The discussion of the Wikipedia article about the EMH is interesting too. One discussant suggests that the EMH should not be considered mainstream anymore. Another discussant claims that it is proven that it even never existed (!). Another discussant comments to this
"Even Fama has of late recognized certain shortcomings in the EMH. Nonetheless, we teach it because it helps us understand how the market works."
Guess I havn’t hit zen master status yet, But I’ll keep meditating.
You seem to have the understanding of one.
"remember it can be a difficult task to seperate good tea from bad."
If only you were going to read this, it would have ended with the first statment.
> Guess I havn’t hit zen master status yet, But I’ll keep meditating.
You seem to have the understanding of one.
Not at all. I only quote one and seek to understand by keeping my
tea cup at least half empty…
but I didn’t stay in a Holiday Inn
express last night.
http://www.google.com/tools/firefox/toolbar/FT3/intl/en/index.html
Strongly recommended by a really bad speller (me):
“SpellCheck
ARe yu a raelly bad tyipsst? Google Toolbar’s new SpellCheck button finds any spelling mistakes whenever you type into a web form, including web-based email, discussion forums, and even intranet web applications. The AutoFix option even corrects all of your text with a single click.”
Sam I have google on my tool bar but can’t figure how to access the spell check to activate. How would I do that? Thanks
> Sam I have google on my tool bar but can’t figure how to access the spell check to activate. How would I do that? Thanks
Please call Google tech support ;-).
Best of luck!
>This is my last post addressing the multi-personality syndromatic in this thread, as he has a continuing need to get in the last word, even when he has nothing to say…
I get it. You are obviously saying nothing…
>That is decided by a judge or jury, not by asylum escapees…
I get it. You escaped from an asylum. Please leave C2 and go back to it forever…
>Frank, it would help if you read the posts instead of skimming them.
The kettle calling the pot black…
>There is no need to argue whether they were experts or masters, or to introduce your opinions about whether experts should be ignored a la Palsun, and whether there was an argument about who appoints them.
I get it. Mr. Ross is the accepted Master Trader. Who accepted him? Blank out. Who appointed him? Again a Blank out.
>I would say that if you and I stood in front of 100 accepted masters of trading"
>Please don’t torture the thread with semantic trivia.
Ross, you are a inevitably tortured. To you failure means suffering - and so does success. Success of your kind (in blanking out reality) is a threat, attainment brings anxiety, desire is guilt, self-esteem is self-loathing, pleasure is laced with hangover, joy is overcome by pain. Whatever the name of such a state, it is not “happiness.”
MK, I think this thread can be closed as well. If anyone has something significant to add, he can open a new thread.
Jules, I thought about two additional explanations for your results. One is that the number of new systems added each month has increased considerably over the years (I think you mentioned this as well as a caveat). Thus, where a 3 year old system comes from a pool of 25, a 3 week old system comes from a pool of 200. Obviously the larger the pool, the larger the spread in Sharpe ratios you’ll find. Or in other words, if 200 systems were launched every month 3 years ago, you would observe a reasonable spread at the right-hand end of your graph, instead of just a point estimate of zero.
Second, it is possible that extremely good or bad systems will be terminated sooner than less extreme systems. E.g. vendors of extremely profitable systems will sell their system or themselves to a fund, or trade it themselves exclusively after having gathered enough capital from subscription fees. Vendors of extremely bad systems will quickly realize that they made some error when designing the system and understand they’ll never get any future subscribers with a 70% drawdown on record. Thus, what may appear as regression to the mean is partly just vendors terminating systems.
Another thing that worries me is that not all terminated systems are included, e.g. http://www.collective2.com/cgi-perl/systems.mpl?want=publicdetails&systemid=9302690
doesn’t seem to be included because it doesn’t show up in the grid, even though is does appear under “1-Click searches”. There seem to be many more systems like this.
Finally, it happens sometimes that after terminating a system, the vendor forgets to close any open trades, which means from that point on the systems will start to lead its own life (I guess that is true for the system under the link above).
Finally, a good explanation instead of the bizarre theories we were getting exposed to by self-appointed, self-accepted gurus of trading…
I don’t agree with the first point, at least not as far as it deviates from my previous caveat. I thought of this when I wrote the caveat. But a larger sample does not systematically increase the variance. If only the number increased, the variance should remain approximately stable. What I mentioned as caveat is that there can be a different selection mechanism, leading to more variation. That means that there should be relatively less average performing systems and relatively more extremely (either good or bad) performing systems. What counts for the variance are the relative frequencies, not the total frequency. That the relative frequencies changed is indeed a possibility, but it is not implied by a change of the total frequency, and therefore it is not so obvious as you seem to suggest.
Your second explanation is possible. To eliminate such selection effects, it would be needed to follow the same set of systems during their history, as I suggested in my second post too. Such a longitudinal study has also limitations, however. For one, you can do it only with the older systems, but then you might argue that these systems shared some special historic events in the forex market, that coincidentally happened at the beginning of their history, but that will not happen anymore. If you use only younger systems then you cannot see what will happen when they are old.
So there will always be alternative causal explanations. The basic reason is that the data do not come from a controlled, randomized experiment. You can try to control the most important disturbing variables, but it usually take years to accomplish such studies (e.g. see how long it took to prove that smoking causes cancer). I hope that you will forgive me that I won’t do that. My being too lazy for that does of course not disprove your explanation.
The main reason why I believe in the effect is that it is predictable with purely theoretical statistical reasoning. If trading involves a probablistic component then you will see that positive Sharpe ratios tend to decrease on average, although they do not need to go to 0. This theoretical expectation is what made me presume the existence of this effect in the RT Forex discussion, and I wrote it later more explicitely in this thread. The figure of the first post is no more than an interesting illustration of the possible size of the effect.
(I apologize if I wasn’t tactical somewhere, I’m in a hurry before my favorite bar closes ;-))
I must confss I haven’t read all of the posts in this thread simply because I don’t have to time for some of these long arguments. My apologies in advance if I make a point already discussed.
I think the answers may be less theoretical and more obvious. First there are abandoned ships. I have two systems OBSOLETE1 and 2 that have not been supported for some time. But I didn’t close the positions. Now these systems drift onwards. From all appearances they are active systems. They still indicate that you can sign up for such-and-such per month. Presumably the stats are still being crunched. How many other systems out there like this?
Second thing is that 95% of systems come onto P123 with obvious mis-use of leverage. And this may be a conservative number. Some will perform extremely well at the beginning and others will blow out immediately. But all of the over-leveraged systems will eventually suffer a major blow, subscribers flee, etc…
This explains why the Sharpe Ratio is so extreme and dies down with time as only the conservative money managers are left.
Steve
>This explains why the Sharpe Ratio is so extreme and dies down with time as only the conservative money managers are left.
I disagree with parts of that statement. I would correct it as follows: This explains why the Sharpe Ratio is so extreme and dies down with time as only the conservative/prudent/very aggressive systems are left in the long-run and the extremely aggressive systems are left out (in the long-run).
I think the answers may be less theoretical and more obvious.
I am not sure if you want to suggest that the theory gets too much emphasis, but note that the discussion started with my theory, before the plot existed, that high Sharpe ratios will show the tendency to decrease. It is logical that I give attention to that theory, because it predicted this pattern, and it was the reason to make the plot in the first place.
First there are abandoned ships.
Indeed that has been addressed already. But this does not explain why the profit factor exhibits the same pattern on the number of trades. The number of trades does not increase anymore for abandoned systems.
Second thing is that 95% of systems come onto P123 with obvious mis-use of leverage.
Leverage in itself does not affect the Sharpe. The profits will be x times larger, but the losses too, leaving the Sharpe the same.
But all of the over-leveraged systems will eventually suffer a major blow,…
This is not an alternative explanation; it is just another formulation of the initial theory. So I agree with it. BTW, the properly leveraged systems will also suffer a blow, but a minor one.
This explains why the Sharpe Ratio is so extreme and dies down with time as only the conservative money managers are left.
Your last argument is that only conservative money managers have a profitable system in the long run, and that only these survive. But conservative, profitable systems have a positive Sharpe. So your theory predicts that the average Sharpe should increase and tend to a number larger than 0. This is not what happens in the figure.
"Leverage in itself does not affect the Sharpe. The profits will be x times larger, but the losses too, leaving the Sharpe the same. "
Except that once the large blow occurs the system provider is forced to alter his / her strategy if continuing to trade.
Does profit factor only consider closed trades?
The long term providers (conservative money managers + Palsun) average out to zero.
The profit factor for Diversified Value dropped from over 5:1 to 3:1 last week. So the answer is that profit factor is affected by open trades.
Thus profit factor would also be affected by the abondoned ships theory.
Steve
>The long term providers (conservative money managers + Palsun) average out to zero.
Are you implying that I am not/cannot be a conservative money manager? I do have a conservative system, by the way…
Wrt. leverage: Yes, a blow out can do that, that is why I said "in itself". The blow out situation is part of the initial theory. So it is not another explanation.
Wrt. open trades: I know that the profit factor is affected by open trades, but my argument was that the number of trades increase anymore when you abandon the system. And the profit factor was plotted on the number of trades. In the graph such a system will drift vertically to the middle, but it will not drift to the right. So that does not explain why systems at the right side are vertically in the middle.
Wrt. the theory: To make the theory visible, I made the first plot again but now with the theoretical prediction in it. In this specific plot, I have assumed a simple model where the daily excess log return rates are independent and have a normal distribution with mean 0. The plot shows the area within which 90% of the Sharpe ratios should fall. See
http://www.xs4all.nl/~julese/Sharpe_forex_theory.pdf
As you can see, the theoretical prediction describes the observed values quite good. But this theoretical curve reflects only the effect of increasing statistical reliability of the estimates.
Correction: " number of trades increase anymore " -> number of trades does not increase anymore
PS: Actually, I take back my statement and re-correct it as follows: This explains why the Sharpe Ratio is so extreme and dies down with time as only the conservative/prudent/aggressive systems are left in the long-run and the speculative systems are left out (in the long-run); because even an aggressive system can be successful in the long-run if the Issue Selection/Market Timing/Money Management is good enough…
