PDA

View Full Version : Fooled by Randomness


Matt Bowen
10-13-2005, 12:23 AM
Trading Tip (From the Van Tharp's Weekly Newsletter)

On Randomness and Streaks

By

Van K. Tharp, Ph.D.

One of my favorite books is Fooled by Randomness by Nicholas Taleb

http://www.amazon.com/gp/product/1587990717/104-4349839-6925558?v=glance&n=283155&s=books&v=glance

I’d suggest that you purchase it and read it cover to cover several times. The major theme of the book is that people totally fail to understand randomness – even logical, educated people. In fact, sometimes logical, educated people can be fooled even more than the non-logical, and uneducated.

For example, last week in an article by Chuck Branscomb, you learned that good traders learn to tolerate long losing streaks, including making 12 losses in a row. We got several comments on that one.

Here is one of them:

You say that a winner still knows that he's a winner even after an expected 12 losses in a row. Who in his/her right mind would "expect" 12 losses in a row?

If a system is designed to win 50% of the time, the chances are only 2 in 10,000 of getting such a result if the system is performing as expected.

Even if the system is designed to win only one out of ten (and presumably make more than enough on the one win to make up for the nine losses), the probability of 12 losses in a row is still less than 50/50 at only 28%. Getting a result like 12 losses in a row would more than likely mean that the system is not working as intended.

I understand the intent of your statement, but show me a person who "knows" that he/she is a winner with a system that loses 12 in a row, and I would like to see that system ... so that I can take the other side of the trades.

My guess is that the author of this letter is very intelligent and understands quite a bit about probability. For example, he was able to illustrate that the odds of 12 losses in a row with a 50% system are only 2 in 10,000. He’s a little off — it’s actually 0.000244.

However, his calculations were based upon making 12 losses in a row in 12 trades with a 50% system. What if you make 100 trades each year? Or what if you are a short-term trader and make 1000 trades each year? Then the chances of a long losing streak are quite large.

For example, I did 20,000 Monte Carlo simulations of 100 trades with systems that are 25% correct; one that is 50% correct, and one that is 75% correct just to see what the losing streaks might be. What I found is shown in the table below:

1.)The first column is the win percentage of the system.
2.)The second column shows the length of the losing streak that will occur100%of the time in this system. That is, you are guaranteed of having a losing streak that long.

3,)The third column shows the average losing streak for that winning percentage. You have a 50% probability of getting a losing streak this long.

4).The fourth column shows a 10% probability losing streak. Losing streaks will happen this long 10% of the time if you make 100 trades each year.

5.)The next to last column shows a 1% probability losing steak. In other words losing streaks this long will happen 1% of the time if you make 100 trades each year.
6.)And the last column shows the maximum losing streak in 20,000 simulations of 100 trades.
Losing Streaks As A Function of Winning Percentage of Your System

See Chart Attachment


Notice that with a 50% system, you are almost guaranteed to have five losses in a row in 100 trades and you’ll probably get six losses in a row. If you simply calculated the probability of getting six losses in a row, you’d say it’s unlikely because the probability is 0.0156. You’d conclude that its nearly impossible.

But it’s not impossible in 100 trades. In fact, it’s almost certain.

In our simulation, we also found that you have a 10% chance of nine losses in a row and a 1% chance of 12 losses in a row. One percent is unlikely, but not impossible. And just when you decide it’s impossible, it would probably happen for you.

I don’t have a simulator that will easily do 20,000 simulations of 1,000 trades, which is still quite likely for a short term trader to make in a year. I will easily have over 1,000 trades this year and I’m not a day trader. Many of my clients will easily have 1,000 trades each year. I’m right on about 45% of my trades and I probably have a losing streak as big as 15-20 losses in a row this year. Some of my students (who are superb traders) frequently report losing streaks as long as 20 in a row.

Most long-term trend following systems are not right 50% of the time. They are more likely to be right 30-45% of the time. Thus, the probability of having 12 losses is a row in 100 trades is no longer just a possibility – it’s a distinct probability.

So what does this all mean for you?

It simply illustrates the point that the average person does not understand randomness, even the average highly intelligent person.

Second, it says that long losing streaks are quite possible. Most people who insist on being right will typically give up their system, thinking it is no good. In reality, the system is doing what you probably should expect.

And, lastly it shows the critical importance of position sizing. If you risk 1% on each trade, then after 12 straight losses you’d probably be down about 10% (i.e., you’d be down less than 12% because you’d only be risking 1% of what’s left after each loss).

Your next trade might be a 20R winner and you’d be up — even after 20 losses in a row.

Don't worry, only 2 out of every 100 traders will pick up on this information, the other 98 will be paying for your winning trades the rest of the year :D

Matt Bowen
10-13-2005, 12:35 AM
Here is another chart I have used to show people what the probability is of having at least X consecutive losing trades within a 50 period sample size of trades.

The column on the left shows you the winning probability of the system

The column on the Top shows you the number of consecutive losing trades.

In other words, if you took 50 trades from trading system here is a good idea of what's going to happen...this is nothing new folks, it's called drawdown (and they do happen).

hanover
03-16-2006, 04:03 PM
Assuming that you are quoting Dr Tharp correctly, he is wrong.

"2.)The second column shows the length of the losing streak that will occur 100% of the time in this system. That is, you are guaranteed of having a losing streak that long."

Absolutely nothing is guaranteed. His table states that (for example) if a system has a 50% win rate, the probability of 5 successive losses in 100 trades is 100%. That's the equivalent of saying that if I toss a coin 100 times, I'm guaranteed a streak of at least 5 heads. The probability of obtaining such a streak may be extremely high, but it is not guaranteed. In a randomly generated event, anything can happen. Randomness is, by definition, random.

David Louisson
Hamilton, New Zealand

mrkam
03-16-2006, 04:34 PM
lets' not lose sight of the overall point here. We are talking about the mentality of the trader more than the minutia of statistics....The WINNER/SUCCESSFUL trader is the one that is PREPARED for such an (even highly probable) "eventuality".

So, I for one would rather count that 5 (or 12!) loss streak as a CERTAINTY, and be postured to sustain it to live to trade another day, than not. So if it does not happen (granted, like you say, it is just a probability), then I'm even happier....

hanover
03-19-2006, 02:28 AM
Point taken (I don't disagree with you), but IMHO using words like "guaranteed" is, especially to the statistically uninitiated, misleading.

IMHO it's also dangerous to assume that a prior win rate will be sustained into the future, even if one's system has been tested historically over an extremely large sample size that somehow purports to cover "all" market conditions. Given that the probabilities of future price movements can not be calculated with any real accuracy (all price action contains elements of a random walk), there is always the possibility that pretty much anything can happen, and that it can - and will - distort win rate. Hence I believe that the above tables, however valid mathematically, lose some of their usefulness when applied to the "probabilities" that underlie the markets. In other words, you could have a system that has generated 900 winners from 1,000 trades in testing, and then later quite feasibly attain a streak of 20 losses. That is not because the math is incorrect, or that you have been incredibly "unlucky"; it is because the forever "changing market conditions" betray a mathematical model.

The disclaimer along the lines of "past peformance offers no GUARANTEE (there's THAT word again!) of future success" is very true - on so many levels!

I have tried to apply statistics to many aspects of (mainly short term) trading, and while they can occasionally provide some kind of rough guide, in my experience there are frequently "inexplicable surprises" in store, especially for the uninitiated. I am being driven increasingly to the conclusion that trading successfully is an art rather than a science. IMHO tables like those above are of limited value, and create the impression that the markets can be conquered by precise statistical analysis. This misconception is further compounded when words like "guaranteed" are used.

David

Matt Bowen
03-19-2006, 12:49 PM
Hi guys,

I thought I'd throw my 2 cents in here...

First, of all Van Tharp is talking about a "specific system" (set of rules that has an expectancy level).

Second, he's talking about a guaranteed loss (meaning that no matter how you slice it, the end result
is a loss). He's not talking about winning anything, but rather a sure-thing of loss.

I don't know maybe we both need to go look up the definition of "Guarantee":


--------------------------------------------------------------------------------------------------------------------

guarantee
One entry found for guarantee.


Main Entry: 2guarantee
Function: transitive verb
Inflected Form(s): -teed; -tee·ing
1 : to undertake to answer for the debt, default, or miscarriage of
2 : to engage for the existence, permanence, or nature of : undertake to do or secure <guarantee the winning of three tricks>
3 : to give security to
4 : to assert confidently

--------------------------------------------------------------------------------------------------------------------


Let me sum this up real easy... the market is a probability distribution... The casino's in Las Vegas (and all of the world for that matter) are built on a 2% house advantage. All because of what? Games of chance, people want action. You could stand in front of a casino with a huge sign saying: "You are going to lose all of your money in here" and what do you think people are going to do? I guarantee you they will walk right past you. Why? because they don't want to hear it... people only hear what they want to hear, you and I have no control over that. Nor do we have any control over crowd psychology. The only thing you can control in this business is the EXIT. I agree with you the statistics only go so far, they are nothing more than a guidelines. Too many people want to make trading black and white, and expect the performance numbers to be the same on future results, that's not real, because the market is made up of thousands of traders all over the world. How many days do you see the same market action? In 22 years of trading I've never seen two days repeat the same. Statistics are like a tire gauge...on any given day the pressure could be more or less. Thinking in terms of stats will only lead to frustration and a lot aggravation...because trading action does not repeat exactly like it did in the past. The market is going to do what it's going to do, with or without you. Kelly Angle said it best: "I think there are a substantial number of people who literally are professional hunters for the holy grail in trading. They don't want - or can't - acknowledge the reality, and instead want to believe in the fantasy. They want to believe there is a magic system that will rarely have loss, only catch the big moves and require no time or capital to maintain"...this is probably why they end up going to a seminar to learn how to trade options, ha ha. (Again, people are going to go after WHAT THEY WANT, not necessarily what they need) they are after the illusion of safety.

mrkam
03-19-2006, 10:35 PM
One of the better threads, IMHO.... All good. No dis.... all good.

I think we understand ourselves... and thanks for the book recommendation, Matt. I have bought all I have seen recommended here, and would say that they have helped me alot....


mk

Vernon
03-25-2006, 01:45 PM
Hi, Steve!
1. Does the Scanner find also the Advanced ABC setups ? Or maybe the Advanced ABC module is a manual one?
2. It looks like Advanced ABC module is something different from the classics TS1,TS2,TS3 setups , in that the first "leg" of this pattern starts in the middle of "nowhere". So that : A quite new MTP "pattern"???
Regards,
Vernon

Steve Griffiths
03-26-2006, 10:12 AM
Hi Vernon,

Why do you keep on posting the same question in different threads on this board, especially when you have already asked this question before and I have alraedy answered it a while ago ?

Does not make sence to me.........

Steve