The ACD Method

Quote from flip:

Mav, do you calculate the sharpe ratio of your trading results? (I'm aware that it's not so straightforward to calculate in a prop trading environment due to not having a specific capital but rather a flexible buying power)

The reason I ask is because you mention turtle techniques, the difficulty of trend following etc.

I actually find it not that hard to develop a trend following system (diversified, trading the usual suspects, i.e. indices, bonds, commodities, currencies) that resembles what typical CTAs are doing. However, that's in the region of sharpe ratio ~ 1, which might be okay when thinking in terms of what CTAs are delivering, but I can imagine that's not really attractive in a prop trading environment, where sharpe ratios will be much higher, although on a smaller capital basis.

So I guess when you say trend following is very very difficult you're thinking of results far better than a sharpe ratio of about 1?

A related - more ACD specific - question if you don't mind: How many trades / positions do you typically have open at the same time using this method? I guess it will vary a lot, but just a general number would be interesting. Do you focus on one / a few positions, i.e. just trading the cleanest/strongest signals? Or do you have many open positions on at the same time (as a typical diversified trend following system would have)?
More...

It's kind of complicated what I do as I'm running so many different strategies. I have an intra-day equity program that is a grey box system. I have a market neutral options strategy that focuses on equities and a swing trading futures system.

Ideally you want to break down the products you are trading into groups so that you do not double up on correlated assets. For example being long crude and ES. They are both the same position for the most part. Or long wheat and corn. Pick the strongest grain or the strongest risk asset.

I don't really buy into most of the standard trend following stuff. The drawdowns are unbearable. That is why their sharpe ratios are so bad.
 
Okay thanks, yes the drawdowns can really be severe with the usual trend following systems...

Interesting idea regarding grouping the assets, your copper vs. ES example was actually a good application of that.
 
ES is flat and Copper is down 5%. Product selection!

Live Cattle breakout still working. Getting ready to take out the highs.

Also FXI down 2.5%.
 
Quote from Maverick74:

ES is flat and Copper is down 5%. Product selection!

Live Cattle breakout still working. Getting ready to take out the highs.

Also FXI down 2.5%.
More...

Chart5.png


http://www.ritholtz.com/blog/2011/0...ampaign=Feed:+TheBigPicture+(The+Big+Picture)
 
I just found out about this blog, and I am impressed with the discussion that you guys are having. I was up all night going through every page - taking notes. Thanks to every one for contributing several key ideas.

I have been living in vacuum not knowing there are so many users of ACD.

Great stuff Maverick74. You have really done the trading community a favor with a lot of information that helps one fine tune ACD and share the different ways to trade it. There is no magic formula, every one has to come up with their own variations of trading the ACD methodology or any strategy in that case.

Since I have had decent success with ACD, I wouldn't mind sharing some of the ideas that I have massaged using Fishers method.

I trade hft strategies in equities, but also have a chunk of my capital running ACD on Futures Black Box.

Since my background is in engineering, and making "Trading Decisions" based on factoring market conditions is somewhat of an art form, I try to stick with a systematic approach. I take every signal I get, but this is where I can fine tune my or super size when multiple signals are triggered - and make my black box better. For example, I take longs in Silver, Gold and Copper - when all of that happens and my trades seems to be very correlated. I would want to work on trading optimal names based on risk / reward ratio and only taking 1 of the 3 or hedging against using a correlated asset.

The strategy has delivered sizeable profits in volatile market conditions (last 2 months + april and may 2011) , however in choppy market conditions I go through drawdowns (which is financed by my HFT strategies that make money every day)

For example, shorts in gold have hurt me - because the machine takes those trades no matter what. However, a human trader would asses other metals before taking the short. On the other hand, I have captured 100+ points in gold taking long trade in the 2008 environment when everything was going down - so sticking with taking everything has worked out so far - but the pain is in the drawdowns when I am flat for 2 months with no action.

I also size up based on Number line and its beautiful when the momentum is building up in the direction of the trend (i.e Number line going from 14 to 16 to 19 etc), however when the number line goes from 24 to 18 over a few days - the computer still thinks that the trend is long and the market stops me out on my AUp trades.

When 30 days ago in case of comodities and in case of stocks 21 or 20 days ago, was an Aup and you get an Aup - those days tend to have big moves - especially in names like AAPL, GOOG etc when funds tend to invest on a monthly basis.

The OR is statistically significant - I m attaching a spreadsheet - which demonstrates, the OR is the high of the low of the day greater than 30% of the time in most commodities and just trading the breakout - gives you a good positive move in the direction of the long term trend.

If you check out the summary sheet, the other thing you will see in the Details of all the commodities is the overnight jump is significant as compared to intraday move. Hence my reasoning to keep my profits run overnight. (Send me a PM for excel spreadsheet, I wasn't allowed to share an excel via ET)

The overnight move is about 1.45 X intraday move in stocks [Did this study on Russell 2000], but you cant have overnight stops in stocks.

Currently, I trade break outs only [Keep positions overnight on my winning trades] - but I have been working on creating a fading system for intra-day trading on non volatile stocks and commodities.

maverick74 - your ideas such as using FAT OR for fading has given me good ideas to think about and work on the Fading aspect of the trades. However its fairly challenging to come up with exits since the prices could be choppy for a while and it doesnt go anywhere when fading at time.

Moreover, as a computer trader - its hard to explain in code, if price hasnt moved in a while - get out. Timed out stops were very effective when trading and watching several markets when I was trading manually. I have 9+ strategies running so its hard to keep track of everything. ACD takes < 10% of all my trades and makes up greater than 30% of my profits as of this year.

I am also planning to trade every volatile and liquid asset using the ACD system as an automated strategy - anything that is volatile and meets the OR criteria of >20% as high or low is definitely a symbol worth watching / backtesting.

I realize that I will have tons of losers but my winning trades are 2 to 4 times as big as my losers and its overall portfolio performance that counts. The MFE to MAE ratio is about 1.4 to 1.8 depending upon what names you are trading. Thats a significant edge when you make expectancy trades.

Further, its hard to quantify - Big Pivot Range and Big OR - but I have come up with an ATR based formula and stick with it. Have noticed these days are choppy days and a good idea to sit out.

Also, use the OR time of the equity market (930 to 940 or 930 to 950) for GC, SI have tend to work better in the recent times - because there is strong correlation between SPY and GC that is existing now - which wasnt the case - 2 years ago.

The other thing that maverick74 mentioned was that Pivots around A levels are good for fading, I have gotten chopped out of 62 trades out of 72 trades when taking breakout around these pivots for the last 2 years. So thank your for sharing this, I will be modifying the code and testing new ideas around this pivot levels. In the book fisher mentions, let the prices pierce through your Pivot if it right above your Aup and stay there for half the time. Hence I would take the breakout only to see prices move against me.

Sharpe Ratios is just a number. I tend to have HFT strategies which have high sharpe ratio of 5+, but a bad day can wipe off significant chunk of your profits when you run mean reversion strategies. I tend to look at sortino for momentum and trend based strategies, and my realized trading performance on silver has as sortino of 8.92 which is a good bang for a buck.

Currently I use the simple rule of letting winners run - however are if there any exit strategy ideas that you guys can share would be appreciated. I hate to have a winning trade in silver which will be 4000 dollars/contract in my favor and then let it go against because I did not take profits out as the market was reversing.
 

Attachments

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