But is Phil a good weatherman? In this blog entry, I’ll do my own statistical analysis on this paradoxical prognosticator. Full disclosure: my middle name is Phil, named after the famed Groundhog Ambassador from Punxsutawney (he is also my father. And before you ask, the groundhog was named Pete when my father was born – so technically the rodent was named Phil after my dad.)

Let me start by saying that I am not the first to analyze Phil’s predictions. The Internet is filled with dozens of differing analyses that boil down to “how often is Phil right?” Most of them fail to do a thorough, fair analysis with data and methodology posted. In fact, I have never seen a solid analysis of our beloved furry forecaster, so I will attempt to do that here.

So let’s get into it. For the geek and the academic transparency oriented, here is my data, my Excel sheet, and my R-script used in this analysis.

As I understand it, the tradition goes like this:

“If Phil sees his shadow on February 2^nd, then there will be 6 more weeks of winter. Otherwise, there will be an early spring.”

That’s pretty vague. Does Phil’s predictions apply to just his area, or to the whole country? What constitutes an early spring? Do we even know Phil’s track record?

First, the good news. We know Phil’s predictions all the way back to 1887 from here (with 10 missing years) – so we’ve got that data. As for “early spring” data, that’s much more difficult. If you want to spend about $1000, you can buy temperature data for Punxsutawney all the way back to the mid-1800s, but I don’t have that kind of budget.

So, I went with the freely-available average temperature and precipitation data for the entire state of Pennsylvania going back to 1898. With that data, I am making the following assumptions:

• Locality to the state of Pennsylvania: Despite Phil’s international fame, I will assume that Phil’s predictions only apply to the weather of the state of Pennsylvania.
• Temperature and Precipitation. I assume that early spring is a function of the temperature and precipitation.
• Back to 1898. Sorry Phil, but I don’t have weather data going back to 1887… analyzing back to 1898 will just have to do.

According to history, Phil predicts that we will have an early spring 11.8% (n=13) of the time. That’s not very often, although considering the climate of Western Pennsylvania I wouldn’t be so optimistic either.

Let’s start with a naïve approach. Let’s label certain years as having an “early spring”? Sure, it’s a subjective measure that varies from person to person, so let’s go with my personal measure of wintriness.

“Andy’s Measure of Pennsylvania Wintriness” is defined as follows:

If the average temperature for the state of Pennsylvania is above freezing (32° F) for the month of February, then Pennsylvania has had an early spring.

According to my metric applied to the historical data, 16.4% of Pennsylvania’s last 110 winters had an early spring. That’s pretty close to Phil’s rate of 11.8% – so maybe there is something to this tradition.

Now that we have labeled each year, we can see that Phil can make two different kinds of mistakes: he can predict an early spring and be wrong, or he can predict a long winter and be wrong. In statistics, these are called false positives and false negatives, respectively. Understanding which kinds of mistakes Phil makes more often is important to interpreting his prediction year after year.

For example, if Phil always saw his shadow, he would still be right 83.6% of the time, but would never actually predict an early spring. That’s why overall accuracy is a poor validation measure in this situation, that is, when positive experiences are as rare as 16.4% of the time. Instead, let’s borrow two measures from medical diagnostics – sensitivity and specificity. Below are the results of Phil’s hits and misses (along with the interpretations).

Interestingly, the last time Phil predicted an early spring and we actually had an early spring was 1999, and the other two years were 1997, and 1990. Before that, Phil never predicted an early spring correctly, yet was still right most of the time.

So you might think that having an accuracy of 77.3% is pretty good (after all, I believe that is a passing grade for Punxsutawney High School). Except that we can see that Phil got our hopes up many plenty of times (FP=10) and missed the early spring even more (FP=15). Having a very low sensitivity (16.7%) also tells us that Phil is generally a pessimist in his predictions.

Now you may disagree with Andy’s Measure of Pennsylvania Wintriness, so let’s approach this more objectively. The two best metrics we have for post-prognostication wintriness are temperature and precipitation in the month of February and March in Pennsylvania. If Phil is good at predicting early springs, then the years that he has predicted an early spring should, on average, have a higher temperature in both February and March. Well, here are those means:

So, according to those results, in both cases of February and March, the mean temperature of Pennsylvania during those next six weeks is actually slightly lower when Phil predicts an early spring. But, more importantly, the difference between the two mean temperatures is negligible (more on that in a moment). Precipitation also shows similar results:

So if temperature and precipitation are usually more wintry when Phil misses his shadow, should we reverse the tradition?

In short, no. The difference is just too small. To test the difference between these two means, I used the Mann-Whitney-Wilcoxon (MWW) test.  What the MWW test does is rank all of the temperatures, and measures if the “Early Spring” temperatures rank higher on average than the “Long Winter” temperatures. In my tests, none of the measurements were statistically significant at a p<0.05 level. Not even close, actually.

In other words, Phil’s predictions are, historically, no better than a random guess. (More technically, after a century of data, we don’t have enough evidence to believe Phil’s predictions are any better than a random guess – perhaps with another century of data, a statistically significant difference may eventually emerge. Maybe.)

That’s a pretty conclusive result, but I have to say that this is still fascinating. According to my wintriness metric, Pennsylvania has an early spring about once in 7 years (14.6% of the time). Phil also predicts an early spring about once in 9 years (11.8% of the time). Those are similar probabilities, there’s just no historical evidence of a correlation between the years that Phil predicts an early spring and actually having an early spring. If Phil’s predictions were that of a coin flip (i.e. 50% early springs), then the tradition probably would have died because a coin flip is way too optimistic. Several interesting years support my theory (e.g. 1983, 1975, and 1950) where Phil predicted an early spring, but then the early spring happened the next year, or one year before the prediction.

My point is this: the fact that both early springs and groundhogs seeing their shadow happen about equally often (and rarely at that), and that nobody can remember which year was which, is probably why this myth has stuck around for over a century.

Or maybe it’s because groundhog tastes really good in February.

Happy Groundhog Day!

PS. – As a side note, let’s all remember that Phil really doesn’t have much evidence to act on for his prediction. Even the mean temperature of January is not a statistically significant predictor of the mean temperature of February, so predicting the ends of winters in Western Pennsylvania is actually much more difficult than it sounds.

There are about a million eggnog recipes out there, but mine is probably the only one that is customized for lattes. Here it is:

Andy’s Latte ‘Nog

In my opinion, the key to a good eggnog latte is to keep it light but not thin. If you’re buying eggnog for making a latte, the light ‘nog will steam better and work with the espresso. Most eggnog recipes you’ll find are very heavy (and calorific!), and don’t work well with espresso. This one is a lot lighter while still using real eggs.

4 eggs, separated
2 tbsp flour
1/2 cup sugar
2 tbsp of sweetened condensed milk
4 whole cloves
Pinch of allspice
Pinch of cinnamon
2 cups skim milk
1 cup of fat-free half & half
Pinch nutmeg
1 tsp vanilla

Makes 1 quart, or 6 lattes.

1. Combine the milk, sweetened condensed milk, flour, cinnamon, cloves, allspice in a thick-bottomed saucepan. Using low/medium heat, heat until steaming but not boiling. Temperature should get to about 120-140 degrees. (If you don’t have a thermometer, until it’s steaming)

2. While the milk is heating, separate the eggs and beat the egg yolks in a large bowl until they get lighter in color. This can take a while by hand, so I use a hand mixer with the whisk attachment. Once it’s fluffy and thick, whisk in the sugar a little bit at a time.

3. Also, whisk the egg whites in a second mixing bowl until thick. Add a teaspoon of powdered sugar, whisk again until the mixture is even thicker.

4. Pour most of the hot milk into the egg yolks in the bowl a little at a time (so the eggs don’t curdle). Keep whisking. Pour the whole mixture back into the saucepan. Pour in the egg whites.

5. Turn up the heat to medium-high. Stir constantly, bringing the temperature up to 160 degrees. Now be careful with this because the temperature can accelerate quickly. Eggs are safe when cooked to 165 degrees for a few seconds, but they will congeal very quickly beyond that – especially if you’re not stirring. I usually turn off the heat when it gets to 150, then remove the saucepan from the burner once it’s been over 160 for a few seconds.

If you’re not using a thermometer, you will notice that the mixture will thicken enough to coat the back of a wooden spoon.

6. Pour mixture into a pitcher. Remove the cloves (via strainer or just a fork). If there are some congealed chunks, use a blender (I like a hand-blender). Stir in the cream.

7. Cool in the fridge for an hour.

8. Once cooled, mix in vanilla and nutmeg. Chill for at least 3 hours (overnight it gets even better). Stir or shake before serving.

To make the latte, make sure the eggnog is very cold. Use about 6oz of eggnog for each shot of espresso. Eggnog cools down much quicker than milk, so I recommend only making one 6oz drink per person. Only steam the espresso to 140 degrees (instead of the usual 160 for milk). Again, eggnog’s temperature will accelerate quickly, so keep an eye on it. Add one shot of espresso. Top with a tiny bit of cinnamon for aroma if you like.

Nutrition facts per 6oz serving: 196 calories, 4g of fat (6% DV), 1g saturated fat (5% DV).

Myth: Open source is just an academic pipe dream.

This was a more prevalent argument 10 years ago. Today, all I have to do is point out that open source software is everywhere. Mozilla Firefox is one of world’s most popular web browsers. A huge proportion of the Internet’s websites are run by products like Linux, Apache, MySQL, PHP, Perl, Tomcat, WordPress, and Java. Google’s Android platform is becoming one of the biggest platforms for mobile devices today. Software developers themselves use a plethora of open source software tools and libraries, the most famous is probably Eclipse. Open source is so ubiquitous, you’d be hard-pressed to argue it’s not relevant.

Myth: All software should be open source!

I have to disagree there, too. I think we will always have proprietary software, and that’s a good thing. What we have seen, though, is that the open source business model has proven itself in just about every domain of software – which many people did not expect.

Myth: Open source is just crappy versions of “real software”.

If this is your opinion, then maybe you’re looking in the wrong places, or not at the whole picture. To me, the negative examples don’t say nearly as much as the positive examples.

Yes, there’s a ton of horrible open source software out there. Dive into Sourceforge.net or any other open source portal and you’ll be swamped with dead projects. Why? Because making software open source is trivially easy (not a bad thing!) and most people in the world are bad programmers.

The fact of the matter is that good software comes from good developers. Period. If you have a good development team that works well together, you might have a good product – regardless of whether the whole world can see the source code or not.

To disparage open source software because you like Word more than OpenOffice isn’t really fair to the whole concept of open source. I think we all know there is no perfect open source alternative for everything, but there’s a whole lot of good, free, and free alternatives out there (…free as in both speech and beer).

So it’s like a charity, right?

Um, no. Companies that market open source software actually make a lot of money. Sure, their product is not necessarily a shrink-wrapped box of software you buy at Staples – but that’s not how most people get their software anyway. In this day and age, if people hear about a good piece of software, they just go download it and use it.

Here’s the key: software is more than just source code, it’s a complex maze of knowledge. This knowledge is often accessible to only a few, highly-trained professionals – and that’s what can be sold. The Linux kernel is a great example of this: the source code has been open for a long time, but what companies like Red Hat do is sell you the service of compiling, installing, configuring, patching, fixing, and any other kinds of support you need for a big fee. Sure, you could go learn everything about Linux for free, but do you really have that kind of time?

Other companies take ad-supported approaches. Mozilla Firefox mostly makes their money from selling advertisements in their product (such as having the Google search bar in the upper right corner – yep, Google pays millions for that little search box).

There are tons more myths out there on open source – so I think we’ll have a “Part 2″ on this in the future. Until then, what are your opinions and experiences on open source software?

The Semantic Web is an idea set to bring a higher level of semantics to the internet. The concept is all about structuring data to be both human-readable and machine-readable.

For example, suppose I have a website for selling books, and I have the words “War and Peace” and “$10″ sitting right next to each other. For humans, we can clearly see that “War and Peace” refers to a book and that $10 is the price. A computer however, would not be able to figure that out so easily. If there were multiple book titles and multiple prices, how could a computer determine which book goes to which price?

In a world where Semantic Web reigns, you would need to annotate that “War and Peace” text as the title of a book, and provide some kind of reference to the $10 as the price of that book. That annotation and linkage is what we call metadata (data about data), and it’s the backbone of the Semantic Web. With the right metadata, software can automatically traverse our data the way our brains do, making tasks like searching and aggregating much easier.

Many proponents of the Semantic Web believe that we should be structuring our content in a way that’s not tied to being human-readable (e.g. font size, colors, formatting), but being more machine-readable with an emphasis on interrelations and semantics. Then we should translate our machine-readable data into something visually pleasing. (Note that this a slightly different idea from separating document structure markup from formatting as in HTML/CSS and LaTeX – there would need to be a lot more non-structure annotation involved in the Semantic Web worldview.)

Turns out – about once every thousand miles.

The fact is, though, our brains are wired for pattern recognition. What we don’t realize is that the number of patterns we would recognize are much larger than we are aware of, significantly raising the chances we run into a pattern. This is the foundation of numerology – the study of the significance of patterns.

Years ago, I came across an interesting numerological example. A group of three couples were doing a “white elephant” Christmas party. What happens is that everyone comes to the party with one gift. Everyone then draws a number, and people open a gift in turn. (There’s also stealing involved – but that’s outside the scope of this example.)

So our six people all drew numbers. Immediately, we figured out that if you paired off the spouses, the pattern was one couple at numbers 1&6, another couple had 2&5, and another couple had 3&4.

Spooky, huh? What are the odds??

Actually, a pattern like that is quite probable. First off, there are 720 ways the numbers could have been drawn – that is, 6! (factorial).

Next, if a couple had reversed their pairings, we would have noticed the pattern. For example, if Adam&Eve had 1&6 or 6&1, we would have noticed the pattern, so we can flip each couple two ways. There are 2^3 = 8 ways to do that.

Now, we would have noticed the pattern regardless of which couple was assigned to which number pair. So if you had Adam&Eve got 1&6 and Samson&Delilah got 2&5, or vice versa. There are 3!=6 ways to do that.

So that now puts us at 8*6 = 48 ways to get the pattern we got.

However, I would also argue that we would have noticed if the couples were assigned consecutive numbers, eg. 1-2,3-4,5-6. These cases behave as the other pattern does, but don’t overlap any of the previous counts, so that doubles our count to 96.

Another pattern we might notice would be a difference of three: 1-4, 2-5, and 3-6. Again, 48 ways to do that pattern. So now our count is now 144.

So we would notice this pattern 144 times out of 720 – that’s 13.3% of the time.

Still sound small? Now consider how many numbers you encounter every day. Phone numbers. Account numbers. Highway numbers. Street addresses. The time. I still sometimes marvel at how often I notice when the time matches my birthday, for example. Throw in a little confirmation bias, and the chances that you will find a pattern in there becomes very likely very quickly.

Finding patterns is fun, though. Ever notice interesting patterns in your everyday life?

What is the central limit theorem? In a nutshell, “everything eventually is normally distributed”. What does that mean? Well, you’ll just have to read on.

The general finding of the central limit theorem is: if we sample something a lot, we know how the general population behaves. More specifically, if we take repeated observations of something, and each of those observations are identically distributed and independent of each other, then those observations are normally distributed.

Now the normal distribution is what we might recognize as “bell curve”. As an example, take peoples’ heights. The average height of an American male age 20-39 is 5″4.5′  (according to this page). Now, we know that there are people who are above that height, and below that height. If heights are normally distributed (and they generally are), then the farther you get away from that mean, the less likely you are to find a human of that height. For example, a 7-foot guy is very uncommon, but a 5’6″ guy is much more common. Since we know that the distribution is normal, we can use math get a good estimate of how likely a 5’6″ male is versus a 7-foot male.

The field of statistics is founded on this concept of taking random samples and seeing how they are distributed. In some simple cases, statistics assumes normal distributions of the population – but not always. In most situations, statisticians will assume a normal distribution of their errors, but not the population itself. That is, if we try to explain many observations, and for each observation we make there’s a small chance of an error on our part, then according to the central limit theorem, those errors are normally distributed regardless of how the actual population is distributed.

Now, such a foundational concept like the normal distribution ought to have an elegantly simple formula, right? Not in this case. Below is the probability distribution function of the normal distribution:

Where mu (μ) is the mean and sigma (σ) is the standard deviation – which have their own formulas.

Wha?? If you’re keeping score at home, this formula includes all kinds of weird elements to it. First of all, it’s got every math operation we know and love: multiplication, subtraction, division, square root, exponentiation. But there’s more… it’s got two famous transcendental constants:  pi and e. Do you know of any formula with both pi and e?? The derivation of this formula even includes integration of both Cartesian and polar coordinates using trigonometry. Oh, and that particular formula has no anti-derivative, so the integration cannot be done algebraically… so we have to use computers to actually use the darn thing.

Isn’t it fitting that the “normal” distribution involves this big confluence of mathematical concepts? So the central limit theorem is not just central because it’s foundational to all statistics, but because connected to all other kinds of math, which is really the mathematical definition of centrality. That’s like a double-entendre wrapped in a square root wrapped in a function with no anti-derivative.

Woah.

Every software development team has a process, even if it’s not explicitly articulated anywhere. The process tells you what you should do, how much you should do it, when you should do it, and why. Sound boring and bureaucratic? It can be, but it doesn’t have to be.

In the late 1990′s, a bunch of software engineers decided to shake things up. They were tired of the current set of software processes out there, and decided to come up with a new philosophy called agile. The foundation for these new software processes culminated in the Agile Manifesto. (It’s 68 words – feel free to go read it.)

The 17 founders of the agile manifesto were reacting to the plan-driven nature of many software processes going on at the time. To many practitioners, the plan-driven philosophy emphasized a get-it-right-the-first-time attitude, or a plan-everything-out-ahead-of-time approach.

However, as I’m sure everyone has noticed by now, technology changes quickly. So do software markets. And if you’re making software that will take 18 months to finish, the technologies and markets and  might drastically change by the time you’re done.

In agile processes, the goal is to be able to change the scope of a project in the middle of the project. Have a plan, but plan for change. Instead of having a product with a fixed scope and a variable release date, have a fixed release date with variable scope. Instead of making a vague plan for a whole year, make an explicit plan for one month, then make a new plan next month. Communicate with market experts as much as possible (e.g. have developers talk to the customers themselves). Have a stable product ready as often as possible, continuously integrating everyone’s work together and automatically testing it out every night.

Agile is certainly not without its controversies. Over the last decade, consultants have sold the advantages of agile pretty hard and many software engineers (and researchers) have retaliated with incendiary skepticism. As with all religious wars, words have been re-defined, various myths have propagated, and poor characterizations have arisen. For example, some people claim that agile processes are just a do-whatever-you-want process. Agile consultants are often portrayed as disingenuous hucksters selling the latest fad diet to people. On the other side, agile proponents often act as if all software teams must use agile even if their current process is working perfectly. Lack of sound research hasn’t helped, either. While agile does have some strong evidence, agile proponents mostly cite personal experience, leading to  unsupported claims on the effectiveness of their practices.

In recent years, though, the wars have quieted. Many software companies use some agile practices, selectively leaving out other practices. Generally speaking, agile has been widely adopted over the last ten years and the Agile Manifesto has remained accepted among agile proponents as foundational doctrine.

While the agile philosophy works incredibly well for many situations, it doesn’t work for everything. From my perspective, it looks like both agile and plan-driven philosophies are here to stay.

Yes, that's a picture of me.

If you are not familiar with the sport, you might look at that title and go “ultimate what?” No, it’s not an adjective, it’s a noun. The sport known as Ultimate (aka Ultimate Frisbee) is a truly fantastic game, and is something I am very excited about.

In Ultimate, two teams of seven players occupy a field larger than a football field. There’s one disc (aka “frisbee”), and players cannot walk with the disc. Instead, players must throw the disc to each other across the field without dropping or getting intercepted. Catch the disc in your endzone, your teams gets a point.

Ok, enough of the rules – why am I excited about Ultimate?

1. It kicks your butt.

I was once at an Ultimate tournament where one of my teammates was the sister of an NFL safety. This NFL player came to play with us one day, and he commented that this game “…is relentless – it really kicks your [butt]“. Ultimate demands tons of agility, endurance, and coordination. In Ultimate, you’ll find yourself sprinting eighty yards to go out-jump other players to grab the disc, then you’ll need the finesse and clarity of mind to toss that disc to another teammate.

2. Not much to it

Learning Ultimate takes no time at all. All you need is a field, a disc, and some people. That’s it! The more you play, you’ll probably want to get some good cleats (wide receiver cleats are great, so are soccer cleats).

3. Can be tons of strategy

Despite what you’ve heard, playing Ultimate takes a lot of tactics and strategies. All of these offensive and defensive techniques have their own vocabulary: huck, force, cup, zone, man-to-man, poach, last back, stack, ho-stack, break, swing, dump, weave, etc. We’re not a bunch of stoners out there goofing off.

4. No refs with Spirit of the Game

To me, one of the most distinctive characteristics of Ultimate is the lack of referees. Teams compete at the national and international levels all the time, and yet the players all call and contest their own fouls. The rules of the game are designed so that no refs are needed, and I suspect that referees will never be added to the sport.

5. The community

This is the best part. I have played Ultimate in Philly, Raleigh, St. Louis, and various places in Western Michigan – every one of those communities are filled with great people. It’s amazing to me how the Ultimate community has people who are competitive, yet friendly. This reason deserves a whole blog post by itself, so I’ll leave it at that

So stop reading silly blogs and go play some Ultimate.

Seem strange? Actually, designing board games is not much different than other design activities. In this article, I’ll talk about how embarking on the journey of design is a wonderful way to spend your time.

“Every block of stone has a statue inside it and it is the task of the sculptor to discover it.”
-Michelangelo

Design is a fundamental process that we all experience. Writing good novels is a design process. So is graphic design. Even developing software or writing math proofs are big design processes. For me, I like to design board games.

You’d think that the design process for each of those domains is different. In fact, board game design is not much different than design in other domains. One of the best nuggets of design wisdom that I have ever encountered was from a hobbyist board game designer podcast, where the host said something like this:

• Work at it until it’s good;
• Then work at it until it’s great;
• Then work at until it’s obvious

To me, this summarizes the entire design journey. There’s a moment where you get that game to a point where it just makes sense. It’s more of a “discovery” moment than a “refining” moment. That’s why I love that Michelangelo quote – you’ve discovered the sculpture inside the block.

Does that mean you’re done? Not by a long shot. According to poet Paul Valery, “A poem is never finished, only abandoned”. I think that’s true about all design.

Here are some other pieces of wisdom I’ve learned from my journey in board game design:

Be willing to abandon ideas (just archive them)

Sometimes it’s hard to part with a good idea. Often, I find that my ideas for a single game end up being two games, for example. Afraid of losing a good idea? What I do is keep all of my old ideas around and periodically go back to them. You’d be surprised how looking at old ideas with fresh eyes can help you get a better perspective of what you haven’t explored.

Finding Good Ideas

Good ideas don’t usually come fully-formed. It’s usually a series of hard work of trying new ideas left and right. Sure, you’ll get a flash of inspiration every now and then that leads to a great idea  – but those are moments you have little control over as a designer. What you do have control over is how much you record those ideas – good and bad. It takes discipline, but I find that skilled designers will work through 103 good ideas to get that 104th amazing idea.

Let your ideas ripen.

I learned this from my 11th grade English teacher with writing, and it applies just about everywhere. After a while, separating yourself from your own work is the best thing you can do for the work. Putting down a design for a few weeks (or even a few days) will give you a fresh perspective the next time you pick it up. Trust me, it’s amazing how well that works.

Anyway, those are some nuggets of wisdom I’ve taken from my journeys in design.

Anyone have their own take? What’s your favorite thing you like to design?

Not convinced? Then you need this list.

1. Lattes are simple!

A latte is essentially steamed milk with espresso. That’s it. You’d think that would be a boring drink, but if you froth the milk just right, it actually comes out to be a smooth, very textured drink.

2. Lattes are healthy!

A latte has the same nutritional value as a short glass of milk (around 90 calories for skim milk). If you’re worried about caffeine, a regular latte has less caffeine that a can of diet coke, and far less than a typical cup of coffee. Personally, when I drink lattes, I never go into caffeine withdrawal. Only when I go back to drinking drip coffee do I get headaches.

3. Latte art!

Just YouTube latte art and you’ll be amazed at what people can do with just pouring perfectly frothed milk.

4. Crazy flavors!

I’m not normally one for doing different flavors in my espresso, but the latte is a great base for all kinds of crazy drinks. Probably the most popular is the vanilla latte. My personal favorite latte is the cinnamon & cayenne pepper latte. Yeah you read that right.

5. Great for coffee drinkers

If you like the taste of coffee, then the latte is a great drink for you. Being such a simple drink, you can really taste the espresso without being hit over the head with flavor. (If you do want to be smacked with flavor, try a cappuccino or just straight espresso.)

So go to your local coffee shop and go order a latte. Tell them I sent you… you’ll get a funny look because they probably don’t know me.