resources
visualizations
presentations
lunch break
Distractions and amusements, with a sandwich and coffee.
fun + amusement
#googlemapschallenge
The Google Maps Challenge — Longest Google Maps Driving Routes
Routes updated 13 June 2017.
CHALLENGES
1 — Longest land leg
3 — Longest land route with ferries
other visualizations
Take a road trip to clear your mind. Take in a few sights and bring home a spoon or other collector's item.
According to Google Maps, how far could you go?
And the most pressing question: can you reach the sorrow at the world's end? Yes you can.
the challenges
Each of the challenges below involves finding points A and B that yield the longest driving route in Google Maps. Each challenge has its own parameters, but certain rules apply to each challenge.
- the route A→B must be generated by the Google Maps algorithm—it cannot be manually adjusted
- the shorter of A→B and B→A must be used
- when multiple routes are available, the shortest must be used
- avoid highways and avoid tolls options must be off, except for the 3-point route challenge
- the "no ferry" land route may include a ferry to cross a river or any ferry shorter than 10 km
- the 3-point route cannot double back along the same path at any point on the route
- the route must be generated by typing in the waypoints, not by dragging them
Any solution to the challenge will surely have a shorter route (not available to the routing algorithm) as well as many more longer ones (duh—it's always easy to pessimize a route).
Updates to Google Maps are frequent and unannounced, making most of you read here historical artefacts. I try to periodically make sure that the current challenge routes are at least functional.
This topic has been previously discussed on xkcd forums.
continuous updates to Google Maps — old routes unavailable, records reset
As Google Maps updates the routing network, some of the old routes are no longer available, or significantly shorter. This maps challenge page may therefore be out of date.
Don't be surprised if links to old routes show a significantly different distance, or point to a route that no longer exists. Such links, or out-dated entries, are annotated as historical.
30,000 km limit broken—again
historical Thanks to a route by Dean Sych, the 30,000 km limit has been broken again. And how! By over 3,600 km. Thanks Dean!
Previously, because of new routes through central Africa, route distances dropped by about 6,000 km. Whereas before routing avoided Congo and Algeria, they now go right through them. This made a lot of the 30,000+ km routes much shorter.
historical The longest route used to be 33,540 km from Quoin Point, South Africa to a dirt road in Indonesia. It starts with "Head northeast toward R317." and ends with "Turn left". After Malaysia, it's mostly ferries. Unfortunately, it's gone. Also, routing in Africa is now more optimized, shortening all the trips from south Africa.
historical (Paren' no longer available as a destination) | The 30,000 km limit has been broken by a route from Paren' to Pearly Beach. This was furthered by discovering the bizarrely remote Chimchememel', Russia (from Chimchememel' to Danger Point).
historical (Uelen still not accessible) | The next milestone for a route with ferries is 32,000 km. Unless you have the money to build a road to Uelen, this new limit is a significant challenge. Interested individuals should start digging immediately.
historical Many of the interesting past routes are now gone—the world is getting boring. This fun 191 hour drive from Portugal to Malaysia, shown below, is no longer available. Too bad. Stop off in Turkey to go to the bathroom. Pick up a few aluminum centrifuge tubes from Iran, too. That sounds like fun.
change for the worse
The details of the routing algorithm is the epitome of Rumsfeld's unknown unknown. It used to be that Portugal to Malaysia was a very doable and relaxing 8 day drive at 15,787 km. And you managed to avoid China.
Later, the route included China but excludes Iran. Apparently, we're no longer on driving terms with Iran. If you want to go from Lisbon to Singapore, you no longer need your Iran for Dummies book.
Still, the updated routing was longer and burned more gas. Prius C drivers with advanced road endurance skills need apply. You know who you are.
The 20,000 km all-land route limit still stands. Africa's complex routing may provide a solution. Or a local conflict that severs a road or two.
Blind Spots and Humor
If you play with routes in Google Maps you'll quickly notice that some parts of the world do not appear to be connected to the smarts of the routing algorithm. For example, you cannot drive from Bejing to New Delhi (As of 31 Aug 2015 this is no longer true. A 5,559 km route exists). These holes in the driving fabric pose a challenge in finding long routes.
historical (no more kayak routes) | Google's subtle humour can be found everywhere (though apparently not anymore, since kayaking has been removed from maps and replaced by decidedly un-funny flight routes), such as in step 9 of this Seattle to Hawaii route, which states "Kayak across the Pacific Ocean — 4,436 km". If you have endurance training, you might wish to continue kayaking to Tokyo, for another 6,243 km. For the purpose of this challenge, kayaking is not allowed.
These examples amply demonstrate that the internet was created by people with humor and is used largely by people without it. I say this because the kayak option has been removed. Doubtlessly because someone tried it, got hurt and the litigation lawyers got involved. It's always been the case that the 1% ruins it for the 99% of us.
going to south africa?
South Africa's Quoin Point is (or was, depending on the current routing) a common starting point for many of the routes.
I'll have to visit this place next time I'm in South Africa and perhaps start driving north.
Recently, I recieved a note from Charlie Kelly who wrote to thank me for inspiring him to put together a list of what to do in South Africa.
Darién Gap
No routes from North to South America exist because of this boggy marsh.
Routing Changes
As new routes become available, long trips become shorter. For example, introduction of a route across Niger and Algeria cut the original land route with no ferry record from 18,260 km to a mere 15,576 km.
Other Google Data Munging and Visualizations
If you are interested in visualization and information, explore my global visualization of Google searches by language and find out where in the US people are searching in Chinese.
For the morbidly curious, of interest might be all the really stupid questions people ask Google.
mobile users
Google Maps routes linked to from this page do not appear to work on iPhone's Safari browser (I have not tested iPad or iPod). A "driving direction not found" error appears. Rest assured, these routes do exist, and can be viewed on a browser on a PC or Mac. Weird.
The routes below are the current answers to the challenge. Do you have a better (longer!) route? Let me know.
starting points
terminus station
Talon is at the end of R481, which goes from Magadan to Talon. This road forks off the P-504 which runs for about 2,000 km from Manday to Magadan.
CHALLENGE 1 — Longest Land Leg
Thanks to James Robertson for finding this exhausting 2,121 km single leg trip. It's a route along Australia's National Highway between Kathernine Hot Springs and (2,121 km later) a fork in the road that splits into the NW Coastal Highway.
I'm sad to see that the route no longer includes Magadan, known as the sorrow at the world's end, which is such an important destination for the Google Map Challenge.
On the upside, this route is so long that "Your destination is in a different time zone". Depending on which way you drive, this might save you an hour.
What happens when you get to the end of the route? You get a much deserved rest stop. Hilarious Aussie humor at its best.
challenge 1 version history
These routes may no longer be available in Google Maps or their routing may have changed.
v6 route Sakha Rebuplitis - Magadan 1,601 km (+44 km)
v5 route Sakha Rebuplitis - Magadan 1,601 km (+44 km)
v4 route Haines Junction - Dawson Creek 1,557 km (+25 km)
v3 route Haines Junction - Farmington 1,532 km (+32 km) (submitted by David Jackson)
v2 route Haines Junction - South Taylor 1,500 km (+25 km)
v1 route Haines Junction - Charlie Lake 1,475 km
CHALLENGE 2 — Longest Land Route
The longest Google Maps route that does not use ferries takes us from Sagres in Portugal to Khasan in Russia. Thanks to Pieter Vandromme for submitting this route.
(13 Jun 2017) I just noticed that there's a 9.4 km ferry in this route now, which crosses the Aldan River. Technically, this disqualifies the route. But instead, I'll change the rules so that the ferry crosses a river and is no more than 10 km :)
challenge 2 version history
v7 route Sagres in Portugal to Khasan, Russia 14,069 km. This route is now significantly shorter (13,322 km).
v6 route Sagres in Portugal to Tanjung Pengelih Pengerang Johor Malaysia 16,280 km. Thanks to Jørgen Adam Holen for pointing out a better destination in Malaysia. (Google Map route no longer available)
v5 route Duyker Eiland, South Africa - Sidi Bettache, Morocco 15,594 km (+18 km) (Duyker Eiland submitted by David Jackson on xkcd) (Google map route, 12,523 km on 26 Nov 2014, significantly shorter with routing through Congo)
v4 route Pearly Beach, South Africa - Sidi Bettache, Morocco 18,260 km (+84 km) (Google map route 12,673 km on 26 Nov 2014)
v3 route Pearly Beach, South Africa - Casablanca, Morocco, 18,176 km (+2,180 km) (submitted by ElWanderer via xkcd) (Google map route 12,708 km on 24 Nov 2014)
v2 route Gibraltar - Paren', Russia 15,996 km (+602 km) (submitted by ElWanderer via xkcd) (Google map route no longer exists on 6 Jun 2013)
v1 route Gibraltar - Magadan, Russia 15,394 km (Google map route now 15,051 km on 26 Nov 2014 and now includes a 9.4 km ferry (a/d Kolyma/M56))
CHALLENGE 3—Longest Land Route with Ferries
The longest Google Maps A–B route that uses ferries. The ferry distance cannot be more than 25% of the entire trip.
Sadly, it appears you can no longer drive from Quoin Point to Indonesia. So, Talon in the Magadanskaya oblast' is back as everyone's favourite gulag town destination—truly the last stop for sadness. To make matters worse, this route is now only 26,320 km, which is shorter than the original entry for this challenge. Sad!
challenge 3 version history
v12 route (7 Sep 2015). This route takes you from Quoin Point on Cape Agulhas to the edge of Indonesia, where there are things to do.
For some reason this excellent route was dropped by Google Maps. Its creator, Christopher Camitta, points out that if you assist the routing algorithm by adding a mid-point in Pakistan, the route can be recovered.
v11 route (7 Sep 2013) Quoin Point, South Africa to Talon, Russia 27,224 km (Google map route). This route takes you from Quoin Point on Cape Agulhas to the edge of nowhere in Russia. During your voyage, you experience the endless visual monotony of sprawling Africa (keeping a wide bearth from Angola), only to end up at the sorrow at the world's end—Magadan, Russia's gulag region.
v10 route (18 Jul 2013) Quoin Point, South Africa to Ola, Russia 24,201 km (Google map route. Thanks to Pieter Vandromme for submitting this route.
v9 route (18 Jul 2013) Quoin Point, South Africa to Unknown Road, Indonesia 33,540 km (Google map route no longer available, Map geek Earl Higgins found this route after invalidating v7. The 33,000 km barrier is again, and legitimately, broken.)
v8 route (16 Jul 2013) Unknown Road, Indonesia to Groot Paternoster Reserve, South Africa 32,433 km (Google map route no longer available, Map geek Earl Higgins pointed out that the reverse of the v7 route is significantly shorter. This means that the 33,000 km barrier has, in fact, not been broken. Pity.)
v7 route (6 Jun 2013) Groot Paternoster Reserve, South Africa to Unknown Road, Indonesia 33,634 km (Google map route no longer available, submitted by Sue DoNym on xkcd). (33,557 on 16 Jul 2013; reverse of route is significantly shorter—see v8)
v6 route Chimchememel' Russia - Duyker Eiland, South Africa 31,766 km (+626 km) (Google map route no longer exists, submitted by David Jackson on xkcd)
v5 route Chimchememel' Russia - Pearly Beach, South Africa 31,140 km (+15 km) (Google map route no longer exists)
v4 route Chimchememel' Russia - Danger Point, South Africa 31,125 km (+692 km) (Google map route no longer exists, submitted by nerd-7hi+42e via xkcd)
v3 route Pearly Beach - Paren', Russia 30,433 km (+602 km) (Google map route no longer exists, submitted by ElWanderer via xkcd)
v2 route Pearly Beach - Magadan 29,831 km (+66 km) (Google map route, 24,191 km on 26 Nov 2014)
v1 route Bregasdorp - Magadan 29,765 km (Google map route, 24,125 km on 26 Nov 2014)
CHALLENGE 4—Longest 3-Point Land Route
Thanks to Nico Tsaousis for suggesting the format of this challenge. The 3-point route must be constructed by typing in the waypoints, not dragging them.
Nico came up with the first entry (see v1 route below), a 33,156 km route from an Unnamed Road in South Africa to Magadan via Norway.
The current longest route in this category is a 48,504 km trip from the Mou Shkola P.kharp high school in Severnyy Kvartal in Russia via South Africa back to Talon in Russia. The route was submitted by Brad Bailey and requires the "avoid ferries" option.
v3 route Shchukozero in Russia via Izingolweni in South Africa to Talon in Russia (Google map route 46,176 km). This route requires that you select "avoid ferries" as well as "avoid tolls". What makes this route really unique is that it starts and ends in the same country.
v2 route Mpahlane in South Africa via Riksveg in Norway to Talon in Russia (31,377 km).
v1 route Unnamed road in South Africa to Talon in Russia (Google map route, 33,156 km on 14 Jun 2017). Because this route starts at an "unnamed road" rather than a specific address, I'm retiring it and replacing it by one that starts at Mpahlane in South Africa.
news + thoughts
Curse(s) of dimensionality
There is such a thing as too much of a good thing.We discuss the many ways in which analysis can be confounded when data has a large number of dimensions (variables). Collectively, these are called the "curses of dimensionality".
Some of these are unintuitive, such as the fact that the volume of the hypersphere increases and then shrinks beyond about 7 dimensions, while the volume of the hypercube always increases. This means that high-dimensional space is "mostly corners" and the distance between points increases greatly with dimension. This has consequences on correlation and classification.
Statistics vs Machine Learning
We conclude our series on Machine Learning with a comparison of two approaches: classical statistical inference and machine learning. The boundary between them is subject to debate, but important generalizations can be made.Inference creates a mathematical model of the datageneration process to formalize understanding or test a hypothesis about how the system behaves. Prediction aims at forecasting unobserved outcomes or future behavior. Typically we want to do both and know how biological processes work and what will happen next. Inference and ML are complementary in pointing us to biologically meaningful conclusions.
Statistics asks us to choose a model that incorporates our knowledge of the system, and ML requires us to choose a predictive algorithm by relying on its empirical capabilities. Justification for an inference model typically rests on whether we feel it adequately captures the essence of the system. The choice of pattern-learning algorithms often depends on measures of past performance in similar scenarios.
Bzdok, D., Krzywinski, M. & Altman, N. (2018) Points of Significance: Statistics vs machine learning. Nature Methods 15:233–234.
Background reading
Bzdok, D., Krzywinski, M. & Altman, N. (2017) Points of Significance: Machine learning: a primer. Nature Methods 14:1119–1120.
Bzdok, D., Krzywinski, M. & Altman, N. (2017) Points of Significance: Machine learning: supervised methods. Nature Methods 15:5–6.
Happy 2018 `\pi` Day—Boonies, burbs and boutiques of `\pi`
Celebrate `\pi` Day (March 14th) and go to brand new places. Together with Jake Lever, this year we shrink the world and play with road maps.
Streets are seamlessly streets from across the world. Finally, a halva shop on the same block!
Intriguing and personal patterns of urban development for each city appear in the Boonies, Burbs and Boutiques series.
No color—just lines. Lines from Marrakesh, Prague, Istanbul, Nice and other destinations for the mind and the heart.
The art is featured in the Pi City on the Scientific American SA Visual blog.
Check out art from previous years: 2013 `\pi` Day and 2014 `\pi` Day, 2015 `\pi` Day, 2016 `\pi` Day and 2017 `\pi` Day.
Machine learning: supervised methods (SVM & kNN)
Supervised learning algorithms extract general principles from observed examples guided by a specific prediction objective.We examine two very common supervised machine learning methods: linear support vector machines (SVM) and k-nearest neighbors (kNN).
SVM is often less computationally demanding than kNN and is easier to interpret, but it can identify only a limited set of patterns. On the other hand, kNN can find very complex patterns, but its output is more challenging to interpret.
We illustrate SVM using a data set in which points fall into two categories, which are separated in SVM by a straight line "margin". SVM can be tuned using a parameter that influences the width and location of the margin, permitting points to fall within the margin or on the wrong side of the margin. We then show how kNN relaxes explicit boundary definitions, such as the straight line in SVM, and how kNN too can be tuned to create more robust classification.
Bzdok, D., Krzywinski, M. & Altman, N. (2018) Points of Significance: Machine learning: a primer. Nature Methods 15:5–6.
Background reading
Bzdok, D., Krzywinski, M. & Altman, N. (2017) Points of Significance: Machine learning: a primer. Nature Methods 14:1119–1120.
Human Versus Machine
Balancing subjective design with objective optimization.In a Nature graphics blog article, I present my process behind designing the stark black-and-white Nature 10 cover.
Nature 10, 18 December 2017
Machine learning: a primer
Machine learning extracts patterns from data without explicit instructions.In this primer, we focus on essential ML principles— a modeling strategy to let the data speak for themselves, to the extent possible.
The benefits of ML arise from its use of a large number of tuning parameters or weights, which control the algorithm’s complexity and are estimated from the data using numerical optimization. Often ML algorithms are motivated by heuristics such as models of interacting neurons or natural evolution—even if the underlying mechanism of the biological system being studied is substantially different. The utility of ML algorithms is typically assessed empirically by how well extracted patterns generalize to new observations.
We present a data scenario in which we fit to a model with 5 predictors using polynomials and show what to expect from ML when noise and sample size vary. We also demonstrate the consequences of excluding an important predictor or including a spurious one.
Bzdok, D., Krzywinski, M. & Altman, N. (2017) Points of Significance: Machine learning: a primer. Nature Methods 14:1119–1120.
Snowflake simulation
Symmetric, beautiful and unique.Just in time for the season, I've simulated a snow-pile of snowflakes based on the Gravner-Griffeath model.
The work is described as a wintertime tale in In Silico Flurries: Computing a world of snow and co-authored with Jake Lever in the Scientific American SA Blog.
Gravner, J. & Griffeath, D. (2007) Modeling Snow Crystal Growth II: A mesoscopic lattice map with plausible dynamics.
Genes that make us sick
Where disease hides in the genome.My illustration of the location of genes in the human genome that are implicated in disease appears in The Objects that Power the Global Economy, a book by Quartz.
Ensemble methods: Bagging and random forests
Many heads are better than one.We introduce two common ensemble methods: bagging and random forests. Both of these methods repeat a statistical analysis on a bootstrap sample to improve the accuracy of the predictor. Our column shows these methods as applied to Classification and Regression Trees.
For example, we can sample the space of values more finely when using bagging with regression trees because each sample has potentially different boundaries at which the tree splits.
Random forests generate a large number of trees by not only generating bootstrap samples but also randomly choosing which predictor variables are considered at each split in the tree.
Krzywinski, M. & Altman, N. (2017) Points of Significance: Ensemble methods: bagging and random forests. Nature Methods 14:933–934.
Background reading
Krzywinski, M. & Altman, N. (2017) Points of Significance: Classification and regression trees. Nature Methods 14:757–758.
Classification and regression trees
Decision trees are a powerful but simple prediction method.Decision trees classify data by splitting it along the predictor axes into partitions with homogeneous values of the dependent variable. Unlike logistic or linear regression, CART does not develop a prediction equation. Instead, data are predicted by a series of binary decisions based on the boundaries of the splits. Decision trees are very effective and the resulting rules are readily interpreted.
Trees can be built using different metrics that measure how well the splits divide up the data classes: Gini index, entropy or misclassification error.
When the predictor variable is quantitative and not categorical, regression trees are used. Here, the data are still split but now the predictor variable is estimated by the average within the split boundaries. Tree growth can be controlled using the complexity parameter, a measure of the relative improvement of each new split.
Individual trees can be very sensitive to minor changes in the data and even better prediction can be achieved by exploiting this variability. Using ensemble methods, we can grow multiple trees from the same data.
Krzywinski, M. & Altman, N. (2017) Points of Significance: Classification and regression trees. Nature Methods 14:757–758.
Background reading
Lever, J., Krzywinski, M. & Altman, N. (2016) Points of Significance: Logistic regression. Nature Methods 13:541-542.
Altman, N. & Krzywinski, M. (2015) Points of Significance: Multiple Linear Regression Nature Methods 12:1103-1104.
Lever, J., Krzywinski, M. & Altman, N. (2016) Points of Significance: Classifier evaluation. Nature Methods 13:603-604.
Lever, J., Krzywinski, M. & Altman, N. (2016) Points of Significance: Model Selection and Overfitting. Nature Methods 13:703-704.
Lever, J., Krzywinski, M. & Altman, N. (2016) Points of Significance: Regularization. Nature Methods 13:803-804.