Reverse engineering why indicators don't work

[MoneyMatthew]

GAT is one of the top contributors on ET and I greatly respect him. But to my taste, only the 2009 onward period is relevant for evaluating the prospects of trendfollowing. The trade became much more crowded post-crash, after the asset allocators all realized they woulda-coulda been sorta-kinda hedged by allocating to the trendies in 2007/8.

I appreciate your insights and I know that you have a ton of knowledge on the institutional side of things. I think generally speaking from a retail trader's perspective who is trading under $200K or even under $10K that these strategies are still very relevant. The alternatives for the retail trader are to put blind faith into untested software and strategies or even worse, roll out of bed every morning and "wing it".

Funds are held to a much higher standard. If retail could learn to just break-even then they would be ahead of the pack.

We have to remember that this is a trading site. The easy answer to pass out to many of the people here is "you are underfunded" and "just get long SPY and go work a day job".

I think if retail was at least aware of some of the basic institutional strategies than that leaves them in a better position to survive than just "I bought super volume order flow fractal indicator for Ninja Trader! Watch me burn my $5K grubstake bankroll!".

 

[Visaria]

How to trade relayed as a song so everyone can understand!

 

[bone]

From my experience, most traders that I have worked with over the years use technical analysis incorrectly. One common mistake I see repeatedly is confirming one indicator with another indicator of the same type (for example, confirming an oscillator with another kind of oscillator). Another pet gripe of mine is the very narrow focus and restrictive use of data sampling time frames - why look at data through such a restricted lens?

My 2 cents, YMMV.

 

[learner2007]

From my experience, most traders that I have worked with over the years use technical analysis incorrectly.

Ain't that the truth!!
While you're saying that regarding the use of indicators, I will say the same for old school TA.

 

[panzerman]

How to trade relayed as a song so everyone can understand!

I've always thought this was a great tune, but the problem is how does Ed define a trend mathematically? It must be defined mathematically so that it can be back-tested and proven scientifically that it is a positive expectancy system. Also, it needs to be able to be coded into an algorithm so that it can be traded automatically if needed.

 

[lovethetrade]

Simple: Backtest it.

Everything else about a theory of how the market works, or how an indicator works, is pure speculation.

While back-tests are an import tool for verification/confirmation, I don't agree with this statement. If you study the market for long enough, you develop an understanding and intuition for what works and does not work without requiring a back-test.

You could spend your whole life back-testing every theory and indicator combination.

 

[tomorton]

I've always thought this was a great tune, but the problem is how does Ed define a trend mathematically? It must be defined mathematically so that it can be back-tested and proven scientifically that it is a positive expectancy system. Also, it needs to be able to be coded into an algorithm so that it can be traded automatically if needed.

Are you trading trends?

 

[Visaria]

I've always thought this was a great tune, but the problem is how does Ed define a trend mathematically? It must be defined mathematically so that it can be back-tested and proven scientifically that it is a positive expectancy system. Also, it needs to be able to be coded into an algorithm so that it can be traded automatically if needed.

The song describes how to do trend following in general terms.

This is my current opinion. Markets have non stationary distributions. In English this means that the win/loss of your 10000 trades you did in the last 5 years are probably not going to repeat for the next 10,000 trades you now do. Compare that to a stationary distribution such as how many reds appeared in the last 10,000 spins of a roulette wheel (it's gonna be a little less than 5000). You can say with very high confidence that in the next 10,000 spins, you are gonna have about the same number of reds.

Back testing may give you some confidence in your approach (i mean no point in trading something that hasn't worked in the past) but to assume it is gonna keep working forever now is a bit ludicrous.

If u understand that, then you realise that the positive expectancy thing and calculating edge is all a bit nonsensical. You either trade well or you don't.

 

[bone]

I've always thought this was a great tune, but the problem is how does Ed define a trend mathematically?

Linear Regression is the most common type of statistical regression analyses, and from my experience regression analyses is the most workable approach to mathematically defining trend in historical trade data. Here are some other types of regression analyses:

• Simple regression
• Ordinary least squares
• Polynomial regression
• General linear model

• Generalized linear model
• Discrete choice
• Logistic regression
• Multinomial logit
• Mixed logit
• Probit
• Multinomial probit
• Ordered logit
• Ordered probit
• Poisson

• Multilevel model
• Fixed effects
• Random effects
• Mixed model

• Nonlinear regression
• Nonparametric
• Semiparametric
• Robust
• Quantile
• Isotonic
• Principal components
• Least angle
• Local
• Segmented

• Errors-in-variables

You'll find that many quants use some sort of statistical analyses toolbox like S-Plus or SAS - there's quite a few good ones out there. Being able to import time and sales data from Bloomberg, for example, is a common and desirable trait.

I am an Engineer by training but I would not label myself a legit "Quant" by any stretch. Hopefully there are some better informed contributors out there.

 

[Simples]

Whatever and however you trade, any directional trading will be dependent on trend. As trading anti-trend is like sailing against the wind, this is something you'd want to avoid naturally.

Blame education and standard tests. These are poor preparations for business, and trading is a business. You can do everything correctly and still lose. You can do everything the wrong way, and win spectacularly. You better figure out why quick! You might be rewarded according to your due dilligence, preparations, study and prowess, but it can take alot longer than you think beforehand! How to survive in the meanwhile?

Proving a negative is pure waste if it doesn't lead you to something that works, or at least make you learn or realize something later.