Perhaps not as fun as a Slashdotting, but my Spamblogging site was posted on Fark today (or last night from the look at the logs).

Here is the thread on my blog. I normally get at most 100 hits a day (looking at my ad impressions), but have hit over 11,000 in the past two days. Hasn't really done too much in terms of ad revenue since a Fark'ing doesn't necessarily bring you those that are looking for ad type things.

It is amusing to see the comments. My favorite is the fellow that says my site is ugly and hard to read - THAT is dead on. I appreciate honesty when I can get it.

I have been using http://stenz.livejournal.com as my blog for a bit now - replacing my journal entries on Slashdot.

In addition to that, I have also started two more blogs.

One blog is for the alumni of my college (Williams College). That blog is at http://www.ephblog.com.

The other blog is for those that are interested in learning about and/or discussing the techniques used in technical analysis of the stock market. Neural nets, genetic algorithms, indicators, oscillators, charts, etc and all of the math concepts along the way. That blog is at http://www.stenslandsystems.com.

I am also considering creating another one for those studying for actuarial exams since there seems to be very little content (or at least good content) for them on the web.

And I am in the midst of a web venture to try to make money. I'm not sure that I can pull it off, but I think the idea is a good one. I'm very surprised that nobody has done it, but I have looked and have been unable to find it yet - so I am going to press on with it and hope I can do it.

It has also been a long time since I posted in this journal.

I have this idea that I want to get geek feedback on. I will post it here in this journal in the hopes that smart folks will read it and laugh at me. I am contemplating submitting a version of it to Ask Slashdot to get even more feedback.

The idea:

I have a bid out on EBay for one of the old school brick cell phones, it looks very nearly identical to this one.
With that phone, I'd like to also buy the largest, cheapest GSM phone that I can find and gut the two phones. I then want to put the working parts of the newer GSM phone into the gutted shell of the big old brick.

The end result is a functioning phone without the health hazards but with all of the coolness of obscenely large phones. I could wall down the street yelling "BUY SELL STOCKS BONDS" and "THE MARKET NEVER CLOSES!"
Doing that with a tiny phone or worse yet a Jabra device and people think you are just insane instead of... well, insane but on the phone and at least talking to someone else... theoretically.

Whatever, the point being I want to swap the guts out and carry around a big obnoxious phone just to fly in the face of conformity.

I am hoping for any suggestions and or ideas on what pitfalls I may run into, or things that might make it easier.
I am not interested in a working old device - that stuff can hurt you. I just want the shell.
Is someone else already selling something like this? If so, I have no interest in doing it - I don't plan on selling them, but I don't want to work on it just to find out I could have just bought one.

I have largely been using my LiveJournal page (stenz.livejournal.com) instead of this page as my journal. But I just finished something sufficiently dorky that it is likely better suited for here for a bit - then will later discuss it more in depth there and then also post up the source code to PerlMonks.

I have been trying to figure out how many possible game boards there are for the game Scrabble. It is a far from infinite number, but it is a huge ass number.

I figured, if given the competition dictionary, then you have a fixed pool of words that can be used, and therefore a solvable end number of possible boards that exist.
One could write a program that would just brute force its way through the entire board space and track what all of those boards are.

The problem is when to stop? How do you know when you have all of the possible boards?

It is easy to figure out how many possible boards there would be if you just tried every combination of tiles available on the board space. (there are 100 tiles and 255 spots on the board - that means you can do easy combinatorial math and see that there are something like 10^80 possible boards - that is off of the top of my head from another time I worked it out - that is not exact).

That number is huger than huge. But it is way larger than the real solution space, which is limited by the rules of the game, as well as the competition dictionary which determines that actual words and a limited set need to be formed, not just every combination of tiles on the board.

That should dramatically reduce the space that we are looking at - but by how much?

Well, I wrote some Perl code just now that looks at the competition dictionary. There are over 172K words in there.

I looked at it from the point of view of a blank board and the first starting move.
We know that the starting move is going to be 7 letters long or less.
That allows us to ignore more than 2 thirds of the possible words as our starting moves - down to over 50K possible words that we can start with.

This is a huge break - if that number were huge to start, it would only get huger real fast.

So from that first move, we know that it can be in the number of positions as it is long (a 2 letter word can be in 2 different positions, and a 7 letter word in 7 different positions on the board to start - at least one tile needs to start on the center tile of the board for those that don't know Scrabble).

From there, we look at each letter in the word and determine what words could be made to spawn off of there. If the same letter occurs more than once in a word, we still count the number of words available for it again since we are counting the probability space.

So we add each of those up and then we multiply that number by the number of letters in the word since there are that number of possible starting positions for the word.

So we iterate over the 50K+ words, and then see how many next possible spaces there are. We are one level deep.

To go more levels deep gets more complicated in the programming and the numbers involved start growing a lot - so I only ran it out one level just to see what sorts of numbers we were going to be expecting.

The number I got with the competition dictionary is 147,993,683,161 possible boards for the first two moves.

Things that aren't accounted for at this point:
1) I haven't looked to see what the cutoff for Perl's int is and when I will need to go to Math::BigInt - I presume fairly soon. The only reason that would be a worry is if that number that I got wasn't the real number and instead some overflow or rollover value.

2) I don't account for shifting off of the board. Technically we could start with a 7 letter word and after that use a 12 letter word that goes off the edge of the board - I don't check for that. Not to mention that also would defy the number of letters available to the second move at that time.

3) I don't watch the bag at all - this doesn't simulate real play. Similar to number 2 in that sense. That means that there might be 40 letter Zs on the board, but there should really be no more than 3 (1 allowed and then 2 blanks).

The issue at hand is do we care about the error words creeping in and throwing off the count?
If we can say that we don't care, then that is very good because we can save a huge amount of computation each turn and therefore crank through this quite quickly.

At first I thought it wouldn't matter - but as I write this I see it does matter.
I was thinking in a linear way - which is how us humans normally thing - but this is growing a non-linear way.
For every error that creeps in and is assumed to be a legit word, it is magnified by the possible moves of the start word - so up to 7 times the errors straight off - and then it will have its own children spawned off of itself - and none of those should count towards the total.

So while I initially thought it wasn't that big a deal if a few hundred thousand errors crept into the first one, it does matter because it adds many many more to the end total.

As a result, I am seeing that this code is a rough draft, but currently too large a number...
In the end, that is a good thing - we want this number as low as possible the whole time.
Both for brute forcing, but especially in terms of tracking the number of boards. Each board is a 255 char string - if there are billions of them (which it looks like there are going to be more like trillions at this point) - then can we really track that?
Say we wanted a database that had a table that was just "boards" - could we have trillions of rows? I'm guessing not, but I honestly don't know.

A database that could handle that would be very nice though since you could then have a massively distributed effort to brute force the boards and have the boards submitted and tracked.
This adds in the issue of error checking the incoming data to see if there are the proper number of tiles and the like.

Anyway, this is the first step in all of that.
I'm hoping the addition of error checking code in my Perl script will allow me to get that number down in the millions - but I'm suspecting it will remain in the billions for now - which worries me for later growth.
At least we know that we will only approach the 10^80 figure but never reach it by a long shot.

...put on sale?
Polyester Bride by Liz Phair (sp?)

She is a cutie.

I am starting to learn interesting things about fractal analysis of the stock market. Not the Brownian motion stuff fully, but along the same lines. geometric instead of cumulative sum stuff.

According to my code, which I have yet to optimize (or more importantly check for massive bugs), it says that for the ideal 10 stock portfolio out of the thousands of stocks that it has access to in my database, the following were the best:
ESS, MED, PNP, GV, MGA, MRGE, TRO, NVEC, CTSH, and INKP
(I wouldn't bother looking at these in the short term - this particular code doesn't look too much at short term bubbles - so more like a year out)

The ideal management would be to watch them during intraday trading and manipulate how much you own of each so that it remains as 10% of each.
That is hard to do without either only doing that all day, or using an automated system to do it for you.

The second way of doing it, and easier to maintain is to wait N days and then reconfigure it (so that you have equal parts of each - normalize it and redistribute).
For a portfolio of 10 stocks, the shortest time that you can wait is about a year.

So you choose any 10 stocks right now - my code suggests these as having the current highest likelihood to give you good returns from what it knows of them.
Put your X dollars into them, wait a year, look at what you have an reallocate them to even out.

If you want to come very close to or perhaps beat the indexes, then you can just spread it evenly to start and each time you go around (yearly in this case).

Were you to be changing it constantly, during intraday trading, then there are fixed formulas for calculating how much of each - or you can still do fairly well with just the even distribution.

My current stock database lends itself well to this analysis since you need at least 2500 days of data to get a pretty accurate end result from the code - and my current database is largely focused on at least that much data (if not more).

That said, I always want to increase the size of the database and the stocks available to me and therefore see if there are more possible stocks out there that would possible be better.

The issues with the larger database of stock are increased rate of problems. Stocks delist and my current system doesn't handle that very well. And it slows things down with each new stock.

If anything, I should probably focus on paring it down - but that is a hard thing to do (to a point).

tomorrow is my last day of work before my vacation for the rest of the month. I got a raise and a bonus today, nice things and higher than I had expected - especially since I hadn't been expecting anything.

I created a live journal.

while I will still read slashdot, I will likely not keep a journal on here any more and instead do so on the live journal.

stenz.livejournal.com

I was going to the bathroom at work and I'm the first to admit that there are a few places where all of my good ideas come to me, and they are all when I'd otherwise be ragingly bored if my mind didn't wander the way it does:
1) college classes - I don't have those anymore, but those were great times for my mind to check out and think of other stuff
2) reading a book - this is unfortunate since it slows me down in actually reading the book when my mind wanders off
3) sitting on the toilet
4) taking a shower
5) falling to sleep
6) commute (usually the commute home - the commute in is more focused on getting there fast, weaving in and out of traffic (I'm on a scooter) and not getting killed)
7) long run and/or walk (I try to avoid walking anywhere when possible - I'm not lazy - in fact I'm quite athletic, I just get too bored walking - as a result, my mind wanders)

Anyway, all of that aside, my eureka moment was when I was thinking through the correlation of the proof that a meandering river is non-random via its relation to Pi, and the price movement of a given stock over time.
I then thought of a way to use that to predict stock movement.

I was both overjoyed with something new to grind on, but saddened that if I tell people about it, they will first blurt out something about the movie Pi since he was looking for a pattern in the stock market via Pi... that was simply a narrative tool and off base.
This on the other hand is 100% valid - in theory.

That said, the more I think about it, the more I think that I can either disprove it, or show that it is no better than just looking at a chart with a certain type of line drawn over it.

That said, I am still going to program it out to see what I can find of it.

This is the sort of stuff I wish I had someone to talk to about it - but largely speaking, either nobody gets it, or nobody cares. To assume the former I think gives me too much credit.

On a side note, I'm signing up for LiveJournal and will be integrating it with an update to my website that I will do starting on Dec 18th (I have off from the 18th until early January so as to use up all of my vacation days that I hadn't used previously).
I'm thinking I will actually pay them $25 for a year of their service. Ever since I got this Mac, I've been paying for everything. I don't know if that is good or not. I guess it is good to those getting the money - Proteus, Pith Helmet, charities, and LiveJournal... okay, I guess that really isn't that much.

http://www.keurig.com/two/index.asp

Those men and women that create that should be worshipped as gods. That is my new religion - we have one of those (the larger office version, not the personal addict version) in the office, as did my previous employer.

As a bonus, it is also apparently bad on the environment since the little K-cups aren't recyclable. They burn everything here anyway... well, not everything - witches they drown. But trash gets burned.

We know that HP, Coke, iTunes, WalMart (or was it KMart), MusicMatch, etc are trying to do online downloads of songs. You pay $N and then get N songs - instant gratification.

The annoying part is that they are all competing for the same customer base. In the process of this, they approach XYZ record label and make deals to that they can get content to distribute. They also sign exclusivity deals, so that means a song/album/discography by band ABC (just an example, not necessarily the young wunderkind Another Bad Creation) may only be on one service.

Because of this, you have to go to iTunes to get your fix for one band, Napster for another, and then maybe the Coke store for yet another - each having their own interface.

Much in the same way that a few open source P2P clients have combined the background processes of several sharing protocols and server groups, it would seem that it would be possible and/or desirable to do the same for the online music purchasing world.
Or in the way that Proteus on Mac and Trillian on Windows have made a client that can use all of the chat platforms.

You would either have a web page or a client that would keep track of your various user ids, and then you could enter a search in one place, and it would then turn up the hits on the various services that you have a login on.
Then you could see that only iTunes carries XYZ, but 3 of them carry ABC and Napster has the best price.

Sort of like Pricewatch, but for the online music downloads.

The issues would be that they likely use their own proprietary communications protocol - even though in the end it is just queries into databases and links to correlate to those and then end data.

Also, on an unrelated note - I saw an incredibly hot girl at lunch today. Either it was Claire Forlani, or someone that looked exactly like her. Not sure how tall she was (aside from shorter than me), and I couldn't tell if she had a Brit accent or not. Otherwise I would say that it had to be her.
Hot.

I went out and bought a 20gig iPod for my fiancee. Here it cost me $630. It comes with a 1 year warrantee on it, but since they have that known 18 month battery issue, I upgraded it to a 2 year warrantee for $90 more. Lots of money.
If she doesn't like it, I will sure as hell use it, but it isn't intended to be the type of gift where you get it for someone in the hopes that you will get to use it.

Maybe after this she will decide to get me the phone that I want - my previous hints that I wanted it resulted in her making fun of me since I hate using the phone. Which is a valid point.