Auto Mechanics vs. Automated Traders

Quote from chaostheory:
Another painful aspect of trends with a fixed position size is the danger of exiting too early.
...

If you exit from a trend with a fixed position and later see the trend suddenly continue, you're in huge trouble. It's because the psychological urge to reenter the trend becomes overwhelming. It's truly terrible to be on the sidelines of a huge trend. Even worse if you took a lot of small losses to get this this point. Sadly, if you make the mistake of reentering you almost invariable enter at the wrong time and take a major loss. Possibly multiple times. The odds are against reentering a trend correctly.

If instead of exiting entirely, you scale out gradually on retracements and scale back into the trade on continuation, then you avoid all those excruciating problems and guarantee that you ride the trend all the way to the very end even if the trend suddenly bursts into action again.
More...

Isn't scaling out gradually on retracements and scaling back into the trade on continuation just the same thing as exiting from a trend with a fixed position and later seeing the trend suddenly continue (psychologically speaking)?

I mean, it's the same problem: when to exit, when to reenter.
 
Quote from TSGannGalt:

Google sub- and super-martingale.

You have a probability set, and you place your bets according to the "statistical edge" you have.
More...

Thanks for the references.

Interesting. According to the definitions found by doing the above, any kind of scaling strategy where you increase your position size is considered a "martingale" if the trade is going against you or "anti-martingale" if it goes in your favor.

So that makes it appear there is some martingale involved in my portfolio after all. But not really. And certainly not with the same kind of risk when you consider the net effect of playing both trend, counter-trend, and chop at the same time.

Plus, most of the descriptions of "martingale" in my opinion only really apply to win it all / lose it all scenarios.

These mathematical formulas never really account for the effect of the average price when trading which never exists with roulette, flipping coins, or rolling dice.

In trading, we're not doing pure gambling. Instead we control an asset at a certain price and as we add to the position, the average price of entry changes.

That requires a different type of analysis and mathematics than martingale, IMHO.
 
Quote from jjw:

it the system is available as a product, why can you not identify it?
More...

Well, since I built it. That would cheapen this post into an "advertisement" which is both beside the point and against the TOS for elitetrader, most likely.
 
Quote from Pippi436:

Of course there are applications available to retail where you can code your own position sizing methods(including scale in/out, pyramiding, mgale or what have you). Smartquant, Neoticker, and i'm sure quite a few others. You are pretty much only limited by your trading and coding skills.
More...

True, technically you can do it with those platforms. You can probably test these ideas on those platforms with minute data and get close enough results--come to think about it.

Personally I demand the accuracy of tick data and tried with Ninja and Neo but not OpenQuant and to my taste found them inadequate since performance was way below acceptable.

Still this OP isn't about the tools other than to say that if you want to experiment with sizing yourself, you need adequate tools and there's certainly many out there that don't support it at all--nada.
 
Quote from WalterEdward:

Isn't scaling out gradually on retracements and scaling back into the trade on continuation just the same thing as exiting from a trend with a fixed position and later seeing the trend suddenly continue (psychologically speaking)?

I mean, it's the same problem: when to exit, when to reenter.
More...

Excellent question. Perhaps it's just a matter of picking your poison.

Still, the rules for exit and reentry are mathematically graduated scales. So there's no decision making stress during live trading, in fact, it happens hands free since it's automated.

But more to your point, are they the same psychologically?

Well with graduated position size, you're always in the trend until it totally finishes and never on the sidelines. No even for a minute.

Also, when you look at the stats for the trades as a combination trade that varies in size and watch your equity curve, it makes you very, very happy.

However, if you drill down to the individual, discrete trades one at a time, there are some with losses mixed in due to the scaling in and out. It just doesn't psychologically feel like losses.

Plus, I'm describing this all rather simplistically to keep it easy to explain. But there's many optimizations on this to greatly increase the profits and reduce the risk.

One is the time proven technique to buy into trends only on pull backs.

So the system keeps a measure of the % of pull back and only adds to the position according to the formulas during at least X% of pull back from highs.

It begins exiting at X+Y% pull back from highs and completely exits at Z% of pull back which still guarantees a net profit on the overall position.

This reduces the risk of loss on new positions to only X% - Y% rather than losing all from the recent high - Y%.

Anyway, that's just to give you an idea that there's tons of variations to how you can do this which you can figure out according to your style and taste.

It's the general principal of graduated position size that's the focus here.

For example, I'm experimenting with the idea of scaling back the position during new highs and expanding it during pull backs. In theory that works but hasn't panned out yet with testing.

Also, you can filter the signals rather than adding or removing on a purely mathematical grade. Certain market conditions indicate to wait a little bit to get a better price to add to your positions.

Still, none of these refinements are absolutely necessary to profit consistently this way.
 
Here's a thought about another reason why gradual position sizing works so well.

Answer this: How do you measure the "strength" of a trend?

There's many theoretical ways to do it.

But mine in simple. It's the price difference from the current price to when the trend started.

The trend starts every time price reverts to 50% of the last rise.

As long as price hasn't reverted 50% from a high, it's still in a trend.

Why do I measure trends this way?

Well, the farther a trend gets away from the 50% point the more people jump on the bandwagon.

The crazy run-up in Oil is a case in point.

As the price rose an exponentially greater number of people were buying into the trend.

Just like any really strong trend, you started reading predictions of astronomical future oil prices at even more than double or triple the highs.

Eventually, an enormous number of late comers join the trend to pay off the original investors in the trend. (BTW, this makes it practically a publicly sanctioned ponzi scheme.)

Finally, the trend shows signs of weakening by pulling back some because just like Madoff's scheme, it starts to run out of enough masses to keep buying into it.

Then the smart money all exits some of all of their positions. But there's enough masses who continue to buy at this point who may even get excited by the pull back as a better buying opportunity.

That provides enough money to get the original investors -- smart money -- out with major profits.

Finally the house of cards comes crashing down to a retracement of 50% from the highs which wipes out most of the late comers to the trend and officially signals the end of the trend.

So, in short, the strength of a trend is measured by how far it is from the beginning of the trend compared to the 50% line.

Something about 50% is almost magical. It seems that even those true believers who think the trend will continue after all the pull back will finally give up entirely when it gets back to 50%.

Plus all the RTM traders (of which I also play part) start selling into the major pull backs and cover their shorts at the 50% retrace point.

You can look at trend after trend and see the 50% retracement point bounce over and over.

That's where the smart money makes their final major cash flow from the trend and switch over to playing chop or sideways strategies until another trend pops up.
 
in dice and roulette and other common gambling games (not black jack), the outcome of the next event is independent of the outcome of the current event. in trading, though one may feel that one's offer to buy or sell (the current event) is so insiginificant that it cannot affect the next event, it can and does. so the math that explains/preditcts common gambling behavior may not, and imho, is not really applicable to trading (the major assumtions about the interactions between events is different).
 
Interesting, chaostheory. Appreciate the explanations.

I can see that there is a psychological advantage in scaling. As you say, it doesn't really feel the same.

So far as the pure question of profitability goes, though, a comparison of strategies (to the extent that one can compare the two) seems to come down on the side of "all-in" rather than scaling in .

Suppose, in a simple example, I buy 3 units and the price moves up 10 points so that I have a profit of 30. On a 25% retracement I exit for a profit of 22.5. On a 50% retracement I reenter at 5, the price goes to 10, and at that point I have a profit of 15. My total results are 37.5

But if instead I enter with 1 unit and then add 1 unit at every 2.5 point increase in price (to my maximum of 3 units), then when the price reaches 10 I have a profit 22.5. If I exit 1 unit on every 25% retracement, on a 50% retracement I have booked 5.0 (LIFO)and have 1 unit in inventory. I repurchase 1 unit at 50% and add my third at 7.5 so that when the price reaches 10 again I have a profit of 17.5. My total results are 22.5.

However, as you point out, there are many different ways to do the scaling in and out, and I may need to work on those until I find the better way.
 
I realized after my last post that the real advantage to scaling-in is the reduction of losses while waiting for the trend.

In my example, the all-in advantage is 15 points. This would be neutralized by 6 two points losses prior to the trend. The all-in profit would be 1.5; the scaling-in profit would be 10.5

So whether all-in is better or worse than scaling-in is perhaps a function of the losses taken while waiting for a trend. But this is a parameter that can be quite different for different traders. If there are a large number of losing trades relative to the winning ones, or if those losses tend to be large, then it may be that scaling-in has a substantial advantage.
 
Thanks to chaostheory (and to WalterEdward, too) for his detailed explanation of the advantages of "scaling into/out of positions" and his many answers to my endless questions and doubts. Congratulations for your patience.

He does however mention some things, here and there (psychological factors and so on), that somehow refer to discretionary trading, which should have nothing to do with automated trading, which is what I am doing - but I don't want to ask anymore questions, because at this point I would get annoying.

I just want to put on record, for myself (for when I'll read it later), that I still don't understand the advantages of "scaling into/out of positions" for automated traders (not for discretionary traders, mind you) and I don't understand how I could effectively backtest a system that uses the technique of "scaling into/out of positions". In sum, since it is unclear (to me) what the advantage is and it is complicated to measure it, I will set it aside it once again, for the same reason I did the first time, when I read about it on Jesse Livermore's book and the second time on Balsara's book.
 
You must log in or register to reply here.
Back
Top