Piker's journal

Hard disk crashed :(

That's because I choose simple but inefficient method of storing quotes. Initially I tried to store quotes in MySQL database but it's certainly not built for tables of 20+ gigabytes. Even simple queries take very long and unpredictable amounts of time. In case of power failure or computer crash (happens quite often), database should be "repaired", which also takes forever.

So I decided to store quotes in binary files under directory structure like YYYY/MM/DD/AAPL.dat. Then another problem arised - when the amount of files exceeded 3 million (and I am only half-way through!), the MFT of my hard disk got full. Performance decreased drastically and I decided to use a tool named Diskkeeper to increase the size of MFT and to defragment it.
Alas! After Diskkeeper played with my disk I was no longer able to boot from that disk. fixboot and fixmbr didn't help. So now I am reinstalling everything.

After googling a bit, I made a pair of batch files which are intended to reserve MFT space for large amount of files. This better should be done on a freshly formatted disk so that MFT comes in one large unfragmented piece.
 

Attachments

-5% in July. Earned gross $117 on 20 trades, paid $229 in commisions. It's too early to tell based on only 25 trades, but all this doesn't look good so far. May be the problem is, system was tested on preceeding months when market rose, and now we see some downturn in the end of this month.

Encouraging signs: average gain on each trade (not taking commissions and leverage into account) was +0.5%, compared to 0.8% the system predicts. Not too bad. And I have few more ideas how to improve my system.

At last finished downloading all 2003-2007 quotes from OpenTick.
Now I have ~5bln quotes in my database. Somewhere within that mountain of numbers GOLD is hidden! I only have to unearth it.

In fact it's a filtered subset of actual ticks, because I'm not interested in sub-second precision and i don't store bid/ask sizes. Now incremental updates will take only few hours on weekends.

Some educating results: I tested and manually optimized one system on a 4-months period. It showed average gain 1.81%, 58 winning trades, 9 losing trades. There are 4 parameters (optimized manually) in that system. However the system completely failed on out-of-sample data, that is 2 months after and 2 months before the optimization period! Talk about statistical confidence. It is possible that it captured some tendency specific to those 4 months, who knows.
 

Attachments

Sh.t, I m not learning from my mistakes. Twice missed profitable signals today.

I sit here, waitng for signals. Program screams "Buy!", I hasten to enter my order and click "Send", and then I receive ... "You were logged out because of inactivity". Because I forgot that I must click that stupid WebTrader window every few minutes.
 
-20% for August.

Nothing to say. It sucks to be a loser. But I'll keep posting until the death of that account.

Questrade is launching new trading platforms today, may be I'll have more luck with better order entry interfaces. I wish one of those platforms had an automation interface. It sucks to spend time in front of computer waiting for signals. And it turns out any discretional trading decision I make is wrong.
 

Attachments

You must log in or register to reply here.
Back
Top