Chabah on Automated Trading

Quote from Hooked2000:

Assuming you go both long and short, are there any times you may not be able to get short at your system prices because of uptick rule.
More...

Yes, I have noticed that a fair number of trades don't execute because the price moves below the sell stop on the market order. I don't know if that is actually because of the uptick rule or just price movement in the stock, but it does happen.

Other that that, I think it is a TradeStation back testing question. I don't think the back tests download each tick I don't see how they would implement the uptick rule in back testing.

That's a good question, I will have to try to identify a situation where it would apply and see if it affects results.

Thanks,

Chabah
 
Well I have been grinding away for days, and the results look very good. I entered the Acme V system, and that seems to work quite well on the daily charts. I entered the Acme Market system, but I couldn't figure out how to set up my chart with all the required inputs, so I am setting that aside for a while.

I have about 8 stocks that I think are tradeable using daily charts. Daily signals are great because you can just review them in the evening and not worry about watching the market during the day. They have not so many trades but good win rates. Here is a sample using a 2 year time frame:

CME, 2.83 PF, 9-3 on 12 trades. This one is pretty illegitimate because all the wins are long, and CME was in a huge uptrend. Not so impressive, but good for long signals.

GOOG short was 5-0.

AAPL 5.61 PF on 5 trades.

SPY 9.73 PF on 9 trades, but very small net profit. Not so impressive but a very reliable signal.

GS PF of 4.02 on 12 trades, nice profit too.

So those are OK, but generally there are not enough trades to make you rich. What I really want are strategies that have a lot of trades but are not in the market very much. Thus, I had to start looking intra-day.

I have some 90-minute candidates. These are OK but still not superb. All are for the last year:

TGT, 4.51 PF on 15 trades, nice 10-5 record.

UARM 2.27 PF on an 18-8 record. Nice profit too.

COH 4.18 PF on an 18-5 record. Nice profit.

All of these were retained because they had high win rates, nice profits (about 15% each), and were not in the market too much (TGT 2.4%, UARM 5.2%, COH 2.7%). If you can cobble 8 or 10 of these together, you should make about 100% per year using 2-1 margin (not 4-1). Not too shabby.

In my next message I'll describe the real gold nuggets that make you say BOOYA and start browsing the Porsche web site.
 
How much slippage do you account for in your tests in terms of percentage?


(I have tried to ask the question about how much slippage to consider in an own thread but haven't got any answers... )
 
Quote from elit:

How much slippage do you account for in your tests in terms of percentage?
More...

I am using either .01 or .02 per share during backtesting, but clearly that won't work for expensive stocks. My plan is to evaluate the actual slippage over a couple weeks of active trading and then use that going forward.

Chabah
 
My buddy and I have been grinding away, and our custom breakout system looks great. In backtesting with reasonable slippage and commissions we're seeing over 100% returns uncompounded per year while being in the market less than 5% of the time and never over night, with the same inputs effective over a good number of stocks (about 15 so far out of 40 or so tested).

Honestly that sounds too good to be true, so it probably is. We won't know until we run a few days of tests to see if the core concepts hold up under live fire, which is why we are going live...today.

More to come.

Chabah
 
Testing has been interesting so far. The structure of the system seems to be working, and it certainly isn't going haywire or anything like that. Slippage seems to be around 2 cents, but I'll need many more trades before I can rely on the average.

I still have the small account limitation, so I'll be arranging for more funding in my account to compensate for the Pattern Day Trading requirements.

The schedule is still to "really" go live January 1st, so we're well ahead of schedule at this point. I'll have two weeks in December to mind the system from home while it runs, at which point I hope to be running a couple other modified Acme systems. All that equals a lot more analysis and development through the end of November.

More to come.

Chabah
 
Trading Plan ... check
Know your market ... check
Good (or great) money management parameters ... check
Psychology ... check

Sounds like everything is in order.

Good Luck,

JJ
 
JJ reminds me to post my money management formula. This is pretty much right out of Van Tharp, but I'll describe my personal implementation for those who are interested.

The account starts with $30k, which I divide into three $10k blocks. The $10k blocks are used as my Equity input into my GetShares function. The GetShares function uses the purchase price and stop loss to compute the risk per share, and a 1% risk parameter to determine the overall risk ($100). Divide risk per share into risk per trade and you have the number of shares.

I actually don't hit the $100 even on most losing trades since I stop out after a couple bars, but nonetheless that's what I use. My average losing trade is much less than $100.

So, each trade initially is .33% risk to my overall account - very small and safe, so I have little concern about going broke without dozens of trades warning.

Often, the number of shares returned based on the risk will cost more than the available equity. For this, I have a BuyingPower function. The BuyingPower function is based on 3-to-1 margin, so it is 4x the Equity, or $40k. It is a very rare trade that uses so much buying power, as the risk and stop loss usually lead to a position that is about 1.5x-2x equity.

So we have the three chunks, each trading on $10k with $40k buying power. Now here's where it gets a little interesting. Because my systems are in the market less than 5% of the time, I can run multiple symbols on the same pool of equity. For example, two symbols will very rarely both fire at the same time, and even when they do the BuyingPower will compensate. Further, since I really have three pools of equity, I can run many more symbols since even if 5 or 6 are in the market at once I will still have enough buying power.

So, that's pretty cool. Each trade risks .33% of my overall account, but I can run against say 10 symbols. So at any given time, I may have several % risk in the market.

Where this gets really interesting is in my newest system, which is in the market less than 1% of the time. I am not sure how many symbols I can run, but I suspect it is north of 50. Each symbol only needs to return 6% uncompounded per year for a 100% total account gain. Perhaps that is "LOL" but that is what I am trying to figure out.

More to come.

Chabah
 
While I am arranging funding and running just a couple stocks, I re-read this famous acrary thread:

http://www.elitetrader.com/vb/showthread.php?s=&threadid=33654&perpage=6&pagenumber=51

I hope to share the same sort of information, however from the beginner's perspective rather than the retiring expert's. I don't know what will work or if I am necessarily on the right track.

In re-reading, I can tell that I have subconsciously incorporated much of Acrary's methodology in developing and testing my system. Near the end of the thread he posts his DUM systems development summary - that is exactly what I have done with my system.

Let me explain.

I started with a concept taken from Professional Stock Trading by Mark Conway. The concept is that volatile stocks want to be volatile, and when they consolidate into a certain chart pattern they tend to break out hard.

The understanding phase has been working up the code and running hundreds of tests over various timeframes for dozens of stocks, also varying the commissions and money management parameters.

The next phase is live testing, which has begun as a drip but will soon be a smooth flow of 10-20 trades per day.

One thing I didn't see in the acrary thread is commissions and slippage. I know from my limited experience that a hugely winning system on paper can be quickly destroyed by commissions and slippage. acrary obviously knew that as well, but I don't see how it was worked in to his examples.

My original custom breakout system looked fabulous on 15 minute charts. The profit factors were just ridiculous. So I turned up the slippage heat, and many of the stocks became huge losers. I switched it to 45 minute charts, and voila, back to profitability.

My biggest concern right now is that my forecasted profit factors are in the "miracle" category as described by acrary. And these aren't based on 10 trades - I have a few stocks lined up that have a PF over 8 and return near 100% annual on over 100 trades. Now that is the exception and most are more like PF 3, but that is still in the "miracle" category. There is about 0% chance that my first system is a miracle (I am not religious) so I expect an unpleasant realization sometime in the near future.

Anyway that's an update on the thinking and progress of the system. I highly recommend the 53 page acrary thread to anyone who hasn't read it. I am not quite bought in on the Monte Carlo analysis stuff, but I agree in spirit and I hope to continue the discussion here on ET.

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