amped up for the Quest start.  There seems to be more visibility, more fans, more everything this year than in the past.  This morning on an oldies radio station they did a lead-in to I-don't-remember-what with "We've harnessed and ready to run our champion string of classic rock hits."

I just took a look at the Quest page at Trackleaders.com, where they had a scrolling window showing updates as they arrive:


Looks like there's been some testing going on and they're nearly ready to roll.  I was hoping that Trackleaders would soup up the interface to the race flow plot so that we could select which teams to show, but even with the display limited to the frontrunners it's going to be an incredibly interesting source of information about what's happening on the trail.

One heads-up: I think that the "Native Chukchi)" in the musher listing on trackleaders.com is pretty clearly Michael Telpin.

Wednesday, February 1, 2012

Following the Quest

The Yukon Quest starts Saturday!  The Quest has really led the way in making use of clever, inexpensive tracking services (Trackleaders) and leveraging social media.

I thought it might be worth a short post describing options for following the race online, so people can make sure that they're set up to take advantage of everything that's available.

The official website
First, the Yukon Quest website itself.  Their race updates will be the authoritative sources on checkpoint arrival/departure times, accumulated rest, etc.  The GPS trackers will also show times but they're being derived from tracking data and because there's some heuristics involved in determining when a team actually arrives (they have to be within <n> distance of the official checkpoint location, for example) don't be surprised if there are differences between that and the Quest website.  Trust the Quest website for this one.

The Quest website also includes trail and route descriptions and a huge collection of other data.  Poke around - you'll be amazed by what's there.

For German speakers, check out the German website, here.

Facebook
The Quest communicates directly with fans on Facebook.  Fans also communicate with each other, but again, information from the Quest itself is authoritative and pretty much everything else should be taken with a grain of salt.  The Quest has been excellent about posting photos and videos and providing links to interesting media and articles as the race is underway.  It's probably the main hub for both information from the Quest organization and posts from enthusiastic fans during the race.

Personally, I've found that there's an awful lot of noise and misinformation from enthusiastic fans and while I appreciate the community and their enthusiasm for the race it tends to crowd out posts from the Quest organization itself and make it difficult to follow "official" news.  Fortunately Facebook lets us control our own environment so that we can see what we want without stepping on anybody else's ability to post and participate.  Options include blocking too-frequent posters as well as choosing to not show posts from anybody but the page owner (i.e. the Quest organization).

Trackleaders
I tend to prefer to go directly to the Trackleaders website rather than following it through the Quest website.  Probably the primary reason is that if you use the race analysis tools (the race flow plot, the projections) you'll end up there anyway, but it's also the case that the map layers on the Trackleaders site include a topo map and the ones embedded on the race websites don't.

This year Trackleaders is also providing a link to a mobile website, and you might want to create a bookmark in your phone's web browser.  To be honest I find the user interface kind of clunky but just having it available is a huge step forward and I absolutely love that they made this available.  The URI is http://app.bluedot.mobi/.  Definitely check it out.

Twitter
You have a few options here.  The most obvious one is to subscribe to @theyukonquest, but they won't be the only ones tweeting about the race, so you can also search for the #yq2012 hashtag through your Twitter client or through Twitter's search interface.  Because not everybody's fastidious about hashtag use I also have running searches on "yukon quest" (include the quote marks).  

Blogs
Partipants' handlers and/or family members have turned into fierce bloggers and a fun source of information.  I hope to post the listing from my RSS reader before the race starts but in the meantime you can search blogs by going to Google's blog search page and starting a search there.

News
Alaska and Yukon newspapers do a fantastic job of covering the races, and it's a good bet to check in with them at least once/day for both news and photos.  The must-reads include

And if you're a total Quest junkie, you can also do a newspaper search on Google, at http://news.google.com.