Ice pellets?

So, I have a little project going.

For a while now, I've been wondering just how accurate those seven- and fourteen-day out weather forecasts are.  So I've started a spreadsheet.

This is how I solve things...with data.  'Cause without data, all you have is conjecture and anecdotes.

I've shown it to a couple of people, and they seemed a little confused about how the spreadsheet works, so let me explain.

Short version: I'm tracking the predicted high temps for my zip code using four different websites and seeing how much those predictions change over the course of the week or two that they're predicting outward.

Longer version:  Going across the top (row 1) is a list of four websites.  Row 2 then are the specific links to those websites' long-term forecasts (as long-term as they happen to offer) for my zip code (45069).  Row 3 then is the website's predicted high temperature for 45069 as of how ever many days out from that specific date - as well as the actual and average historic high temperature according to official Cincinnati records.

Now, to the actual data.  Take a look at December 1, for example (row 31).  The actual high temperature for December 1 was 33 degrees (Fahrenheit, of course).  As of November 30 (the day before December 1, marked in the High 1 day out column), weather.com predicted that the high temperature was going to be thirty-two degrees.  The day before, they predicted that December 1 would be 36 degrees.  As you slide to the right in row 31, you see the predicted high temperature for December 1 as we go backward in time until the first day that December 1 showed up on weather.com on November 22 and was supposed to have a high of 43 degrees.

The next set of columns does the same thing for AccuWeather.com, then WeatherBug.com, then WeatherUnderground.com, and finally the average high temperature for 45069 based on the past however long data has been kept.

(Longer version all done now.  You can come back.)

I don't necessarily know what I'm going to get out of the data just yet.  I've promised myself that I'm going to do this for 120 days before I come to any specific conclusions.  I have noticed that AccuWeather's fourteen-day out forecast shows a lot more variation in the 14-days out time than it does when it's only a couple of days out.  I highlighted a couple of dramatic changes that AccuWeather showed.  December 7 was supposed to be 35 degrees as of 13 days out but moved to 59 degrees as of 12 days out - a change of 24 degrees.  The same kind of thing happened for December 2 when 14 days out, it was supposed to be 60 degrees, but the next day the high was only supposed to be 37 degrees - a 23 degree swing in a day.  The most dramatic change I've seen from the other sites is six degrees from Weather.com

No big conclusions yet or anything...just throwing it out there that I'm working on this because...um...because I...

Correct...

Oh, and today we're supposed to have ice pellets.  WeatherUnderground says so.

At least that's what WeatherUnderground said on Sunday morning when I typed up this post.

PS - I decided to chart only the high temperatures for the various forecasting sites because (a), it's an easy thing to compare quantitatively as opposed to precipitation which comes in weird % chances and everything and (#2) it's really easy to get and type up at a glance.

Distorting the world to reveal the truth

I stumbled across a CNN.com article about a book and website Atlas of the Real World in which computer mapping tools have been used to redraw the world map according to various statistics other than land area.

Up above, you'll see the map of AIDS deaths around the world as of 2002. There's also detailed information about this map available, as there is for every one of their maps found on their website where they offer their six hundred or so maps and information entirely for free.

Always good to see people finding new ways to present information to us.

Three things - not even remotely related

This graph is gorgeous. It presents huge amounts of data - GDP per capita by country, survey results within each country, and average satisfaction on a 10-pt scale by country - but it presents the data in an amazingly clear fashion, allowing anyone to quickly see that there is a clear pattern in the satisfaction vs income data of nations as we shift the average GDP per capita of those nations.

The original article containing the graph can be found here, and it's a moderately interesting article. The graph, however, is way better than the article.

---

I have now, officially had the best brownies in the world. The recipe from king Arthur Flour is surprisingly simple, and it makes a brilliantly fudgy brownie with a nice, slightly-crunchy top. Outstanding stuff, though I do warn you that I had the variety without nuts, 'cause that's how I likes my brownies. I cannot vouch for the nutty version.

Chocolate chips will provide tiny molten pockets of chocolate within the greater brownie landscape. Add them if your desire for fudginess knows no bounds.

And they are correct...yum...

---

Apparently we got a trembler here in the Midwest this morning.

I was up and checking the interwebs at 5:30 but sadly didn't feel a thing.

You can check out the major media's craptacular reporting of the event or some significantly better reporting of the same. Your call.

I won't judge you either way.

It did remind me, however, that there is a heck of a fault line running through the heart of our country, something that Iven Browning did, at least, teach me.

Elegent presentation of musical things

Who doesn't love the combination of great music and data analysis?

This article is hot 'cause it's fly

I love math...especially when it's applied to things that it clearly should never be applied to

The first mathematical guitar genius

I am an absolute sucker for statistical analysis applied in surprising places. I'm guessing atha anybdoy familiar with my blog knows about my jones for baseball stats, but here's a new one for me.

Recently, Bill Simmons mentioned a New Yorker article about some mathematically-inclined who have taken it into their minds to make some sense of movie profits.

They analyzed hundreds of scripts, plotted various aspects, and attempted to find correlations between the script aspects and the final movie revenues.

Nuts, right?

Nope, turns out that they've had a good deal of success.

Take a gander if you've got a moment or two...

In similar things on the web, there are a couple of math movie databases and one simpler movie math site.

The power of visuals

I'm a sucker for a well presented visual idea.

I'm about to dive into reading a cool-looking book that probably suggests a bit about that feeling.

I've also mentioned my digging of various baseball stat sites, and I continue to be excited with new, better ways to present information in a way that makes the information instantly, easily understandable.

And along comes Idiagram - a site of a visual designer in the State of Mass. He's got some really cool ideas about explaining processes and ideas. Some of them are a little complex, and others are absolutely gorgeous.

It's certainly a site worth playing around on.

Thanks, by the by, to good ol' TD for sending the link my way.

Good stuff...

The rule of the stat boys

If you're a baseball fan who's hanging on the web, you've got to be checking some of the cool sabrmetric stuff on the web.

• The Ray Lankford wing of the hall of fame from the outstanding Beyond the Boxscore blog
• THT Dartboard showing just how dominant the AL is (eight of the top nine teams)
• Contending Holes showing where the weakest parts of the NL teams are - #2 is amazing, and the St. Louis Cardinals are being carried by Albert
• Ten Thing I Didn't Know Last Week is easily one of the most informative columns on The Hardball Times
• Calculated playoff odds from the partially subscription Baseball Prospectus

A better way to see things

We're visual people. One of the things that we do best is to look at data and find patterns that we didn't necessarily expect to see. And for a long time, people have known that it's often easier to see those patterns if the data is presented visually.

One of the areas that present tons of data that are kind of enjoyable to look at is baseball. Even in ye olden days, every batter had a dozen stats (HR, avg, slugging, SB, hits, runs, RBI, W, ERA, K), and thankfully, a bunch of websites have begun to find ways to present this data graphically instead of just running all the numbers in huge, long charts. (Not that a website with the stats for every baseball player in history is a bad thing, you understand.)

First up, there's the Hardball Times whose site presents graphs showing timelines of the division races. They're the source of the graphic I've presented above. It's a great way to look at, for example, the NL Central and see just when it was that the Reds hit a rough patch. My honest guess is that the blue line won't be getting closer to the brown line anytime soon.

There's also fangraphs.com that has really been at the forefront of finding effective visual ways to present baseball numbers. They're developed a graph called the win probability graph showing how each and every event - every hit, every strikeout - changed the chances that each team would win the game.

On the graph on the left, for example, you can see that in the Reds-Rockies game from July 16th that it was Rich Aurilia's single in the eighth that really made the difference, taking the game from a 70% for the Rockies to a 90% win for the Reds with one swing. Fangraphs also has some great graphs allowing you to look at individual players and check out how their tendencies (grounders vs fylballs, power vs speed, etc) have changed over their career. You could look at, for example, what's wrong with Mark Teixeira.

Finally there's BaseballGraphs.com that presents loads of data - including the 1990 season summarized. Here you can see that based on runs scored and runs allowed, the Reds were probably the second best team in the NL in 1990 (their last worls series title year). The Mets' stats predicted that they'd be over .600, but they ended up seven games below that predicted record.

And it looks like the Reds were doing it with pitching, rating high on the vertical scale (low ERA = higher) and left on the horizontal (high runs scored = far right).

Of course it doesn't matter because a flag flies forever.

All of this is relevant because I'm heading to Detroit today to check out a game in Comerica. We'll be in the first row over the wall in the very left-most corner - section 144.

Enjoy the day, folks...