Football Tunes

Football and truly great music have never been the best of bedfellows, but there's something about the collective and dissonant tunerizing of thousands of semi-drunk supporter-spectators which can bring a genuine, warm smile to your face.

I'm not talking about the funny/rude songs which different supporters sing at each other or the players, but more the traditional "theme music" type of choral festival.

Everton have never had the best terrace songs, but they do have one of the more whistle-able tunes in 'Johnny Todd' (AKA "That theme music from that there Z-Cars programme, like") and, but for the fact that it has a weird time-signature switch in the third line, it could have become a terrace classic. Except ... that I caught some of the Carling Cup highlights mid-week and what should I hear? Johnny Todd blaring out from the Vicarage Road P.A. system. It was nice of yon hornets to commemorate the day you were handed your rears by a clearly superior outfit in the F.A. Cup Final, but, please: Where do the lyrics mention the M1 motorway?

In truth, though, as mentioned, Everton generally, like God, don't have the best tunes.

West Ham have grabbed a good one. I was once at a game between West Ham and Charlton and the way they sang "Bubbles" (frequently) was almost enough to make the neutral wish they were cockneys. I said "almost", capisce? Sheffield United's ditty always gave me a chuckle. Who can hate a team who so openly celebrate beer, lard and tobacco (of two types) with such abandon? In passing, I guess I can't really veer through this subject without a nod to a reasonably decent ballad sung by the unemployed; something about kerb-crawling, I think it was: "You'll never walk alone"; that's it. It's not a bad choice, in fairness.

Anyway, this subject came back to me with some force when I heard that a friend, who works for a super-huge corporation, had been asked to learn the company ditty.

I'm sure you can agree, it's a powerful and severely sinister image.

Emacs report, week #1

It has issues but, I'm sorry to say, I'm beginning to like the not-so-small beggar. I actually prefer its cut(copy)-and-paste mechanism to vim's, and its buffer handling is nicer, too. On the flip, lisp makes it sometimes unresponsive and there's too much CTRL-related action, really, but it makes a change from banging on the ESC key like a chimpanzee; that ain't working ...

Of Mountains and Molehills

... talking of which, there's no 'mountain of the week' this week (I know: you're all gutted) as I've spent what little spare time I've had toying around with the PFC (or rather with a tangential problem area spawned by the PFC).

Check this out:

The bottom lines are the standard deviations from all algorithms. Their colours match the lines higher up to which they apply. The 'middle' strata of lines are measurements from the proportional solution, AKA 'The Garlic Algorithm' (which I unwittingly copied unseen in a later comment). The top lines are from a solution I've been playing about with. The numbers on the x-axis are inputs to the function which produces those outputs (they don't apply to the Garlic figures, which are there for comparison's sake). The numbers on the Y-axis represent either mean averages, or standard deviations. The "jaggier" lines come from smaller sample sets. With those deviations, you have to crank up the number of iterations to really see the curve emerge.

Here's the thing: I don't know why my function works that way. Why there is a "zone of wickedness" between inputs 3.0 and 4.0 (with broadly decreasing std-dev as the figures get higher) is beyond me. It may be "emergent" or it may be that I haven't quite understood it (<-- likely), but it's pretty good at what it does.

The highest profit I've had so far is over 3 million: 3,331,724 to be precise. But that's after thousands of cycles; even the worst algorithms will throw that eventually.

[UPDATED]: At least a couple of people seemed to take a look at the dirty, ad hoc test script linked-to and discussed on the PFC page. This is just an update to say that it has been retro-fitted to take into account all the bugs and errors mentioned, and has had an upgrade in that:

• It now uses the popen() interface (pipes) instead of system calls to get the value from the algorithm program. This means it's still using IO on some level, but at least it doesn't continually have to make hundreds of disk reads now, which was always the clunkiest part to begin with. It also means it should work on Windows installations of python as no unabstracted system calls are being used.
• It's had a couple of [very basic] command-line arguments added. The number of cycles can be changed manually and there's the ability to suppress the primitive progress bars and the list dump (less IO etc.).
• It'll calculate standard deviation.
• It'll give a rough indication of the time (wall time) the entire process took.

As before, it's still horribly dirty code and assumes nicety on the part of the client program. Read the comments. Use at own risk, etc.

"And that's why I don't like cricket"

Hail, Australia. Looks like you're going to be getting that little jug back pretty damned soon. We apologise for large-ing it quite so classlessly last time.