my journal 2

[Yamato]

Replying as I read.

The problem with stoplosses is not only that they complicate things and have many negative implications (adding a stoploss increases the danger of over-optimization, as you're adding an extra parameter), but also that in back-testing they do not add profitability.

Yes, stoplosses would have some advantages: I would know how big my biggest loss could be. But the advantages of things being the way they are now are more, so I am not considering adding them. Also, with 120 systems created and a humongous infrastructure with 7 excel files, being burned out... it is simply not worth it, practically impossible. In this field, I have learned from the beginning that I just have to say "no" from the start, and not start doing something I cannot conclude.

The same applies to trying a different platform. Huge implications. It's like starting a new business. You have a restaurant, it took you years to get it going, and you're telling me to dismiss it and turn it into a swimming pool

Oh, good point. I read it only now: "I know that is a massive change and a lot of work with 120 systems though...".

I can't get started on this now. However I thank you for the detailed feedback and advice. But it's just a route I cannot get started on. I mean: we're talking about one year of extra work without any clear advantages (as I said there are more disadvantages than advantages).

 

[Yamato]

Mercedes Sosa - Gracias a La Vida - YouTube

http://en.wikipedia.org/wiki/Gracias_a_la_vida

"Gracias a la vida" (English: Thanks to life) is the name of a song composed and first performed by Chilean musician Violeta Parra, one of the artists who set the basis for the movement known as Nueva Canción. It was released in Las Últimas Composiciones (1966), the last album Parra published before committing suicide in 1967.

Gracias a la vida que me ha dado tanto.
Me dio dos luceros que, cuando los abro,
perfecto distingo lo negro del blanco,
y en el alto cielo su fondo estrellado,
y en las multitudes el hombre que yo amo.

Gracias a la vida que me ha dado tanto.
Me ha dado el oído que, en todo su ancho,
graba noche y día grillos y canarios,
martillos, turbinas, ladridos, chubascos,
y la voz tan tierna de mi bien amado.

Gracias a la vida que me ha dado tanto.
Me ha dado el sonido y el abecedario,
con él las palabras que pienso y declaro:
madre, amigo, hermano, y luz alumbrando
la ruta del alma del que estoy amando.

Gracias a la vida que me ha dado tanto.
Me ha dado la marcha de mis pies cansados;
con ellos anduve ciudades y charcos,
playas y desiertos, montañas y llanos,
y la casa tuya, tu calle y tu patio.

Gracias a la vida que me ha dado tanto.
Me dio el corazón que agita su marco
cuando miro el fruto del cerebro humano,
cuando miro el bueno tan lejos del malo,
cuando miro el fondo de tus ojos claros.

Gracias a la vida que me ha dado tanto.
Me ha dado la risa y me ha dado el llanto.
Así yo distingo dicha de quebranto,
los dos materiales que forman mi canto
y el canto de ustedes que es el mismo canto,
y el canto de todos, que es mi propio canto.

Gracias a la vida que me ha dado tanto.

http://www.cancioneros.com/nc/685/0/gracias-a-la-vida-violeta-parra

 

[pecas]

I would consider using the sum of drawdowns, instead of the combined drawdown, as a selection method (all relativized).
The sum of drawdowns well represents a worst case scenario, especially if you double that number.
It's not the most efficient way to make money given the drawdown of a combination of systems, but definitely reduces the risk to reach the uncle point again next time you start trading.

Another suggestion is grouping your systems based on the futures they're trading: equities, metals, energy, etc. I would prefer having GBP/USD and Gold enabled, more than GBP/USD and EUR/USD.

 

[pecas]

Oh, and I read Wiley: the handbook of portfolio mathematics.
Indeed, hard to read, a lot of math that becomes harder very quickly.
I'll try to watch the videos and read the links you posted, on Monday.

 

[Yamato]

Yes, this sounds safe, but if I only wanted to be safe, I would not have a problem. The problem is that I have to be safe and yet at the same time I need to step on the gas.

You see, if I add up all the drawdowns for the top 25 systems I have, there is no risk of reaching uncle point, but then the uncle point has to be set so high that no investors will ever accept it.

My 120 systems have an average max relativized drawdown of 9000. Multiply that by 25 and you get about 200k. Multiply that by 2, and you get an uncle point consisting of a loss of 400k, when in fact my investors were willing to only lose the 40k we had made.

Actually, with the number of contracts we've been trading (about 60), our uncle point would have had to be about 1 million dollar. The problem was not that they weren't willing to lose 1 million dollar, but that did not appraise correctly a potential drawdown for the future, and we didn't manage to do it, because we used max relativized combined drawdown, and, after optimizing it hundreds of times, we came up with an optimal combination (too optimal, because curve-fitted and over-optimized) that returned, only for the past, a very high profit with a very small max drawdown. It is a good concept to find systems that compensate each other, but not though guesstimates, optimizations and rule-of-thumb methods. There has to be a fixed formula that guarantees a balanced and constant way of achieving this, and that makes sure we're not just coming up with a lucky brute-force combination.

Now you see my problem. Your proposal cannot be applied, but it clarified where we stand: we have to find a way to at once minimize risk while maximizing profit.

Of course, if I just wanted to be safe, I'd allocate a potential drawdown of 1 million per system, and I'd be safe. Then I would make 1% per year and no one would be interested.

But keep suggesting things because it will help me focus on the problem and identify its elements. We're getting closer.

Regarding this suggestion, "Another suggestion is grouping your systems based on the futures they're trading: equities, metals, energy, etc. I would prefer having GBP/USD and Gold enabled, more than GBP/USD and EUR/USD", we need once again to translate it into a formula, so there's no risk for our momentary preference to negatively affect the outcome of our investment.

As you see, we have to come up with a formula that has a very large number of ingredients.

Regarding the videos i posted (there's a total of about 10 videos), they're pretty pleasant to watch and interesting, but only give you a taste of the issues ahead. The problem with his paper instead is that it's heavy for me, like anything wih formulas, but it is only 16 pages. I think it will be the best investment for me. Especially since I'll be able to read them at work.

 

[barjon]

travis

ok, I guess I shouldn't have reproduced your picture and drawn those damn lines on it because I'm not suggesting utilising trend line, MA crossovers or anything like that.

What I am suggesting is:

1. Your money management relates solely to your Master System (Combined Performance) and that such management is applied across the board of the supporting trading systems and not by selective tinkering with those individual trading systems. I think that squares with your own thinking?

2. It is the nature of your trading systems that the risk (R) in your Master System (Combined Performance) is unlimited (or, maybe in reality, to what capital you have left when the plug is pulled on the trading systems through lack of sufficient margin).

3. It is possible for you to set R at any level you want in your Master System (Combined Performance). If it was a single tradeable system (which it isn't, of course) this would be the stop-loss point that closed the trade. In your case it would be the point where the supporting trading systems were turned off and the completion of a trade cycle.

4. Should this be the case the trading systems would only be turned off for actual trading but would still run on a "paper" basis enabling a similar "paper" continuation of your Master System (Combined Performance). It is possible for you to set a level (R-x) where you would turn the systems on again for a fresh trade cycle,

5. Neither R nor (R-x) levels need be discretionary, but born from analysis of historic drawdowns and recoveries in your Master System (Combined Performance).

6. You might consider R to be a point where you seek to reduce further risk rather than a "stop-loss" point. However, further risk can only be reduced by reducing your capital in play and if this is not possible by reducing position size then I can't see any alternative but to tinker with the supporting trading systems.

jon

 

[Yamato]

Replying as I read.

Well, I didn't draw all my conclusions and elaborate a reply from just your picture, but also from the other things you wrote.

1. Given this usage of "master system" and how it confuses me, I am not sure we even agree on this. So I will define the details better for you and for my own benefit. My Master System is not enough and it is not self-sufficient, if "master system" is defined as a synonym for "combined performance", as you seem to use it. As written before, i want to develop a formula that takes into account both the combined performance and the individual performance of systems. Systems are enabled as soon as they have a sharpe ratio of 2 or more, and yet their dosing (how many contracts) depends on the Combined Performance, and it also depends on the margin available and probably on something else as well, such as the maximum loss. I think we should stop using the term "master system", because it confuses me as to what it means. If we can find other words for it, it's better. If it means "combined performance" let us use that term, which is clearer. I don't see why we should call "combined performance" with a new name, when the original name is clearer.

2. Too complex phrasing. You need to simplify things or I will not understand you. I don't have to grade your thesis and we don't have to impress anyone. I don't mean to be critical but here we will end up like with bbmac, in a fight, because he was writing theses like you. I am not a professor looking for any essays to grade. Anyway, whether it's because English is not my first language or because you're writing a thesis, but I really do not understand.

3. With this formula-like sentence you send me again into a short-circuit. You want to show me I am not good with formulas? Ok, you're better at them than me. But if we're talking about ice cream what is the point of translating such a sentence into a formula? Damn, this sentence is just as hard as bbmac's. You guys are competing at short-circuiting me. When I talked about the scientific method, I didn't mean every sentence has to look like a formula, but only that there can be nothing that is not univocal, so let's make 3 formulas, simple ones, and follow them. I am not saying every word we utter has to be uttered as a formula.

4. This is another needlessly overcomplex sentence. Hey, I am writing a lot of stuff, but I don't think it is necessary to turn every sentence into a maze. So, once again, I am into a short-circuit because not just every sentence is a formula, but it's referring to previous sentences and the post is now a single formula, as in a treasure hunt, where the elements from one discovery lead to the next step. So, since I didn't understand the previous sentences, I am now into the dark as far as this new one as well.

5. Still do not understand, and I am getting more and more lost.

6. I can't make sense out of this either. I know it must be as frustrating for you as it is for me, but I am not doing it on purpose. We have to keep it simple. Pretend you're talking to a child, like a 10 year old.

The point is I am not an academic. When I said let's make things in a scientific way, I didn't mean we have to talk like two professors. I just meant let's ban all discretionary action. Other than that, let's keep everything as simple as possible. I am prejudiced against needless complexity and formulas, so you have to be patient with me. I have like a phobia for academic language. At once I am for science and against academics (**** school).

 

[counter_violent]

Following on from todays discussions ( and i'm really going to try and keep this simple )
The drawdown /equity chart posted is of the combined systems in play.

Would it be possible to break this chart down into individual instruments and their, real time performance?
If this is possible then it may show up the levels where the collective systems start to fail ( in terms of drawdown ) for example, if 70% ( fig plucked from the air ) of systems go into drawdown condition then place a direct opposing hedge on each system in drawdown until the systems recover / normalize. Rinse and repeat as required.

The levels for automated intervention could only be determined after some analysis of the individual and collective performance.

 

[Yamato]

Replying as I read.

Yes, the drawdown/equity chart posted is of the combined systems in play at each specific period. As i said before, that equity line merely represents profit and it represents, depending on the period, profit achieved via different systems and via different required margin. Given those changing elements, you can only draw so many conclusions by looking at the equity line. But it is still a perfectly correct representation of profit. So that is I am keeping it that way.

Yes, of course it would be possible to break that chart into individual systems. But I can't post it, because we have traded 60 different systems during the last year and a half. Second of all, it would not be very useful because we did not trade throughout the period (as I said, we changed them). If I know the point of this, I can come to some compromise. Otherwise I cannot spend several hours to post 60 different charts.

Yeah, but this suggestion is far from being univocal and it's extremely complex, and it cannot possibly be back-tested due to its complexity. It was a simple post, but suggesting a nearly impossible task, if intended in a univocal way. If it's intended as "modify the balance of systems here and there as you go along" then I cannot do it, because it would imply not being automated. Also, it's hard to automate things based on how they look on a chart, so we would still need to come up with a formula.

I think you are very far from understanding my problem, because you're in a different situation, like almost everyone else who's been reading this journal, and who's tried to help me.

Thanks for the effort, but, understandably, there is going to be only one person out of 200 who understands exactly where I am at and what I need. I am lucky enough that those like you who don't understand what i am doing don't insult me, and tell me stuff like "you are beating about the bush" and "if you don't want to hear my questions just say so". And I am even luckier if there's people who read at least the last month of posts before asking questions.

 

[barjon]

Replying as I read.

Well, I didn't draw all my conclusions and elaborate a reply from just your picture, but also from the other things you wrote.

1. Given this usage of "master system" and how it confuses me, I am not sure we even agree on this. So I will define the details better for you and for my own benefit. My Master System is not enough and it is not self-sufficient, if "master system" is defined as a synonym for "combined performance", as you seem to use it. As written before, i want to develop a formula that takes into account both the combined performance and the individual performance of systems. Systems are enabled as soon as they have a sharpe ratio of 2 or more, and yet their dosing (how many contracts) depends on the Combined Performance, and it also depends on the margin available and probably on something else as well, such as the maximum loss. I think we should stop using the term "master system", because it confuses me as to what it means. If we can find other words for it, it's better. If it means "combined performance" let us use that term, which is clearer. I don't see why we should call "combined performance" with a new name, when the original name is clearer.

As you said earlier, name is unimportant. Just trying to confirm it as central and not merely the sum of the parts.

2. Too complex phrasing. You need to simplify things or I will not understand you. I don't have to grade your thesis and we don't have to impress anyone. I don't mean to be critical but here we will end up like with bbmac, in a fight, because he was writing theses like you. I am not a professor looking for any essays to grade. Anyway, whether it's because English is not my first language or because you're writing a thesis, but I really do not understand.

Your systems do not have stop-losses, therefore the combined possible loss is unlimited. Albeit that you have a calculated expectation of the worst drawdown.

3. With this formula-like sentence you send me again into a short-circuit. You want to show me I am not good with formulas? Ok, you're better at them than me. But if we're talking about ice cream what is the point of translating such a sentence into a formula? Damn, this sentence is just as hard as bbmac's. You guys are competing at short-circuiting me. When I talked about the scientific method, I didn't mean every sentence has to look like a formula, but only that there can be nothing that is not univocal, so let's make 3 formulas, simple ones, and follow them. I am not saying every word we utter has to be uttered as a formula.

You can set the maximum drawdown you are prepared to stand before calling uncle and you can set that at whatever level you like

4. This is another needlessly overcomplex sentence. Hey, I am writing a lot of stuff, but I don't think it is necessary to turn every sentence into a maze. So, once again, I am into a short-circuit because not just every sentence is a formula, but it's referring to previous sentences and the post is now a single formula, as in a treasure hunt, where the elements from one discovery lead to the next step. So, since I didn't understand the previous sentences, I am now into the dark as far as this new one as well.


If you're going down a turn off, turn on route you will need to know how the combined systems perform after they are turned off in order to know when to turn them on again.


5. Still do not understand, and I am getting more and more lost.

Since you say you are against discretionary intervention I was saying that the turn off, turn on levels can be established by some formula based on historical evidence of past combined performance.

6. I can't make sense out of this either. I know it must be as frustrating for you as it is for me, but I am not doing it on purpose. We have to keep it simple. Pretend you're talking to a child, like a 10 year old.

Since you seem against turn off, turn on I was saying that you could instead protect yourself against a similar degree of continued drawdown by reducing your exposure - although I can't see how this can be done without reducing position size. That only leaves tinkering with the individual systems in some way, which you are against I think/COLOR]

The point is I am not an academic. When I said let's make things in a scientific way, I didn't mean we have to talk like two professors. I just meant let's ban all discretionary action. Other than that, let's keep everything as simple as possible. I am prejudiced against needless complexity and formulas, so you have to be patient with me. I have like a phobia for academic language. At once I am for science and against academics (**** school).

Travis

Sorry I was not trying to be academic - just trying to get away from charts and lines and tried to put it in different terms - I've tried to explain better above.

Jon

 

[Yamato]

Sorry if I sounded pedantic in recommending simplicity. You did give me simplicity in the following post and i replied, but this might not get us anywhere.

You see, the problem here is that no one can understand me, because few of you are fully automated (the only one I met so far is pecas, and he is not fully automated). Probably the closest one to understanding my money management problem has been pecas and bbmac, but he was criticizing me too much so I had to temporarily ban him. I can't accept too much criticism, due to the over-critical father i've had.

Thanks everyone for your efforts to help me, even the banned ones.

Ok, now I will reply to your red lines within my post (I hadn't noticed them at the start).

Replying as I read.

As you said earlier, name is unimportant. Just trying to confirm it as central and not merely the sum of the parts.

Yes, it is not just the sum, but also its parts matter, taken individually.

Your systems do not have stop-losses, therefore the combined possible loss is unlimited. Albeit that you have a calculated expectation of the worst drawdown.

Yes and no. Or rather: "no". Look, if I have time exits, and the longest a trade lasts is 48 hours, the losses can only run for 48 hours. Then there will be a stop to them. I don't know how much they could lose in those 48 hours, true, but the loss in not unlimited. Furthermore, with a limited number of contract, the loss can only go as far as the contract goes. For example, with one CL contract, even if CL went from 70 to zero, I would only lose 70 thousand dollars, which is a lot, but it is not unlimited. So I would definitely not say that they are "unlimited", but rather that the max loss is "unknown". But you can still get a very good estimate from the past.

You can set the maximum drawdown you are prepared to stand before calling uncle and you can set that at whatever level you like

Given that your sentence by itself doesn't make sense to me, I looked at the original one:

3. It is possible for you to set R at any level you want in your Master System (Combined Performance). If it was a single tradeable system (which it isn't, of course) this would be the stop-loss point that closed the trade. In your case it would be the point where the supporting trading systems were turned off and the completion of a trade cycle.

In part I don't understand, and I don't think I would agree, from the small part I can understand. I will try to say something regarding this subject, even though I don't understand your sentence. I will not use stoplosses: I will not add them to my systems. I do not understand the concept of "trade cycle"

If you're going down a turn off, turn on route you will need to know how the combined systems perform after they are turned off in order to know when to turn them on again.

Yes, I forward-test my systems, which means I record the trades exactly as if I were trading them with real money, but they're only "paper trades". So I can resume trading. Having said this, I really do not contemplate the possibility of turning them off, except if the individual ones have failed or if we don't have enough money to run them anymore.

As i said before, turning a system off means that after several months, it has gone below a sharpe ratio of 1. Turning all of them off only means that there has been a failure in our money management. It is not planned in a regular situation to turn all systems off: if it happens it should mean there's a general failure.

Since you say you are against discretionary intervention I was saying that the turn off, turn on levels can be established by some formula based on historical evidence of past combined performance.

Yes, I am against discretionary intervention, in that I am for total automation, so we know what we will do at all times and we avoid discretionary mistakes. Yes, good point. I need a formula that will decide how much combined drawdown we will accept and expect. Having said this, I would not call this "turn on, turn off" levels, but just "turn off levels" or even better "failure point" and even better "uncle point". Because, as i said, it's not like I devise as part of my strategy to turn them on and off all the time. If I ever turn all systems off, it means we have failed (or "i have failed" if I am investing my own capital). Part of my studying should help me understand how likely it is that I will fail. I want to be able to choose uncle points so that I will never reach them, but not so unlikely to fail as to trade systems that don't make any money (relative to the capital invested).

Since you seem against turn off, turn on I was saying that you could instead protect yourself against a similar degree of continued drawdown by reducing your exposure - although I can't see how this can be done without reducing position size. That only leaves tinkering with the individual systems in some way, which you are against I think

I am absolutely against "turn off/turn on" and I am hoping to never see again a situation where I have to turn them off (all at once), but it's ok to turn one off if it has failed, or gone below a given performance. As I said before, I would be ok with a plan to enable all systems above sharpe ratio 2, and disable a system if it falls below 1. My estimate is that this will happen to one system every 3 months. If not, then i might tweak the value (as part of a formula, not at will).

Correct. There is no way I will fight the drawdown by tampering with my systems in terms of contracts traded (all the way to disabling them). The drawdown has to be taken entirely. As I said before, here:
http://www.trade2win.com/boards/trading-journals/85510-my-journal-2-a-314.html#post1690404

1) with futures it is not possible, because the contracts are too large to be able to reduce bet size with the limited capital size i have. I either trade or not trade. I can't possible reduce bet size, given that i am usually trading 1 contracts.

2) The other problem is pointed out by the kelly criterion wikipedia entry:

Quote:
Kelly assumes sequential bets that are independent (later work generalizes to bets that have sufficient independence). That may be a good model for some gambling games, but generally does not apply in investing and other forms of risk-taking.

In other words, if we lose money today, it is not like rolling a die, but it may come in a specific sequence, so that we might hurt our investment by betting less the second time. Let's picture a case where you lose today and make money tomorrow for a few times. You would not end up even but broke, because you lose 1 today, make 0.5 tomorrow (after reducing bet size), then you double bet size again and lose 1 again, but then you reduce it, and gain 0.5 the next day... you get my point.

So, all things considered, I do not believe in the approach of reducing bet size, nor in monitoring the trend lines of the equity line, nor the moving averages crossovers in the equity line (and disabling the systems when there's a crossover).

I believe we must keep trading constant throughout the period. And most of all, I believe so for the reason that we have backtests telling us the entity of potential drawdown, and all these back-tests would be worthless if we altered the investment during the drawdown.

So I have listed 3 reasons to not do this: the first that I can't do it, and the other two that i don't believe in it and it's not scientific.

 

[Yamato]

Portfolio selection

THE PROCESS OF SELECTING a portfolio may be divided into two stages.
The first stage starts with observation and experience and ends with
beliefs about the future performances of available securities.

DONE: I HAVE BELIEFS ABOUT THE FUTURE PERFORMANCE OF MY SYSTEMS

The second stage starts with the relevant beliefs about future performances
and ends with the choice of portfolio.

THIS IS WHAT I WANT

This paper is concerned with the
second stage. We first consider the rule that the investor does (or should)
maximize discounted expected, or anticipated, returns. This rule is rejected
both as a hypothesis to explain, and as a maximum to guide investment
behavior. We next consider the rule that the investor does (or
should) consider expected return a desirable thing and variance of return
an undesirable thing.

YES

This rule has many sound points, both as a
maxim for, and hypothesis about, investment behavior. We illustrate
geometrically relations between beliefs and choice of portfolio according
to the "expected returns-variance of returns" rule.

YES, I NEED IT

One type of rule concerning choice of portfolio is that the investor
does (or should) maximize the discounted (or capitalized) value of
future returns.l Since the future is not known with certainty, it must
be "expected" or "anticipatded' returns which we discount.

CLEAR

Variations of this type of rule can be suggested. Following Hicks, we could let
"anticipated" returns include an allowance for risk.

NOT CLEAR

Or, we could let
the rate at which we capitalize the returns from particular securities
vary with risk.

NOT CLEAR

The hypothesis (or maxim) that the investor does (or should)
maximize discounted return must be rejected.

NOT CLEAR

If we ignore market imperfections
the foregoing rule never implies that there is a diversified
portfolio which is preferable to all non-diversified portfolios.

NOT CLEAR

Diversification
is both observed and sensible; a rule of behavior which does
not imply the superiority of diversification must be rejected both as a
hypothesis and as a maxim.

CLEAR

Ok, I read the first page.

 

[barjon]

[QUOTE}....Yes, good point. I need a formula that will decide how much combined drawdown we will accept and expect. Having said this, I would not call this "turn on, turn off" levels, but just "turn off levels" or even better "failure point" and even better "uncle point"....

Travis

I think I have been much more of a hindrance than a help to you, but just one final comment relating to what you say above.

At the moment I assume your systems are turned off so far as real money is concerned and that you will turn them on again when you have the capital. For want of a better phrase I describe the initial turn-on to the turn-off as a trade cycle. When you turn them on again it will be the start of a new cycle.

To protect from total drawdown (failure as you describe it) you could have a formula that sets your "uncle" point much higher and another formula that tells you when to start afresh. A simplistic example with a simple trade would be a stop-loss close when 25% of capital is lost and a re-entry when estimated capital (ie: where capital would have got to assuming no stop-loss close) rises by 25% from its low point.

Anyway I think I'd better leave you in peace

take care

jon

 

[pecas]

I would like to stay on the problem of determining the max combined drawdown of a combination of systems, because I think this is necessary to go further on.
I mean determining it with a formula, not combining past results, as this could lead to luck/curve-fitting, like probably happened to you.

Let's assume the systems are independent, not correlated. If this is the case, you can have any drawdown, at any moment, which is between 0 and the sum of max drawdowns.

Using the sum of max drawdowns as worst case scenario is not efficient and not an option, like you said.

But comparing the series of "current drawdown" of any system, you should come up with a statistical interval of confidence of what the combined drawdown could be. Don't know if my english is clear, but assuming you have two systems with max drawdown of 4000 and 5000, what I mean is you have a 100% confidence that max combined drawdown is <=9000... probably, comparing the two drawdown series, you come out finding that you have 99% confidence that the max combined drawdown will stay under 6000; 95% that it stays under 4500... and so on.

How you come up with this result from a math standpoint, is one thing I still have to investigate, and I'm not sure to have the ability to do this, but for sure it can be done... if math is too hard because series are too long, it can still be done with a Monte Carlo simulation.

If we can solve this problem you will also have the chance to say for sure if your latest combo was actually curve-fitting.

 

[Yamato]

Travis

I think I have been much more of a hindrance than a help to you, but just one final comment relating to what you say above.

At the moment I assume your systems are turned off so far as real money is concerned and that you will turn them on again when you have the capital. For want of a better phrase I describe the initial turn-on to the turn-off as a trade cycle. When you turn them on again it will be the start of a new cycle.

To protect from total drawdown (failure as you describe it) you could have a formula that sets your "uncle" point much higher and another formula that tells you when to start afresh. A simplistic example with a simple trade would be a stop-loss close when 25% of capital is lost and a re-entry when estimated capital (ie: where capital would have got to assuming no stop-loss close) rises by 25% from its low point.

Anyway I think I'd better leave you in peace

take care

jon

Replying as I read.

Well, except for the company you give me and the caring you guys show me, anyone who is not going directly in the same direction I am going will indeed distract me from my objective. It's inevitable. In this case, more than ever before, I find that only one person out of 100 might be going in the same direction as I want to go. And all the others are giving me the wrong directions, because they don't know where I am to begin with (only bbmac and pecas do, so far). But even questions from people trying to understand where I am are useful, provided they don't tell me "i am beating about the bush", while I am wasting my time trying to tell them where I am, and after they did not spend time reading my previous posts.

That is why I've been answering dozens of posts where people don't even know where I am, let alone telling where I should go next. Because it clarifies my mind. I should be able to answer any questions and criticism (e.g. bbmac's posts), if my mind is clear. If I don't, then it means I should reason more, and your questions help me. But of course there's also a limit to the amount of questions and criticism that can help me. Beyond a certain extent, it becomes counter-productive.

For example, as i said, bbmac gave me useful advice, but then he just as much distracted me and demotivated me and demoralized me, by subtly yet constantly insulting my systems and my work. Others were so superficial that I felt insulted by the simple fact they were posting anything. It's complex.

I want the feedback, but in this case more than ever, I need to keep going in one direction, whether right or wrong, I need to get there, to find out if I was right or wrong, and it's hard to keep going in that direction, if there's people on the side of the track telling me "no, wait: go here" and "no, wait: go there" and "where are you coming from?" and "where do you have to go?". If there's one thing I do not when I am tourist in a city is ask for directions from people in the street. You're better off using a map. This here is more or less as if I were asking for directions, except I didn't even ask.

It's as if you walked down a street with a map in your hand, and everyone stopped you and asked you "where do you have to go? because you know, you might be going the wrong way...". But this is implicit in the nature of a journal being written here vs. in my own computer, with no one reading. So, I should really say "thank you for the feedback" rather than "don't give me any directions: I know what i am doing already", which is what I sound like.

Regarding the rest of your comments, here's my answers.

Yes, my systems are "turned off" so to speak (in terms of trading real money, because of course I am forward-testing them).

Yes, I will turn them on again not only when I have capital but when I'll have this whole thing figured out. I need a mathematician really. Even though I just told someone else who said he needed a programmer to code his systems to become a programmer himself, it is not likely I can make myself become a mathematician and solve this problem. The second page of markowitz's paper is full of formulas.

Oh ok, now I know what you meant by "trade cycle".

Well, yes, of course, as I said to pecas, it's easy to find formulas to set uncle point higher, even to a million, and then I am also confident that not only will we not exceed it, but that my systems will prove to be profitable, as they have in the past.

The problem is that I want much more than that. I want a formula that optimizes profit relative to risk, and I want to know how likely, according to different combinations, I will be to reach a given level of drawdown if I trade a given portfolio for a given number of years.

No, sorry but your example once again is implicitly bringing up some sort of average crossover on the equity line ("exit when" and "re-enter if"). There is no re-entry, as i said several times. There is only uncle point (with myself or with others) and there is failure and there is having to do things all over again, because you might have done something wrong. That is why I said I would not call it "turn on / turn off" but just "uncle point". We must not find a level to turn the systems off and then back on, but a level which is so unlikely to be exceeded that it will not happen during your lifetime. And if it does, it will mean that it is very likely that you did something wrong, like we did this last time. Furthermore, I am expecting any investors to disappear if we reach such a level. So by all means, in any "uncle point" concept there can be no "re-entry". To be clear, "uncle point" means "it's all over, we ****ed up" rather than "let's stop investing for a while".

 

[Yamato]

I would like to stay on the problem of determining the max combined drawdown of a combination of systems, because I think this is necessary to go further on.
I mean determining it with a formula, not combining past results, as this could lead to luck/curve-fitting, like probably happened to you.

Let's assume the systems are independent, not correlated. If this is the case, you can have any drawdown, at any moment, which is between 0 and the sum of max drawdowns.

Using the sum of max drawdowns as worst case scenario is not efficient and not an option, like you said.

But comparing the series of "current drawdown" of any system, you should come up with a statistical interval of confidence of what the combined drawdown could be. Don't know if my english is clear, but assuming you have two systems with max drawdown of 4000 and 5000, what I mean is you have a 100% confidence that max combined drawdown is <=9000... probably, comparing the two drawdown series, you come out finding that you have 99% confidence that the max combined drawdown will stay under 6000; 95% that it stays under 4500... and so on.

How you come up with this result from a math standpoint, is one thing I still have to investigate, and I'm not sure to have the ability to do this, but for sure it can be done... if math is too hard because series are too long, it can still be done with a Monte Carlo simulation.

If we can solve this problem you will also have the chance to say for sure if your latest combo was actually curve-fitting.

Replying as I read.

Yes, let's stay on that problem. Yes, good point. Let us not accept the past combined drawdown value, especially if we obtained it by over-optimizing and testing hundreds of different combinations of systems (to achieve the lowest drawdown and highest profit).

Ok, let's assume that they are not correlated (which is not the case of course), as if we were tossing a lot of coins. Yes, very good point. Then the drawdown can go from zero to the sum of all max drawdowns, PLUS you have to allow some room for a worsening of your systems profitability. But let's keep reading your hypothesis.

Yes, just adding up all the drawdowns is not good enough, because obviously we would come up with a very high uncle point, but investing this way (without taking risks) is easy for anyone. We have to elaborate something more complex which will enable us to maximize the use of capital. Of course if we allocate a million dollars to each system (for drawdown and margin), we will not be disappointed by the drawdown. However, we will be disappointed by the returns.

We are looking for a way, and a formula, that will optimize and define how to minimize risk while keeping returns at the level we want them (and viceversa).

Yes, so far I follow you: if we are trading two systems with a drawdown of 5000 and 4000, we are pretty confident that 9000 will not be exceeded... but we are not at all confident of that in case they are correlated, even though your example excludes that they are. So I would sum it up by saying that in your example we're confident of it, but in reality, with just 2 systems we are not. We would need 50 systems to be confident that 9000 total would not be exceeded (you also have to consider that the future performance is worse than the past).

Another problem is finding a reliable formula for giving us all these estimates you're mentioning (99% confidence that this will happen, 95% confidence about this other event, etc.). I have no idea how this can be done, and it certainly does not seem like something it's going to be easy to do.

Well, there you go. I agree, since you say what I said just now: it's going to be hard. "Monte Carlo simulation" sounds like it is a recipe to solve the problem, but I am not using any formulas unless I fully understand them, or rather if I really have to use a formula, it will have to be the formula i understand the most, so I will not be jumping from one simulation formula, portfolio theory to another, but I want to choose something that is simple, that makes sense, and then I will focus on that (right now i am focusing on markowitz).

So far we're both in the process of assessing all the issues and concepts at stake. You're showing patience, and I am glad. We definitely cannot afford to have the attitude of "it's not a big deal" or "it's easy".

 

[Yamato]

Nobel Laureate Harry Markowitz & Mark Hebner Talk Smart Investing

#1. Nobel Laureate Harry Markowitz & Mark Hebner Talk Smart Investing: Part 1 - YouTube

Oh, this is great. The interviewer quotes him: "...a good portfolio is more than a long list of good stocks and bonds"... great quote, I need to find the book:

"A good portfolio is more than a long list of good stocks and bonds. It is a balanced whole, providing the investor with protections and opportunities with respect to a wide range of contingencies."

- Harry Markowitz in his 1959 book "Portfolio Selection: Efficient Diversification of Investments"

I am getting closer and closer to answering my questions. This focus exactly on my problems.

Found the .pdf here:
http://cowles.econ.yale.edu/P/cm/m16/m16-all.pdf

Here's another good link (it has a long list of links on markowitz):
http://investingforcatholics.com/Articles/Harry_Markowitz.aspx

At this last link I provided, they sum it up very clearly, and they show why I should keep on focusing on markowitz:

The theory developed in Portfolio Selection was a theory for optimal investment in stocks that differ in regard to their expected return and risk. Investment managers and academic economists have long been aware of the necessity of taking both risk and return into account. Markowitz's primary contribution consisted of developing a rigorously formulated operational theory for portfolio selection under uncertainty. His theory evolved into a foundation for further research in financial economics. Markowitz was the first to place a number on risk relative to investing. Risk was previously discussed in general terms and based more on feeling or intuition. He was able to quantify the "undesirable thing" an investor tries to avoid by using a range of possible return outcomes, based on the past variability of returns.

Ok, here's another good quote from that link:

Markowitz concluded that risk is central to the whole process of investing. He then wondered how to measure the appropriate amount of risk to undertake. Markowitz came to realize the cruel truth of investing: investors cannot earn higher returns without taking on greater risk, and the greater the risk, the greater the possibility of loss. He set out to devise ways to help investors apply tradeoffs between risk and return. Using mathematics to solve the puzzle, Markowitz discovered a remarkable new way to build an investment portfolio, which he called the "efficient portfolio.” It offers an investor the highest expected return for any given level of risk, or the lowest level of risk for any given expected return.

So obviously, pecas and I (and whoever else cares) should focus on studying this "efficient portfolio".

 

[megamuel]

Hi Travis,

Hope you are well and had a nice break. I've only scanned over the last few pages as, to be honest, it's getting a bit too complex for me! So apologies if my suggestion is a bit too simple or if it has already been suggested in a more complex way!

But, if the problem is that the drawdown of all 120 systems combines is larger than the account can handle, could you not just start off with a reduced number of systems? Say you have X amount of trading capital, work out what drawdown such an account could handle, then find out the optimal number of systems you can trade that historically have not exceeded that drawdown, and that is your starting point. Say it works out at starting with 30 systems, for example. Then as equity allows, more systems are added. Now I know you want to remove all discretion from your trading but I think that you could easily automate this in Excel.

In the end you would have something like this:

$30,000--------->30 systems trading
$40,000--------->35 systems trading
$50,000--------->40 systems trading
$60,000--------->45 systems trading

etc etc etc.

Until at the end you have all systems trading with enough capital to handle the worst historical drawdown. You can start off with some of the more conservative system and add the riskier ones as capital allows. I think this is what I would do if I had 120 profitable systems.

Anyway, just a suggestion, sorry if it's retarded. Good luck.

Sam.

 

[Yamato]

Well, of course your advice makes sense, and thanks for understanding that you might have missed a few shows, and thanks for the modesty and at the same time the detail of your advice, which is such a rare combination (the last user I banned combined arrogance with lack of detail and wrote: "you're being reckless!").

The problem is that we're much further than that, and we were further even a year ago. The problem is not that I don't have enough money to cover the drawdown, nor that I cannot pick a safe combination of systems.

The objective is to increase the amount of systems traded, and therefore profit, without increasing the risk. I have to maximize this ratio. As I said before, it would be very easy to trade a couple of systems with capital of 1 million dollars, but the profit would not make it worth it.

So the problem is finding a univocal and automated way of selecting systems/contracts so to maximize profit while minimizing risk.

You're suggesting rule-of-thumb concepts, while I am looking for the automation of the whole process and of course I am trying to do so in the most scientific and reliable way.

Until here it was rule-of-thumb mixed with some scientific approach (not enough), and we failed.

If we keep using rule-of-thumb, it might happen again. We need a fixed method and formula, and it has to be efficient. And this is not going to be easy.

As bbmac pointed out, where there's discretion there's a risk.

You see, you say that you know that I want to automate it, but your advice is still discretionary.

Who says that trading 1 system for silver and 1 system for GBP is an automated approach? We have to first study markowitz or at least develop a theory of our own on whether it makes more sense to enable 2 contracts on the good GBP contracts (small leverage) than to enable 1 contract on a silver system (huge leverage).

As markowitz wrote:

A good portfolio is more than a long list of good stocks and bonds. It is a balanced whole, providing the investor with protections and opportunities with respect to a wide range of contingencies.

I can't just say "let's enable 1 contract on all systems", because it might not be optimal. I need to choose systems/contracts according to a clearly defined and univocal theory and formula.

And regarding drawdown, I cannot just say "it's too big with all these systems, so let's disable this and that system". Indeed, how will I decide which systems to disable? This would still need discretionary intervention. And instead I want a good and safe formula to automate and optimize all this.

And markowitz promises all that. And that is what I want. Cfr. my previous posts.

Furthermore, your method is simply advising me to keep doing what we've been doing. But what happens if your appraisal of past drawdown is wrong, as it's been the case this time? We need to bring this much further than rule-of-thumb portfolio selection.

The previous method was very close to science already, closer than what you are suggesting, but we made the mistake of optimizing the combinations by testing hundreds of combinations and appraising their drawdown to profit ratios and choosing the best ones. This is curve-fitting in my opinion. So I need to start doing thing from scratch and do it right this time.

As someone said, "a little learning is a dangerous thing":
http://www.phrases.org.uk/meanings/a-little-knowledge-is-a-dangerous-thing.html

We used a lot of metrics, too many, and we overlooked an important detail, just because we thought that with all those metrics we were safe. If I had to come up with a quick fix I would not use the max relativized combined drawdown, but the average of the 20 max relativized combined drawdowns. But since this, too, would have to be defined univocally, I might as well go back to the drawing board and do it all from scratch. Besides, right now I have no capital so I can spend my time studying.

But if you told me: you need to quickly start investing, then I use the existing method, but look at 20 worst drawdowns and yet I'd also have to consider the largest drawdown. The formula would be as complex as hell. Furtuhermore ,there's all those other questions and doubts I raised.

If I will never get to scientific univocal solution, rather than not investing at all, I would use the previous non-automated approach and a lot of rule-of-thumb. But I consider that still better than what others have advised me to do on this journal.

To further clarify the concepts I am bringing up in this post (as in the previous ones), you simply need to watch this short video which I posted earlier:

If I understood formulas better, I would have gone straight to its two or three formulas and by now I'd be done. But I suck at formulas, so I am having to watch hours of videos, write hours of posts, and read hours of books. It might be faster to go back to school and get acquainted with math all over again.

Here's part 2, pretty pleasant:

Part 3, very good:

"efficient frontier"... "risk/return tradeoff curve"...

 

[Yamato]

more markowitz

Financial Crisis Discussion: Harry Markowitz - YouTube