Rol's Trading Journal

[Kohanz]

Quote from Rol:

Kohanz, thanks for the interest. It helps keep me thinking about things and not get lazy. I am a Medical Technologist at a major medical center hospital. Working around automated analyzers in the laboratory helps me appreciate the role computers can play in strat automation. The instruments follow programming instructions to perform functions tirelessly. A lot of success in trading I believe is just about being at the right place at the right time, and a human just can't keep up. I mean what person would just sit there for weeks to months at a time with their finger on the buy button, hoping it doesn't turn out to be a "Fat" finger.

Totally agree. Impressive results last week, btw. The equity curve is nearly ideal. You have been running live for just over 4 months now, correct? How closely have you found that your live results are reflecting your backtesting results and how do you perform this comparison (statistically). Do you monitor this relationship (backtesting vs. live) closely and frequently?

Quote from Rol:

I use the TradeStation platform and the limit is 1000 symbols in what is termed their RadarScreen. I don't trade off of charts, otherwise I would need a separate chart for each symbol. The RadarScreen is similar in function, it appears, to what others are doing with Excel.

So you use a market "scanner" to find candidates for your strategy. Do you find that such a scanner (1000 symbols) has bandwidth or CPU issues or is it relatively lightweight?

Quote from Rol:

I have developed into my code the ability to handle disconnects and position mismatches, so that is no longer a big problem for me.

I think I might know what you mean by "position mismatches", but could you please elaborate?

Quote from Rol:

When markets are quiet I am more in cash, but when things start moving, I am slow to enter, but quick to get out.

I think I understand the first part, but the "slow to enter, quick to get out" I'm not sure about. Do you mean that your strat only takes a smaller portion of a larger trend?

Quote from Rol:

My position size is an equal dollar amount per position. This is necessary for me to be able to calculate the maximum number of positions I can take. A lot of what I have been doing recently has to do with determining that sweet spot between position size and maximum number of positions to maximize net profit.

Without going into details if you don't want to - I'm interested in a general idea of what this calculation looks like. Do you take into account commissions, liquidity, etc.?

Quote from Rol:

With regard to entries and exits, I always use limit orders for entry and market orders for exit. I use market orders for exits because when the strat says to go flat, I am willing to give up a few points to do so. There is a bit of slippage in backtesting because it assumes I buy at the bid and sell at the ask, when the opposite is usually true. I am not scalping so it really is not a big issue. If one is developing a daytrading strat with tight margins, I would suggest they assume they have to buy at the ask and sell at the bid to get a realistic perfomance report (unless they are a quant trader working next door to Wall Street).

Sounds like good advice. Thanks!

All the best this coming week.

 

[Rol]

Quote from Kohanz:

Totally agree. Impressive results last week, btw. The equity curve is nearly ideal. You have been running live for just over 4 months now, correct? How closely have you found that your live results are reflecting your backtesting results and how do you perform this comparison (statistically). Do you monitor this relationship (backtesting vs. live) closely and frequently?

I began code development on a portfolio basis using Amibroker because TradeStation didn't have the capability (they say it is coming).
Each night I import the trades that Amibroker took and compare that to what I did. Some days it is 100% the same, but if there are too many signals, they don't always match. When I was discretionary trading all the time and taking big risks, I would compare my performance to Amibroker's and it always did way better. I felt like it was taunting me with a high win rate, while I was being blown out or just trying to break even. I was convinced that somehow it was programmed to cherry pick the winners using hindsite. I thought what would be the motivation of the developers, because the backtesting engine was a one time fee, no subscription. Then I looked closer to the symbols it was trading over several years and noticed that more symbols started with the letter "A" than randomness would suggest. I thought, is that the secret to great results" Just buy symbols that start with the letter "A"? I actually considered doing that just to buy the exact same stocks that Amibroker was. Then it dawned on me that it was just going alphabetically down the list of tickers, and when it got to a trade signal, took the trade until it reached its maximum allowed trades. (since then I have learned to code it to be random). This was great news though because it meant that the strategy didn't depend on picking only the winners to work . It just had to include enough different symbols to hit the expectancy.

Currently, I am performing as well as backtesting, although not exactly the same symbols. Actually the bactester took a long trade in RINO (now in the pinksheets) in november, but is still at all time highs.

So you use a market "scanner" to find candidates for your strategy. Do you find that such a scanner (1000 symbols) has bandwidth or CPU issues or is it relatively lightweight?

The scan is performed when markets are closed and takes about 2 or 3 minutes to complete. I have actually included a 1000 symbols in RadarScreen and it takes a while to load, but doesn't seem to get bogged down during market hours, even with a cheap Dell from Walmart.

I think I might know what you mean by "position mismatches", but could you please elaborate?

position mismatches are when the number of shares your strategy thinks you have doesn't match the realworld number. If the strat tried placing a sell order for say 100 shares and you only had 99, it would get rejected.

I think I understand the first part, but the "slow to enter, quick to get out" I'm not sure about. Do you mean that your strat only takes a smaller portion of a larger trend?

No, it has more to do with the idea of that say you have a target price to buy at thinking that is close to the bottom, wait until tomorrow, because you'll probably be able to get it at an even better price :eek:.
What I mean by quick to get out is the strat monitors intraday price action and once conditions are met, it exits. Some of my trades last only a few minutes. Overnight holding can also be disastrous if the stock were to gap down the next day.

Without going into details if you don't want to - I'm interested in a general idea of what this calculation looks like. Do you take into account commissions, liquidity, etc.?
As a starting point you would take your
Networth x2/Position Size to determine the most number of positions you should be holding. My calculation is a bit more complex though, but that is where I started.


Sounds like good advice. Thanks!

All the best this coming week.

 

[Rol]

Well exited my position in CCME around the open for a $320 loss. At one point I was down $1900 so I consider this a gift.

Reversed a losing position in LPHI but this was not part of the strat. I saw support breaking down on it midday and when I pulled up the news, I saw that too may law firms were hounding it. I know I said I would go 100 percent strat, but things can get boring when you auto trade. These stocks had already broken the rules of a typical entry, so I felt the need to micromanage them. I avoided the temptation to backup the truck with them, which is definately a step forward.

Another nice thing about using a strat on a large number of positions at once is that when orders are firing off, you realize you shouldn't interfere with the system because it is screaming at you "Don't mess with me."

Even with a net loss on the day, my networth closed at new highs since starting this journal. Go figure.

Real-time Realized P/L (Today) ($600.90)

Entries: 6
Exits: 15

 

[Rol]

Placed a buy to cover on LPHI at the open. Not part of strat, but did help with damage control.

I don't plan to give play by plays of all the trades I make in this journal. There is already plenty of that here.. Besides, there are just too many trades. I don't give any one trade more importance than the other. The stocks in my portfolio are all equals to me, so no favorites. I am their Master.
I will, however, be providing equity curves regularly to see how my equity progresses. Also, today I was thinking it would be instructive to me as well as others if I showed side by side charts of my equity curve verses the E minis at the end of each month. What I want to see is how my system performs during the various market conditions that occurred during the month. This is something I have not payed attention to greatly. The backtesting seems to be good in downtrends, sideways trends, as well as uptrends. We shall see.

Entries: 7
Exits: 9
BuytoCover: 1

Real-time Realized P/L (2/8/11) $440.65

 

[endsongs]

Quote from Rol:

Hi all, I have been working on a mechanical trading system for about 4 years now and 4 months ago went live with it starting with $30,000. Some reasons for starting this thread are to chronical my performance, force me to go 100% auto and not interfere with the system. Also feedback, questions are welcome. I could go into the specifics of the strategies I use, but it probably wouldn't produce profit anway if not fully automated. Frankly, I'm not even sure what method to call what I do. Swing trading seems to imply holding times of several days to weeks, while mean reversion is considered intraday (correct me if I'm wrong). Some intraday charts look more like trend following. My average hold time is 2 days. My strat does not look at support or resistance levels, or moving averages, or volume other than to screen out low volume stocks. The couple of indicators I use are not used in the typical ways. It seems to come down to an edge I have discovered that backtests well or even better on out of sample data on a portfolio of about 1400 stocks (no futures, currencies, options, etc.). My theory was that if it worked on around 68% of the stocks out there, why not minimize individual stock exposure by diversifying into multiple symbols. I knew after backtesting on a portfolio of stocks, that automation was the only way to achieve this. Fortunately, just in the past year or so software, programming language and trade server connectivity has finally made it possible to go 100% black box, which is my ultimated goal. I wanted to be able to take the human element out of it completely,even including stock selection and the maximum number of stocks to hold. The results I have included are all the trades me or my system have made. I would say it is about 90-95% system. The discretionary trading dips on the equity curve are when I messed with the system. Starting today I plan to include weekly updates on the system performance and will begin a new equity curve with trading equity of $52000.

So, how did you backtest on a portfolio? Did you purchase an add-on?

 

[Rol]

Quote from endsongs:

So, how did you backtest on a portfolio? Did you purchase an add-on?

Click here to find the answer

 

[Rol]

Saw some nice action day today, with the system working beautifully. I was able to recover my losses from two days ago and then some. It is a process of taking 2 steps forward and 1 step back, which is what you would expect with 66% profitability.

This is how I will guage whether my system is comparing well to backtesting:

• 66% or greater profitable
  [*]Profit factor > 2
  [*]Average $ Profit of Winners > Average $ Profit of Losers
  [*]Maximum system drawdown of 8%

If the sytem begins missing these numbers, then I need to reevaluate it.

Entries: 12
Exits: 11

Real-time Realized P/L (2/10/11) $657.08

 

[Kohanz]

Quote from Rol:

This is how I will guage whether my system is comparing well to backtesting:

• 66% or greater profitable
  [*]Profit factor > 2
  [*]Average $ Profit of Winners > Average $ Profit of Losers
  [*]Maximum system drawdown of 8%

If the sytem begins missing these numbers, then I need to reevaluate it.

Good plan. I assume that as you move forward, older live results will become part of the "backtesting" you compare newer live results to?

 

[Rol]

Quote from Kohanz:

Good plan. I assume that as you move forward, older live results will become part of the "backtesting" you compare newer live results to?

Yes they would if I backtested from today's date (which is what I generally do if I am only looking at recent history, such as the past 4 months). I'm not too concerned about curve fitting with my particular strat because it has never done poorly over a time frame of say > 6 months.

I would like to challenge anyone to post a From-To time frame of > 6 months and < 10 years, where they think my strat would do poorly and I will post the Amibroker equity curve of that time frame. I know it would only be simulation, but you have to start somewhere.

 

[endsongs]

Quote from Rol:


Click here to find the answer

I'm referring to the Tradestation equity curve you posted. That only works for a single symbol unless you buy an add on, right?