Journey from investment bank to independent automated trader

Just checking in. Still working on writing software and tweaking algorithms to get the strategy running. I've been writing a business plan of sorts and a bunch of little routines to help me quantify my trading system performance.

Believe me, the algorithm and the mathematical legwork is what takes the longest. I've spent hours and hours, but a lot of it has to do with really making sure the edge holds up statistically.

There was no -need- for this entry, but I want to point out that people don't just show up one day with magic trading algorithms and get rich. They have to put in a lot of sweat and sift through data, formulate an idea, test if the idea works, and build the framework to test the idea.

One aspect of this framework is components that can be re-used. Starting from scratch implies you have no components and no ideas to re-use or tweak. Starting from scratch is the hardest part.

Part of the insight I gain over time is the fact that many traders take an idea and just assume it will work. There was a panicky thread on ET today or yesterday where the guy lost 60% trying to trade support and resistance on the ES. See, that guy wasn't all that unlike me at one point -- he figured he could take a simple idea, run out, and make money with it. I've been at this game for years, failed, had to go work at the IB, and I've gone through the all the technical indicators, the basic ideas like S and R, and so on and so forth. The point is, an idea is just an idea. It really needs to be tested, and it has to be tested thoroughly. A person has to make the commitment to sit down and test cases, write their results, and continue.

I'm currently working on some "regime-switching" models that I picked up on from some of the books on time-series. I have a concept and some products I'm attempting to apply the trade to and the code in R is now pages upon pages. I have confidence some aspects of the strategy will work, but there are trial and error components to the strategy that are, at times, frustrating.
 
I'm freeeee, bitches! Done with grad school. Time to crank out the trading systems. I'm going to code all night long and implement dis beeyatch.

I feel alive!
 
Quote from lolatency:

I'm freeeee, bitches! Done with grad school. Time to crank out the trading systems. I'm going to code all night long and implement dis beeyatch.

I feel alive!
More...

So were you in grad school or lying about working at an investment bank? Neither one of these can occur concurrently. Looks like grad school.

Anyway, I was looking into the percent bid percent ask functionality in Ninjatrader, but never got around to it.
 
Quote from bwolinsky:

So were you in grad school or lying about working at an investment bank? Neither one of these can occur concurrently. Looks like grad school.

Anyway, I was looking into the percent bid percent ask functionality in Ninjatrader, but never got around to it.
More...

In NYC, most of the financial/math graduate programs can be done in the evenings. They usually either offer the classes after 6PM or mirror the morning classes in the evening.

Moreover, I realize this is the internet and that the tone isn't quite there, ... but you come across as a serious dickhead when you accuse people of being liars. I'm going to assume that you aren't a dick; however, I'm going to insist that if you communicate with me, that you use a more polite form of English. i.e., rather than saying "were you lying", you could simply state the question and ask how I managed to do both at the same time.

I simply have no patience for rude people anymore.
 
Haven't been posting to the thread because I actually have been working 12-15 hours a day at the new place.

That having been said, I did a few things. I decided against NinjaTrader and went ahead and just signed up for TradeStation. It's not ideal, but I needed a personal brokerage for my long-term trades and I wanted to get rid of E-Signal, so I took care of two birds with one stone and got Tradestation.

Unfortunately, it's the weekend and I have a lot of computers and TS requires that each computer get authorized. I also lost internet access because the cable company screwed up and closed my account. I've been trading my personal account at work, made a few trades.

I'm eager to get into the market for trading; however, the real barrier I'm trying to overcome is managing and working with the data. I have gigs upon gigs of .CSV files and I've isolated what I want to trade. It's just that the sheer volume of data and the code required to work with it is a lot of work.

I'm using memorial day to write more python code to help me process very large data sets. R chokes on gig+ files, and C++ would take way too long to write the code in. I need to be able to shift my books, so to speak, from one subset of the large data to the next.
 
I spent about a month+, full time, testing dozens of strategies with traderstation using easy language.
Some were my ideas I was sure would work, others were professional systems, some modified by me. Some were based on many indicators you find in trading books. Tried all kind of combinations and then some.

NONE worked in a consistent, reliable manner when backtested long term across changing markets conditions.

Some generated meager returns consistently other worked well until they blew up.
Eventually moved on.

It is possible but it it quite difficult and time consuming. Don't waste too much time doing this. Better to use your skills working FT for a good paycheck.
 
Quote from lolatency:

Haven't been posting to the thread because I actually have been working 12-15 hours a day at the new place.

That having been said, I did a few things. I decided against NinjaTrader and went ahead and just signed up for TradeStation. It's not ideal, but I needed a personal brokerage for my long-term trades and I wanted to get rid of E-Signal, so I took care of two birds with one stone and got Tradestation.
More...

have you looked at interfacing python to the NinjaTrader api? i've done it, using netpython (http://pythonnet.sourceforge.net/), and i think it could benefit you. i dont remember exactly how i installed it (other than including the NinjaTrader.Client.dll in the python directory), but it is an easy interface to work with. for example, to retrieve the last price for es, it just takes just a few lines
Code: 
from NinjaTrader.Client import *
inst = 'ES 06-09'
nt = Client()
if nt.SubscribeMarketData(inst) == 0:
	last = nt.MarketData(inst, 0)
else:
	print 'unable to subscribe...'
print last

all the dll functions are in the NT help library: http://www.ninjatrader-support.com/HelpGuideV6/helpguide.html -> contents -> automated trading -> automated trading -> automated trading interface -> dll -> functions.

hope this helps you.
 
Quote from pupu:

I spent about a month+, full time, testing dozens of strategies with traderstation using easy language.
Some were my ideas I was sure would work, others were professional systems, some modified by me. Some were based on many indicators you find in trading books. Tried all kind of combinations and then some.

NONE worked in a consistent, reliable manner when backtested long term across changing markets conditions.

Some generated meager returns consistently other worked well until they blew up.
Eventually moved on.

It is possible but it it quite difficult and time consuming. Don't waste too much time doing this. Better to use your skills working FT for a good paycheck.
More...

There's no such thing as a good paycheck, in my opinion. I know I started posting on ET asking about how I could live out of an RV, but I am fairly serious about not working for the man forever.

That having been said, I have been working for the man. Sadly. Perhaps the wrong man, too. I go through these cycles where I wonder if I am fundamentally not a good employee, or what. So ultimately, in the long run, I won't have a choice but to go independent.

That having been said, I've been working on strategies recently, though for my employer and not myself. I was planning on using this weekend to get somewhere. I'm hoping I can post here, so this blog doesn't end up being a "what happened to him" kind of thing. I'm still working on this and I'm going to make it happen.
 
I know it has been tried a thousand times:

Some graduate student thinks he can make big money trading deviations from commercial factor models. He then contemplates raising capital and trading off the commercial models.

I didn't go so far as to get a subscription, but I did build a model to trade the factor relationships and I did get a profitable strategy. Unfortunately, the time series I were looking at yielded only 5-8 trading opportunities per year at low-frequencies, with hold times of 29 days or more.

So, I could:
1) Expand the strategy to several relationships, or
2) Try to come up with smarter ways of trading microstructure phenomenon present in these relationships

The strategies built around the commercial factor models were profitable, but it seems as if the market is generally very efficient when it comes to working around these known relationships -- the faster the frequency, the less efficient the market is, however.

A common problem in a lot of the strategy development was that of structural breaks. I'm taking some time to write some code to detect and work around structural breaks. Mostly re-reading and studying the techniques for dealing with structural breaks. Once I finish this, I will go back and revisit the commercial factor models, mine the relationships, and then implement some kind of "fault tolerance" for these structural breaks in the trading relationships I want to trade.
 
You must log in or register to reply here.
Back
Top