I determined some time ago that I would eventually like to make the switch from discretionary trading to mechancial system trading.
Here are some previous posts from my old blog, over a year old, where I work through Tharp’s questions. I have listed his book in the iBC library under my picks. If you are considering system development, or are early on in the process, they might make interesting reads.
Setting My Objectives Part 1: Self Assessment
Part 2: Defining My Objectives
Part 3: Defining My Objectives
Anyway, I’m excited because I’ve found a system that fits my objectives. Of course, this is the strategy that uses RSI(2) to enter and exit long positions in the ultralong and ultrashort ETFs.
BHH over at IBD Index has been doing some great work, and has helped to take this idea from its infancy into a mature system. Dogwood has also made some great contributions. I will be tracking all the trades associated with this system on Covestor, and have put up a widget on the blog so that anyone can see at anytime how the system is working.
I’m putting this information all into one post so that when I receive questions I can refer readers to one place for answers.
What Is the System?
The system creates a long entry into one or more ultralong or ultrashort ETFs when their RSI(2) closes beneath 10. The system triggers an exit from the ETF when RSI(2) closes above 80. The buy or sell is made the morning after the signal is generated. I am currently using only the ETFs which average at least 100K shares traded per day.
There has been a good deal of tesing concerning stop losses, and it seems that the minimum stop should be near 12%, although a looser stop may work well for traders with a higher threshold for pain. If one can stomach the drawdowns, maybe a stop is not necessary at all. That is not a recommendation, just an observation. I have also been testing time stops, and have found that they decrease net profits without a corresponding decrease in the drawdowns.
BHH has also added another condition, and that is that the ETF must be trading above its 50 day moving average. You can check the results of the strategy with that condition applied here: More RSI(2). I think you’ll find the results promising, to say the least. The use of a percentage stop, coupled with this condition, seem to have minimized the drawdowns produced during earlier testing. Right now, I am not using that condition, although I may switch over to using it, soon.
The system typically produces better than 70% winners. This number will change based on any stops or other conditions that are applied.
Possible flaws in the system are introduced by the fact that most of the ETFs have not been trading for even a full 2 years. In order to account for his, I tested the S&P, Dow Jones, and Nasdaq, for at least the past 20 years. The ratio of winners to losers was very similar to the ETFs, but the annual returns and net profits were not nearly as good. Regardless, the system was still profitable but did not beat the indexes. Part of this can be mitigated by the fact that you were only invested about 33% of the time.
The beauty of the system is when it is applied ETFs. The ETFs are leveraged 2x and offer the opportunity to be both long and short. This is why it should beat the indexes.
There is still work to be done on this system. When multiple signals are generated, which ETF should be chosen over another? Also, is the 50 day average the best average to use?
As these questions are answered, and as success is hopefully achieved, I will update this post with future results.
these are the type of systems, especially with a wide or no stop loss, that lead to black swan events.
Remember that the markets can stay overbought/oversold for longer than you can stay solvent.
I would suggest lots of back-testing focused on finding the ideal stop (your 12% may well be the answer) and then stick to it no matter what.
i personally use the 200 day MA as my trend indicator in my trading system.
I am a fan of Taleb, and have The Black Swan sitting next to Tharp’s Trade Your Way to Financial Freedom on my bookshelf. I disagree with your opening statement. No system can lead to a black swan event. The nature of a Black Swan is that it cannot be predicted and is completely unforseen. Now we can plan and try to throw curveballs at the system to try and break it down, but again, if it is the Black Swan we are planning for, this is foolhardy, as one cannot foresee nor truly understand the event before it occurs.
The system is robust. Keep in mind that this system survived the crash of 89 as well as some big dips before it and after, and still came out profitable, albeit not as profitable as had you bought the index on the day of inception and held it until the last day of the testing. My point is, as long as one is not leveraged 10x, this system should survive. It is not reliant upon any particular market direction. It is not picking up pennies in front of a steam roller. It does not rely on fancy quant algorithms. The biggest challenge to the system’s survival is human psychology. If buying strength becomes back in style, it will not work anymore.
All that being said, I’m not saying it should be used without a stop loss. Anyone who has not tested this and seen the results should be scared to death to trade it, as I would had I started trading a system that I did not know and understand. But again, a wide or no stop loss really has nothing to do with a Black Swan.
Fuck Taleb, whoever he may be.
Come now Fly; surely you jest.
Either that, or you are trolling for Ducati.
I do not jest. I am not a fan, in any manner, of technical analysis.
Therefore, I am unfamiliar with your Gods, like Taleb.
You are just picking a fight.
You know damn good and well that Taleb is not a technical analyst.
And if you don’t, your indequacy is startling.
You should read Taleb’s book, just for the mind expanding qualities of it.
Smart guy. “Godliness” remains to be seen, however.
I hear you asshats making reference to this “TAleB.”
Let me tell you, I have the slightest clue who he is, nor do I care.
Taleb’s a master of the goddam obvious. Market returns are heteroskedastic and not normally distributed, and excessive leverage can fuck up your day when one of those “not as rare as you think” rare events happens.
There! That’s the whole crap book in one sentence.
Bill, excellent! Taleb in one sentence! lol
I read Fooled by Randomness last winter on vacation in Kauai and ironically, there were two black swans that lived in a small lake on the hotel property just outside of our lanai. I was a little weirded out by the whole thing.
Mechanical trading definitely takes the emotion out of the equation, assuming you maintain the strict discipline—which can be hard in the face of a large drawdown, as you know.
There is no perfect system. Eventually all of them experience challenging drawdowns.
I’d be interested in periodic updates of your system.
Good luck with your quest.
Taleb made $40 mill in 1 day in 1987. He has the street creds.
http://www.fooledbyrandomness.com/bloombergProfile.pdf
He main point is that not even Mr. Fly in all of his greatness and time traveling can avoid a Black Swan event. And the other side of the observation is that even if Mr. Fly made 100 correct calls in a row, that wouldn’t predict that 101 would also be correct.
Wood, every hedge fund that blow up (including the one i unluckily invested in) said exactly what you just said to me.
The point is that “Black Swans” are not truly black swans, they occur all the time and people call them black swans after shit hits the fan (to justify their blow up) because they started off on day one thinking that their system “survived” this crash and that event 🙂
Also, you do not need leverage to lose all your money… many people manage to lose it just fine without any leverage simply by not using proper risk management.
B2C:
Taleb’s use of the term Black Swan is within the context of “no system can effectively take into consideration everything that could happen” so therefore just because some event hasn’t been observed yet doesn’t mean it will never happen. (i.e., the Black Swan of Australia had never been seen in England, so it’s existence was precluded from the universe of swan colors as known in England.)The randomness of the financial market, he argues, requires the expectation of a Black Swan event.
And therefore prudent risk management (avoiding taking a position that you cannot afford to lose 100% of) is the only way to survive.
No, you don’t *need*leverage* to lose all your money, you can do it with stupidity alone. You do *need*leverage* to lose more than your original investment (no matter how stupid you are).
However, an “investor” with funds evenly split between just five opportunities, in different industries (or asset classes) would find it practically impossible to lose all their money, if unleveraged. Any trader following an anti-Martingale position sizing system, if unleveraged, would also find it practically impossible to lose all their money, even if their system sucked. They may lose more than they’d like, but they won’t lose it all.
It goes back to heteroskedasticity, or what Eckhardt (or was it the other Turtle co-founder?) described as the “infinite variance” of the markets. 3-sigma events are about 4x as common in the S&P 500’s movements as one would expect if using a normal distribution. I covered that here:
http://www.billakanodoodahs.com/2006/11/streaks-distributions-black-swans-and-golden-swans/
Using leverage combined with assumptions of normality, or risk mearsurements such as VAR, just exacerbates the severity of an event that will happen more often than those techniques *ASSUME* they will.
“Survival” means different things to different people. The best-performing newsletter over the last 25 years (according to Hulbert) has averaged close to 20% annualized over that time. That newsletter LOST 60% in the months leading up to, and during, the 1987 crash. They managed to recover from that to lead the field over the entire quarter-century.
Is that “surviving?” They didn’t lose it all. They obviously have a good methodology going for them.
Would you have stuck with a fund (or newsletter) that had a 60% drawdown?
Regarding drawdowns and mechanical trading, has anyone done any work on anti-martingale position sizing during losing streaks? In contrast to game theory, market losing streaks are less “random” IMHO and more a consequence of market conditions being unfavorable to a particular system which further reinforces anti-martingale position sizing as a valid mathod of reducing drawdowns. Whoa, I just saw Bill’s post during me writing this…
“heteroskedasticity”. I love that term.
(Is that anti-gay nomenclature?)
No, although the word “nomenclature” itself is fairly gay.
I hold nothing against faggots (literally! I hold NOTHING against them!), but I do think that sucking dick is woman’s work.
“Anti-Martingale” makes it sound more complicated than it is.
Assuming you already have your entry and exit criteria based on the system, you now need to size the bet (and doing so may include using leverage). There are only three general ways to do this.
You can bet a constant amount of money throughout, which is kind of stupid, because sizing to R=$1,000 won’t make much sense when you’re a gazillionaire.
You can increase bet size as your account gets smaller. This is what happens when you double down on the next bet after a losing bet (the Martingale), and is also pretty goddam stupid, because you’ll generally not have enough money to run the system if you hit a losing streak. This happens occasionally when running with other people’s money, this is pretty much what Jerome at SocGen was doing, doubling up to get even.
You can decrease bet size as your account shrinks, which is the anti-Martingale. If I size my bet based on Kelly Criterion, this is what I get, each bet being a fixed PERCENTAGE OF EQUITY instead of a fixed amount of money. That way, my 2.5% bet is $3,500 now and is $3,500,000,000,000 when I’m a gazillionaire in a few years. I bet more as my account grows, less as it shrinks.
Here’s my recent post mentioning bet sizes.
http://www.billakanodoodahs.com/2008/05/what-is-the-optimum-bet-size/
Now, at 2.5% bet size (for example), if I hit a streak of 10 losses, I haven’t lost 25% of my money!!!!!! I’m actually down “only” 22.4%. You’d think at 100 straight losses at 2.5% I’d be out of business, but no, I have 8% of my capital left!
In a Martingale system that doubles size each time, I’m out of business by the fifth losing bet – I can’t follow the system for the next one!
In a system with constant dollar size, set to what is 2.5% of equity NOW, I can only take a 40-long losing streak.
So the anti-Martingale method is the way to size, but it’s really just as simple as defining bets as a fixed percentage of equity at the time the bet is made …
I really appreciate a hard working woman.
Wow. It’s work now? Who knew…
When trading, I’ll use a hard dollar stop, or simply position size based on the loss (1% of equity) I’m willing to take to my stop price. Oversimplification? Maybe, but I won’t run out of money.
boca, you have a good attitude.
“Blowjoy.”
“What? Blowjoy?”
“Well, its certainly not a job…”
From Saturday Night Live…
Bill, I’m glad you’ve been visiting, and as always, your input is great.
Plus, where else can you go and bash gays, promote oral sex, and discuss systems and bet sizes, all in one comments section?
Bill,
Read your Black Swan/Gold Swan.
Enjoyed it.
Interesting in that LTCM [and others after them] filled with Math wizz kids, failed to account for the statistical truth that was present at their formation.
jog on
duc
zenProfit:
I know what Taleb is talking about, i read all his books.
My point is that this strategy (or any variation of catch a falling knife) does not account for the obvious let alone the unexpected. When a speculator loses money using this strategy they then blame it on a “Black Swan” event.
Woody is obviously smart enough to have risk management techniques in place so i am not too worried about him. Others, following the post without much thinking may not be so lucky.
Just plot any down trending ETF, like DUG or XHB alongside the RSI(2) and you will see what i mean. Many times during the past year both have hit RSI(2) of under 10 and then went on to have an RSI(2) of above 80, the problem is that by then the ETF had dropped another 10-30%. So while the RSI(2) recovered very well the actual trade was a big loser. Try to do that without using a stop loss, couple that with some poor timing and “black swans” and you could lose all (or most) of your money within weeks.
No leverage required. No true Black Swan required either.
Another reason why using a medium to long term moving average is critical IMHO as you are only taking trades in the direction of the longer-term trend. It removes the falling knife scenarios.. or atleast you only try and catch it 1-2 times rather then death by 1000 paper cuts. You stay out of DUG situations.
I agree with you though about a stop loss. Protective stops ALWAYS hurt performance because they inevitably terminate trades that would have ended up profitable but they are critical to sound position sizing and risk managment.
B2C:
Agreed.
Just plot any down trending ETF, like DUG or XHB alongside the RSI(2) and you will see what i mean. Many times during the past year both have hit RSI(2) of under 10 and then went on to have an RSI(2) of above 80, the problem is that by then the ETF had dropped another 10-30%.
Thankfully, with StockFetcher and other back testing tools, we don’t have to eyeball it.
DUG
Worst Day: – 2.9 percent
Worst Week: – 9.7 percent
Worst Two Weeks: – 16.91 percent
Worst Month: – 16.66 percent
XHB
Worst Day: – 5.41 percent
Worst Week: – 10.03 percent
Worst Two Weeks: – 12.45 percent
Worst Month: – 14.85 percent
So while the RSI(2) recovered very well the actual trade was a big loser. Try to do that without using a stop loss, couple that with some poor timing and “black swans†and you could lose all (or most) of your money within weeks.
No leverage required. No true Black Swan required either.
I don’t disagree with what you are saying, but your comment applies to any and all forms of trading/investing, not just the RSI(2) system.
Every trader needs to answer five questions before entering a trade:
1. What to buy
2. When to buy
3. How much to risk on the trade
4. When to sell with a loss
5. When to sell with a profit
If you can’t answer each question, then don’t take the trade until you can. The beauty of system trading is that the answers to the questions are built into the rules of the system.
Dogwood, of course I agree with you, wholeheartedly.
The discussion has moved somewhat away from the actual system that is being traded, and that is fine, as it has been another great discussion. However, I feel I need to be clear about how I’m trading it.
The system I’m trading is only using the leveraged ETFs. As such, XHB is of no consideration. Furthermore, as I’m currently trading it, I’m spreading my capital across a possible six different ETFs.
Believe me, I’ve been watching the interaction between price and RSI(2) for almost a year now. I understand acutely how sometimes RSI(2) will not cross the exit trigger until a significant percentage loss has occured. However, as long as one is not totally invested and leveraged in that ONE etf, and assuming there is some sort of stop protection, there is really nothing to be overly concerned about. Losses are a part of every system. There will be another trade, and another, and another, and the stats show that likely all three of them will be winners.
I will be using stop protection. Right now, until I figure out exactly how much of my capital I’m willing to devote to this, I will use 12%. Eventually I will be using a position-sizing/percent risk system.
Just for the fun of it, I ran the basic system. I call it RSI2_V1, on DUG. No stops, 10K initial capital, 10K per trade, no moving average component. The results are below. The first column is All trades and the second column is the long trades, so they are really both the same thing. Sorry about the formatting. I don’t have time now to do anything else.
As you can see, the system has held up well, despite the very poor performance of DUG.
All Trades long
Net Profit ($901.37) ($901.37)
Gross Profit $3,870.97 $3,870.97
Gross Loss ($4,772.34) ($4,772.34)
Profit Factor 0.81 0.81
Total Number of Trades 15 15 0
Percent Profitable 46.67% 46.67% 0.00%
Winning Trades 7 7
Losing Trades 8 8
Avg. Trade Net Profit ($60.09) ($60.09) $0.00
Avg. Winning Trade $553.00 $553.00 $0.00
Avg. Losing Trade ($596.54) ($596.54) $0.00
Ratio Avg. Win:Avg. Loss 0.93 0.93 n/a
Largest Winning Trade $1,636.40 $1,636.40 $0.00
Largest Losing Trade ($1,682.36) ($1,682.36) $0.00
Largest Winner as % of Gross Profit 42.27% 42.27% n/a
Largest Loser as % of Gross Loss 35.25% 35.25% n/a
Net Profit as % of Largest Loss (53.58%) (53.58%) n/a
Max. Consecutive Winning Trades 3 3 0
Max. Consecutive Losing Trades 3 3 0
Avg. Bars in Total Trades 11.33 11.33 0.00
Avg. Bars in Winning Trades 6.14 6.14 0.00
Avg. Bars in Losing Trades 15.88 15.88 0.00
Avg. Bars in Even Trades 0.00 0.00 0.00
Max. Shares/Contracts Held 337 337 0
Total Shares/Contracts Held 3362 3362 0
Account Size Required $2,715.30 $2,715.30 $0.00
Total Slippage $0.00 $0.00 $0.00
Total Commission $30.00 $30.00 $0.00
Return on Initial Capital (9.01%)
Annual Rate of Return (6.82%)
Buy & Hold Return (58.33%)
Return on Account (33.20%)
Avg. Monthly Return ($56.34)
Std. Deviation of Monthly Return $785.18
Return Retracement Ratio (0.31)
RINA Index (2.54)
Sharpe Ratio n/a
K-Ratio 0.24
Trading Period 1 Yr, 3 Mths, 4 Dys
Percent of Time in the Market 48.05%
Time in the Market 7 Mths, 8 Dys
Longest Flat Period 36 Dys
Max. Equity Run-up $3,584.21
Date of Max. Equity Run-up 02/12/08 16:15
Max. Equity Run-up as % of Initial Capital 35.84%
Max. Drawdown (Intra-day Peak to Valley)
Value ($3,964.87) ($3,964.87) $0.00
Date 05/21/08 16:15
as % of Initial Capital 39.65% 39.65% 0.00%
Net Profit as % of Drawdown (22.73%) (22.73%) n/a
Select Net Profit as % of Drawdown (22.73%) (22.73%) n/a
Adjusted Net Profit as % of Drawdown (102.19%) (102.19%) n/a
Max. Drawdown (Trade Close to Trade Close)
Value ($2,715.30) ($2,715.30) $0.00
Date 05/27/08 16:15
as % of Initial Capital 27.15% 27.15% 0.00%
Net Profit as % of Drawdown (33.20%) (33.20%) n/a
Select Net Profit as % of Drawdown (33.20%) (33.20%) n/a
Adjusted Net Profit as % of Drawdown (149.22%) (149.22%) n/a
Max. Trade Drawdown ($2,127.92) ($2,127.92) $0.00
@Dogwood: we are both on the same side. My original comment was in the context of not using a stop loss. I am a system’s trader that happens to use VERY WIDE stops. But i have back tested thousands of charts across full market cycles, i spread my bets, go short and long and keep plenty of cash on the sidelines.
I’ve invested, studied, witnessed and read about many blow-ups by people no where near as smart as Wood who have made too many assumptions and neglected risk management and lost their shirt.
My whole investing thesis is that anybody can be profitable year in and year out as long as they employ a system with positive expectancy and stick to the rules.
@Wood: thanks for the back-test. As I said, several times :), I am not worried about you, you know your stuff… You used to read the stockrocket blog if I recall correctly, that dude implemented a very similar strategy and made lots of money. Yet some idiots still lost lots of money by poorly following his system.
Born2Code, I’m trying to debate with you yet you keep killing me with kindness. 🙂 Thanks.
Re: your thesis that anybody can be profitable year in and year our with pos. expectancy and money mgmt…I agree too. The hard part is trusting the system enough to follow it, religiously.
I would like to see someone test random entries into the universe of equities, with maybe only a volume filter, and then apply differing money mgmt condition on the entries, testing the effects. I have a feeling they would not perform nearly as poorly as most would imagine.
Born2Code,
How are we supposed to have heated debates and flame wars if we keep agreeing with one another? Talk about sucking the joy out of life. LOL!
Agreed, we’re on the same side of the debate. Profitability is easy, as long as a trader has the necessary discipline.
Wood,
If I can figure out how to truly select stocks at random, I would be interested in running some tests. I’ll poke around the Internets and see what I can find. Didn’t Van Tharp mention testing with random stocks?
Dog, Tharp is probably where the idea came from.
Aren’t you supposed to be on vacation? The beach picture looked very nice.
I’m back from vacation, unfortunately. Every time we go to Michigan, it is a struggle to come home. One day, we won’t.
Figured out how to do random trading and will have a post up in an hour or so explaining the “system”. Should be fun.
Eggselent…Will be looking for it tonight, if I can stay awake.
Wood/Dog
Positive expectancy is a key variable.
jog on
duc
Excellent post (and comments). Which six Ultra ETF’s are you using? I love this system…
Norcal, I am using a watchlist of all the leveraged ETFs. I am using a volume filter where avg. volume > 100K shares a day.
Due to the fact that one has a limited amount of capital some consideration has to be give to selection, and some of the ETFs have a lot more volume than others, the six that typically show up on the screen are the QID, SDS, DXD and QLD, SSO, and DDM>