Hi all,
I have to echo the sentiments of several of the more serious traders/subscribers here and say upfront that I seem to subscribe to a great looking system JUST before its biggest drawdown.
After thinking on this, I wonder if, rather than just being a jinx, there is some other statistically sound logic at work.
Think about it: a system, call it system X, has a great run…and shows up in all the rankings as a major force, gets all the buzz, gets some good reviews, etc. But of course that is BECAUSE it is performing. However, ANY AND EVERY SYSTEM has drawdowns. EVERY ONE. NO EXCEPTIONS. There is no such thing as a 100% win percentage with any system with over 100 trades. I defy anyone to show one that has over 90% for 100 trades or more (and a REAL 90% that really includes fills and commissions, not the version that C2 reports, because even though it is a valiant effort and continually improving measurement, is FAR from accurate). SO, since that IS the case, perhaps it makes sense to find a STABLE system (that is, one with enough of a track record to where you get a feeling of the TRUE max drawdown, or at least a good MEAN of the average drawdown), then if the system is robust and reliable and well-supported otherwise, just WAIT for the drawdowns (because they WILL happen) and only AFTER them get in - because they are likely to actually go UP if the system has a well-documented profitable history! I know it sounds stupid to be contrarian, but if you think about it, it actually makes sense! Note that the system has to be profitable and stable in general over a decent time span so a real feel for the true max drawdown can be grasped.
W
Walter, I agree and am in the same camp (have joined a few systems here just in time to watch them have their worst streak to-date). I agree that it could be a cyclical phenomenon, kind of like when a major magazine like Time puts a positive stock market story on the cover, you know right then that the bull market is over.
Buying into the (hopefully) bottom of a drawdown on an otherwise good system takes both good timing and a lot of cajones (similar to fading a strong trend)
Maybe we need to start a bad-timing support group and start nominating potential systems and when we think it would be a good time to join them. Its like performing TA on the systems themselves, strange but maybe logical.
My two cents:
For lots of reasons, C2 tends to highlight highly risky systems. Part of that is because people are drawn to those fantastic equity curves, and so make those systems more popular, which in turn drives more popularity (mentions on the forums, hot hand listings, etc.)
I admit I need to do a better job of drawing attention to less risky systems. (I am working on some stuff that will do just this.)
There is no such thing as outsized return without outsized risk. It’s that simple. I am not referring to any system in particular. It’s just the nature of this business. You want fabulous unbelievable returns; you will take fabulous unbelievable risks. And that’s fine. There’s nothing wrong with wanting to take big risks and thus potentially make large returns. It’s important that you know you are doing it, though.
Now, if 100 system vendors all roll the dice and try their hands at very risky strategies, then quite a few of them will pan out. But you only hear about the gamblers that won. Those are the guys that draw attention to themselves and attract subscribers. But as all gamblers know, a lucky streak only lasts so long.
People post things on C2 like: “I must be unlucky, because as soon as I join a system, it does poorly.” That is the wrong way to think about it. You join systems when they are behaving particularly luckily, and then you experience reality. If you had searched for systems on C2 that took lots of risks, but failed really really badly, and then you subscribed to them, you’d probably be surprised by how lucky you are – as soon as you joined the loser system, it turned great. But of course that’s not what happened. You found a system that did unusually poorly, and then it just turned normal, which is what you’d expect. (That’s what reversion to mean is really all about. It’s not that “you can only do well for so long” – it’s that, when a human being notices something unusual, that’s exactly what it is…unusual … and things are more likely than not to get more usual going forward).
Anyway, this is all a long-winded way of saying there are two guilty parties here on C2: Me, because I have not yet succeeded (though I have tried!) in creating a set of tools that will allow users of C2 to truly judge risks and rewards accurately. And you (the collective you out there – not just Walter) who think there is some magic formula, that it is possible to make lots and lots of money without taking commensurate huge risks.
If it was that easy, lots of people would be doing it, and we’d all be billionaires. But it’s not and we’re not.
So: My message is, Look further than the skin-deep equity curve when you judge a system. Look at the risks it took in order to make the gains. Look for lower-risk systems.
I will work on adding tools to the site which will allow you to find them.
I want to stress that I am not pointing at any systems in particular. No vendors behave dishonestly. Everyone is trying their best. And some systems really are great – for those users willing to take the risks that the systems do.
But I’m admitting I need to do a better job of helping users analyze risk and make better decisions. I’ll re-double my efforts.
Matthew
I sincerely hope that you don’t get too dependant upon Sharpe ratio. While it has it’s place as one of many tools with which to judge a system, it’s terribly flawed when used with trend following systems. In particular, it drops considerably when there is a large win! And long term systems by their nature will have occasional large winners. Consider: A system that returns exactly $1000 per trade will have a large Sharpe (it would trend toward inifinity…). But throw in an occasional $20,000 win, and the Sharpe will be lousy.
My son does portfolio research for Mellon Finance. He says that they NEVER use Sharpe for their analysis of long term system. Rather they use the Sortie (sp?) ratio, which penalizes -downside- performance, but doesn’t penalize success. (Actually, there is a small success penalty, as large gains will raise the mean, which then penalizes the poorer trades even more, thus lowering the systems overall score. But this effect is quite minor.)
I don’t mean to pick on particular systems, but look at the Sharpe and graphs of the following two high Sharpe systems: DiJiT and ER2 Cash Cow. (which has a Sharpe of 3.226:) In these cases, Sharpe seems a bit unbalanced .vs. the equity graph, at least in my opinion.
And you (the collective you out there – not just Walter) who think there is some magic formula, that it is possible to make lots and lots of money without taking commensurate huge risks.
I don’t think anyone who has been investing, trading, or living life for any length of time has this illusion!
Walter,
Take a look at my BTS Bottle Rockets system. If there ever was a candidate for playing the drawdowns, this would be it. You can draw a nice straight support line for the periodic drawdowns.
Brian
Aha, yes thats EXACTLY what I am talking about…as Matthew said, its about MEAN REVERSION, which is not an unknown concept but it is something we might be able to EXPLOIT and use to enhance results. The caveats are:
1) the system must have enough history for there to BE a dependable mean reading.
2) the system must have a positive equity slope as its mean.
3) it would be helpful if these C2 graphs were updated more than they are. Its impossible to use them as they are - Matthew, is there any way of upgrading those charts - more resolution and more up-to-date? Then we could visually see a good candidate (like brian’s)…I almost feel like we should be able to devise an revert-to-mean formula and be able to score the systems - at least those that meet criteria #1 and #2.
Just as a back-of-napkin example also, my ER2 Cash Cow system has a real drawdown so far of about 20%…if someone turned the system “on” when drawdown was at or over 20%, then shut it off after the max consecutive profit before drawdown (which I’m not even sure of), the profits would be insane. Again, its reversion to mean.
Good food for thought and ideas to explore.
W
Longer term trend following systems will naturally have open equity drawdowns, as positions are closed only after the trend changes - thus you aren’t selling at the high (or buying at the low for that matter.). As such, the drawdown represents sort of a portfolio “cleansing” and makes the drawdown a good time to start trading…
The slight downside to all this is how many systems have we seen on C2 which crash and burn -never to return -I spuppose what we’re saying is let a system establish some sort of consistant record then jump in when it shows a dd as with Walters cash cow which appears to be nicely moving out of dd -good thought about jumping in at that point!!
Good luck with that mean reversion strategy. Buying the dip has its consequences. Unless the trader is proven over a long term period (2+ years) of handling adversity not just ability to make money, you’re gambling at vegas.
But could you actually buy 14,255 shares of ICCA (for example) and get a decent price? Today that stock only traded 50,000 shares which is right at its average. These types of trades can’t be followed in a real account with any hope of getting the fills shown on C2, and therefore the performance shown in the BTS equity curve isn’t realistic.
I don’t know what that question has to do with this thread, but the answer to your question is no, you can’t. You can only get a fill equal to what the market is advertising at the moment of execution. Also keep in mind that historical volume doesn’t determine what your fill will be like. Only the displayed size at that moment does. That’s were the RF lacks in portraying tremendously.
Tarek … It has a lot to do with this thread since the basis of “fading the drawdowns”, if I understand Walter’s point, is identifying the patterns from the equity curves shown on C2 in order to pick an appropriate system. If those curves are not realistic (which they aren’t for the BTS system and such thinly traded stocks) then there is no accurate way to choose a C2 system to try this approach. So I wasn’t asking if an accurate fill could be gotten on such a trade (obviously it could not, which was my point), but challenging Brian’s offer of his BTS system as a candidate.
Ok, I see your point Randy. Thanks for the clarification.
Yes, thats exactly true, Randy - even IF this was a good methodology, we would face two initial hurdles:
1) can we trust the results posted on C2? So far, the jury is out on that question, depending on many factors.
2) does the system/vendor being considered have enough of a track record to be of statistical value?
If we could have faith in #1 and #2 (which I surely don’t thus far), then we could move on to an analysis of the systems that would seem to qualify…
W
The longer the time period positions are held, the more statisticaly significant they become in judging their truthfulness.
The proper term is “regression to the mean”. And you will never be able to exploit it. That’s mathematically impossible. The pitfall to believe that you can exploit it, is known as the “gamblers fallacy”. A simple example of the gamblers fallacy is if you throw a dice a few times, get a low outcome each time, and then believe that the next time you must get a high outcome. It just isn’t true. The events are independent of each other, and therefore the probability that you have a low outcome the next time is exactly the same as it was the first time.
Regression of the mean therefore doesn’t mean that after a series of losses you will get profit. It means that after a series of losses you have still the same chances as in the beginning and therefore probably something that is better than these losses - but not necessarily a profit. With a dice: If you have five times a 1, then the expected valus for the next throw is still 3.5 and therefore better than these 1s (= regression to the mean) and you cannot conclude that probability of a 4, 5, or 6 will now be larger (= gamblers fallacy).
Let’s assume for simplicity a very favorable situation for the system in that there some kind of stable probabilistic pattern in the market and that the system is adapted to this pattern and therefore has a stable probability of a profitable trade. More specifically, assume for example
- the system has a probability of 2/3 to win on each trade;
- you always either win or loose $1000 with each trade;
- the trades are probabilistically independent of each other;
- the system dies whenever the total value of the account becomes 0.
Under these assumptions, you can find in every book on fundamental probability theory a proof that the system will eventually die with probability 1. So there is no point in trying to estimate the ‘maximum drag down in the long run’. It is already mathematically certain that this will be 100%. If you estimate a lower percentage, it’s just because you didn’t wait long enough.
However, I would already be happy if it doesn’t die while I am alive. I cannot be sure of that either under the above assumptions, and besides I don’t know when I will die, so I will even be satisfied if there is only small probability that the system dies soon. Obviously, the parameter that determines this is the probability of a profitable trade. A system with a high win probability will have a long expected lifetime.
There is also another parameter that should be considered here: The number of trades per day. If the system has many trades per day, then the probability of a profitable day will be larger then the probability of a profitable trade (assuming that the latter probability is larger than .5). For example, if the system has 5 trades per day, each with probability 2/3 to win, then the probability of a profitable day is binomial (p = 2/3, n = 5, k >= 3) which is .79. This is just the law of large numbers working.
Now the profit curves are plotted per day. So a system with many trades per day will automatically have a smoother curve with smaller drag downs. However, this doesn’t make it a better system. Obviously, under the above assumptions it doesn’t matter for the expected profit, and neither for the risk, whether you do 5000 trades in 5000 days or in 1000 days. Thus, such as system, although it will exhibit a relatively smooth profit curve, just uses its lifespan five times as fast. This last property might be important in relation to my own limited lifespan, though. I will be rich sooner, or I will be poor sooner; but the probability of these events doesn’t change.
So I don’t think that technical analysis of system curves will do any good.
Randy,
I’ve always had difficulty with this argument here at C2. Yes, I agree that some of the trades don’t add up volume-wise. But it’s because, as you point out, my system is not really meant to be used in a $100,000 account. But what if someone only has $10,000 and want to use this system? If they simply divide the number of shares by 10, then the trades make more sense, and the equity curve DOES make sense in that case.
Sorry about that Brian, you are correct. It would be useful if the C2 system pages allowed vendors to specify a starting account size other than $100K to base everything on for that page, then used a single scaling factor for each system to make the other comparisons it does for best/worst/hot hands, etc. This way at least the very important trade summary tables for each system would represent how the system vendor would suggest the system be traded in real accounts. It would also avoid forcing unrealistic quantities in the tables just to satisfy the $100K starting account requirement (and I understand that competition among systems to show up most favorably in searches, etc. requires that the entire account equity be used to make the most dollars).
What then if you get 10 subscribers each with $10,000 accounts? Then all of them will suffer significant slippage for trying to get into the same stock/price. If your system is really more designed for someone with a $10,000 account, are you limiting the number of subscribers? If you allow 10 subscribers, this is just as good as one with a $100,000 account.
Are you trading the system yourself? If I developed a system which can only practically be traded with a $10,000 acccount, I would not open it up for everyone else to trade it as well and still continue trading it myself.
Chris
