Entries and Exits Don't Really Matter

[dcraig1]

I've said this and I'll say it again: Mr Tharp does not trade. Listening to him about trading would be like listening to your grandad trying to talk like an expert about fishing despite the fact that he has only once tried fishing and he hangs around with a lot of fisherman during his life.

I never knew you could make so much money by charging people for advice that you never put into practice, because you failed at it before. There's an entire institute built because of this. If that's not sad then I don't know what is . . .

Even if he did trade, his book would still not be worth much. I should have known better. Anything with phrases like "financial freedom", "secrets" and such like in the title is usually best avoided.

 

[timsk]

The position sizing methodology is designed so that the daily return stream from each market is statistically indistinguishable from any other, and it does accomplish this goal regardless of the entry/exit methodology employed.

Hi jj,
Either I've not being paying close enough attention (very likely) or you're yet to outline the position sizing / risk management strategies that you recommend? I'm loathed to ask you to 'dumb down' your thread, nonetheless, I'd be grateful if you can keep the maths to minimum and explain your ideas in a manner that would be appropriate for an Early Learning Centre. I know, my request borders on blasphemy to someone like you!
😱
Tim.

 

[nine]

Do it both ways. A simple (lol) english language version and a complex mathematical version.

Personally I'm with Taleb and I always favour Monte Carlo sims for RoR followed by attempts to figure out worse cases than the data can project (the Black Swans).

 

[the blades]

I do believe the blades also trades stocks in this way, hopefully if he has time he may stop by this thread, I'm sure it would interest him greatly.

yep, an interesting thread Mr McQueen.

I'm not yet sure where this post fits in as I'm to time strapped to digest all this at the moment, but here's what I do;

I trade a basket of stocks (about 20% UK through spreadbetting, 80% US through IB) that are from about 20 mechanical screens. I hold aroundf 120(!) stocks with anything from a few % to up to 15% of my portfolio to each screen.

I filter down the stocks that pass the screen by eyeballing the chart - basically looking for stocks in an uptrend or that have broken a downtrend. If they pass that "test" I then set limit orders to trigger entry, based on pullbacks, using fibs / S/R but nothing very scientific. Exits result when the stock fails to pass the screen, a limit order is set pased on fibs again or basic S/R lavels.

My overall portfolio is then hedged using ER2 contracts and spreadbets against the FTSE and £/$. The level of the hedge varied depending on signals that I pay for (there, I've said it) but overall the whole "system" would be profitable and maybe less risky without this.

I devote about 5 hours a week to the maintaining the portfolio. Any mug could do what I do (well I am a mug, afterall), though you need a relatively large account to hold the volume of stocks.

As I say, where this fits in to this thread is debatable but as you mentioned me, it felt rude not to contribute😛

From April 25th I'm a full time "trader" and will hopefuly contribute more valuably to these forums. I'm not sure what I'll add, only that if I can do it just about anybody can.

UTB

 

[Mathemagician]

Hi jj,
Either I've not being paying close enough attention (very likely) or you're yet to outline the position sizing / risk management strategies that you recommend? I'm loathed to ask you to 'dumb down' your thread, nonetheless, I'd be grateful if you can keep the maths to minimum and explain your ideas in a manner that would be appropriate for an Early Learning Centre. I know, my request borders on blasphemy to someone like you!
😱
Tim.

The piece you quoted is as detailed an explanation as it gets. Just as the specific entries/exits used in this example don't really matter, neither do the details of the specific position sizing methodology. Remember, the point of the thread is not to place an original piece of work that is of great value into the public domain. Rather, the purpose is to illustrate that most traders place far, far too much focus on entries and exits at the expense of money management and position sizing, and that there is a good bit of edge to be had in fiddling with the latter. Beyond that, it's up to the individual trader to pick up the proverbial ball and run with it!

jj

 

[Mathemagician]

One thing I would be happy to do is provide an overview of the public domain position sizing methods if you think that would be helpful. For most purposes they fit the bill quite nicely.

jj

 

[timsk]

Sorry jj - misunderstanding. I thought you were going to unveil a new or little known technique that in your view is better than the standard position sizing methods out there and utilized by most traders. My mistake!
Tim.

 

[Mathemagician]

No sweat, tim. FWIW, I've actually been running a small account using really aggressive fixed ratio since 1/17 and tracking it using a t2w blog. I've found it to be quite insructive and it's probably worth a mention here. When I get some time I'll start working up some posts on the standard sizing and portfolio methods.

jj

 

[dcraig1]

yep, an interesting thread Mr McQueen.

I'm not yet sure where this post fits in as I'm to time strapped to digest all this at the moment, but here's what I do;

I trade a basket of stocks (about 20% UK through spreadbetting, 80% US through IB) that are from about 20 mechanical screens. I hold aroundf 120(!) stocks with anything from a few % to up to 15% of my portfolio to each screen.

I filter down the stocks that pass the screen by eyeballing the chart - basically looking for stocks in an uptrend or that have broken a downtrend. If they pass that "test" I then set limit orders to trigger entry, based on pullbacks, using fibs / S/R but nothing very scientific. Exits result when the stock fails to pass the screen, a limit order is set pased on fibs again or basic S/R lavels.

My overall portfolio is then hedged using ER2 contracts and spreadbets against the FTSE and £/$. The level of the hedge varied depending on signals that I pay for (there, I've said it) but overall the whole "system" would be profitable and maybe less risky without this.

I devote about 5 hours a week to the maintaining the portfolio. Any mug could do what I do (well I am a mug, afterall), though you need a relatively large account to hold the volume of stocks.

As I say, where this fits in to this thread is debatable but as you mentioned me, it felt rude not to contribute😛

From April 25th I'm a full time "trader" and will hopefuly contribute more valuably to these forums. I'm not sure what I'll add, only that if I can do it just about anybody can.

UTB

Are your screens both technical and fundamental ? Have you looked at the possibility of hedging by shorting weak stocks rather than using index futures ?

 

[Chorlton]

My overall portfolio is then hedged using ER2 contracts and spreadbets against the FTSE and £/$. The level of the hedge varied depending on signals that I pay for (there, I've said it) but overall the whole "system" would be profitable and maybe less risky without this.

I'm sure this is probably way off-topic now but could you explain this in more detail. I've read similar strategies many times but never clearly understood what is involved and what the process is.

Thanks in advance,

Chorlton

 

[StratOpt]

The last time that I recall JJ doing an aggressive money management testing he was able to achieve over 1000% gains in the testing time period so this should be interesting to follow. It can be a wild ride though for sure.

 

[the blades]

Are your screens both technical and fundamental ? Have you looked at the possibility of hedging by shorting weak stocks rather than using index futures ?

the screens are 90% fundamental - only relative strength or all time high prices on the "technical" side.

I intend to look at shorting weaker stocks when I have more time to devote to it. Some basic strats I've looked at do offer increased return, but at much greater risk. That's not to say it doesn't work, just that I haven't found the answer yet.

UTB

 

[the blades]

I'm sure this is probably way off-topic now but could you explain this in more detail. I've read similar strategies many times but never clearly understood what is involved and what the process is.

Thanks in advance,

Chorlton

no bother Chorlton,

Let's say you have a (long) portfolio of £100K. You would short an equivalent ammount of an index - say a £20 per pt down bet on the FTSE at 5000 (to use round numbers again).

I'm interested in relative performance. So my long portfolio might lose 10%, but if the index loses 15% I'm 5% up. The flip side is I might gain 10% but still lose 5% if the index rises 15%.

I'm relatively confident I can outperform the index. I'm less so that shares in general will go upwards in future - hence this strategy.

Feel free to ask for further clarification, If I've not answered anything.

UTB

 

[Chorlton]

no bother Chorlton,

Let's say you have a (long) portfolio of £100K. You would short an equivalent ammount of an index - say a £20 per pt down bet on the FTSE at 5000 (to use round numbers again).

I'm interested in relative performance. So my long portfolio might lose 10%, but if the index loses 15% I'm 5% up. The flip side is I might gain 10% but still lose 5% if the index rises 15%.

I'm relatively confident I can outperform the index. I'm less so that shares in general will go upwards in future - hence this strategy.

Feel free to ask for further clarification, If I've not answered anything.

UTB

Hi Blades,

Thanks for the explanation. I understand the logic behind this but i'm surprised that this strategy works so well but then again I have no experience with this method. I assume the stocks in your portfolio must match the index performance / movement pretty well then?

Out of interest, would having a filter in place that restricts new trades being taken as a result of adverse Index movements achieve a similar result?

Cheers,

Chorlton

 

[the blades]

Hi Blades,

Thanks for the explanation. I understand the logic behind this but i'm surprised that this strategy works so well but then again I have no experience with this method. I assume the stocks in your portfolio must match the index performance / movement pretty well then?

Out of interest, would having a filter in place that restricts new trades being taken as a result of adverse Index movements achieve a similar result?

Cheers,

Chorlton

the idea is to protect against significant falls when all stocks are hit by sentiment. I'd rather have a 20% return with 5% drawdown than a 30% return with 20% drawdown.

The hedge is about protecting against things that might happen to existing holdings, so a filter to restrict trading after the fact may be useful, but I think different.

UTB

 

[Chorlton]

The hedge is about protecting against things that might happen to existing holdings, so a filter to restrict trading after the fact may be useful, but I think different.

UTB

Ahh..... Yes, Good Point!!!! 😱

 

[dcraig1]

Yeah, I never found much use for the Van Tharp stuff either. Unfortunately, as you've noticed, there just isn't much out there. Ralph Vince is about as good as it gets. His books are mostly repetitive, though, to the point where I can't really tell them apart without looking through them.

jj

Perry Kaufman has a chapter in "New Trading Systems and Methods". Kaufman is suggesting using a genetic algorithm to find "optimal" size for each system. Which is handy, as I was thinking along those lines as well.

I guess overfitting is, as always, a danger.

 

[Mathemagician]

There is no fathomable reason for using a GA for this purpose as it relates to one system. It's a one dimensional problem and can be accomplished by brute force very quickly. If he wants to go looking for optimal weights of a portfolio of systems, this is simply asking for trouble as there are few more powerful methods for overfitting to past data.

As an example as to why (and why you need a GA), suppose you've got 10 models and allow the relative weights to take only integer values between 1 and 10. This yields on the order of 10^9 or 1,000,000,000 combinations, far more than most trading models allow which creates an enormous potential for overfitting as the data set is significantly smaller than this. If each model itself has just 100 possible configurations and these are optimized along with the relative weights, you've now added another 20 zeroes to the number of combinations. This gives you a parameter space containing something on the order of 10^29 possible configurations. This is, um, a lot.

More to the point, though, is that unless you've got a huge amount of good, clean, and relevant data (very difficult to come by) optimizing position size in any way and taking the results seriously is more dangerous than beneficial.

jj

 

[dcraig1]

There is no fathomable reason for using a GA for this purpose as it relates to one system. It's a one dimensional problem and can be accomplished by brute force very quickly. If he wants to go looking for optimal weights of a portfolio of systems, this is simply asking for trouble as there are few more powerful methods for overfitting to past data.

As an example as to why (and why you need a GA), suppose you've got 10 models and allow the relative weights to take only integer values between 1 and 10. This yields on the order of 10^9 or 1,000,000,000 combinations, far more than most trading models allow which creates an enormous potential for overfitting as the data set is significantly smaller than this. If each model itself has just 100 possible configurations and these are optimized along with the relative weights, you've now added another 20 zeroes to the number of combinations. This gives you a parameter space containing something on the order of 10^29 possible configurations. This is, um, a lot.

More to the point, though, is that unless you've got a huge amount of good, clean, and relevant data (very difficult to come by) optimizing position size in any way and taking the results seriously is more dangerous than beneficial.

jj

Your Optimizer doesn't have to adjust individual system parameters, just the weight of each system in the portfolio of systems.

Position sizing a portfolio of systems (however you do it) is a form of optimization. It depends what your definition of fitness is.

 

[StratOpt]

Optimizing the weights is exactly what he is saying that you do NOT want to do .... although I do something very similar to this