Algos out, Discretion back in

Just saw this here by Linda Raschke, spot on imo:



"Discretion is back in...

But, let's not call it discretion!

I was reading a daily hedge fund news letter, and here is a direct quote:


"CHARLOTTESVILLE, VA (HedgeWorld.com)-Since 2005 and especially since last year, quantitative fund performance has not been as good as it used to be. Now that many have come to realize that mathematical models are not flawless ("Duh" - my own injection), a new trend in asset management is emerging: integrating the human touch of a fundamental approach into the blind computer-driven process..."


So...let's see......we come up with computer driven models to take the emotions out of trading (the human element), in the hopes that we can arb a statistically positive expectation. Firms hire dozens of PHds to dissect the same time series data and any edge is quickly arbed out. Not to mention that the future does not unfold in the same way as the past which is what the models are based on.

Now, let's bring back the human touch we tried to get rid of in the first place - this human touch now allows an individual to "override" the computer driven signals. (Of course, this individual will always know in advance which signals will work and which will not). So...what exactly is the point?

There is definately an industry bias against terms like "traders", "technical analysis", "discretion"....but for good reason as well. Too many technicians have been negligent on focussing on risk concepts and instead overemphasize "calling" the market. Who wants to allocate large sums of money to a group who can't quantify risk? How is one supposed to know what type of leverage to use?

Definately a Catch-22. But, it still made me laugh to see that the "trend" is now using human judgement integrated with computer driven models....Let's just call a trader a trader...."


LINK:
Discretion is back in... - Linda Raschke Blog

Too many all doing the same thing is just as unsustainable as a business model like shoving credits down the throats of people who can't afford them.

Algos aren't out. An oil trader was chatting to me the other day. A quant in his office had just gone live. He was walking around as if he was a big bad swinging dick as he'd just made USDx,xxx on a gasoil trade. He was getting ready to retire.

The next day he lost it all. And more.

mauzj said:

A quant in his office had just gone live. He was walking around as if he was a big bad swinging dick as he'd just made USDx,xxx on a gasoil trade. He was getting ready to retire.

The next day he lost it all. And more.

Click to expand...

Ooops !

BSD said:

Just saw this here by Linda Raschke, spot on imo:



"Discretion is back in...

But, let's not call it discretion!

I was reading a daily hedge fund news letter, and here is a direct quote:


"CHARLOTTESVILLE, VA (HedgeWorld.com)-Since 2005 and especially since last year, quantitative fund performance has not been as good as it used to be. Now that many have come to realize that mathematical models are not flawless ("Duh" - my own injection), a new trend in asset management is emerging: integrating the human touch of a fundamental approach into the blind computer-driven process..."


So...let's see......we come up with computer driven models to take the emotions out of trading (the human element), in the hopes that we can arb a statistically positive expectation. Firms hire dozens of PHds to dissect the same time series data and any edge is quickly arbed out. Not to mention that the future does not unfold in the same way as the past which is what the models are based on.

Now, let's bring back the human touch we tried to get rid of in the first place - this human touch now allows an individual to "override" the computer driven signals. (Of course, this individual will always know in advance which signals will work and which will not). So...what exactly is the point?

There is definately an industry bias against terms like "traders", "technical analysis", "discretion"....but for good reason as well. Too many technicians have been negligent on focussing on risk concepts and instead overemphasize "calling" the market. Who wants to allocate large sums of money to a group who can't quantify risk? How is one supposed to know what type of leverage to use?

Definately a Catch-22. But, it still made me laugh to see that the "trend" is now using human judgement integrated with computer driven models....Let's just call a trader a trader...."


LINK:
Discretion is back in... - Linda Raschke Blog

Too many all doing the same thing is just as unsustainable as a business model like shoving credits down the throats of people who can't afford them.

Click to expand...

Good post.. Algo trading cannot cope with NEWS DRIVEN DAYS,,, on Technicial days they can kick ass though.
Most of the quantative models have a sentiment module as part of their overall code but the model itself is the weak link. Hence I would not be surprised if the code gives back most of it winnings in 1 bad day .
I feed the sentiment into my code ( Either go short or long only ) before the market opens every day and the result has been impressive.. If market for some reason switches direction due to arrival of new information I have to switch too. I ofetn only set it once a day .

This is the result of todays trades up until now,,, NO LONGS

Grey1

Well done mate !

That make me happy. wonderful mate !

I think you will find that algos have done particularly well this year. Check out Winton capital for example. I looked at the HF indices the other day and saw that the standout best performer this year was pattern recognition. I am not sure what patterns those are but I wish I did know. I have found the last month appauling.

TWI said:

I think you will find that algos have done particularly well this year. Check out Winton capital for example.

Click to expand...

Indeed.

Trader Daily § Wizard Of Winton

GammaJammer said:

There's room for everyone imho.

GJ

Click to expand...

I agree actually.

I think a lot of ones approach to trading stems from which side of ones brain is the dominant one.





Is this dancer turning clockwise or anticlockwise for you ?

Answer first, then click on the test below:

The Right Brain vs Left Brain test

As I wrote earlier in another post, several studies have shown that the first born - in my case an older brother - gets the logical, linear left side of the brain more strongly developed, while the second born - ie me -, gets a more developed intuitive, holistic right side of the brain.

What can I say, between my brother and me that just totally applies.

Applied to trading people with a clearly right side dominance of the brain like me and I strongly suspect Linda B. Raschke as well will clearly go down the discretionary path where we trade off of visual patterns on charts.

Leftbrainers will be far more likely candidates for mathematical, computerdriven algo-models who spend most of their backtesting for all sorts of inter- and intra- market correlations.

And if Market Wizards clearly demonstrated one thing, it's that there are as many approaches to trading as there are net profitable traders, be the method fundamental, technical, or quantitative, and be the trade management mechanical or discretionary, with the only unifying factor they all had in common was a strong focus on risk-management.

Grey1 said:

Algo trading cannot cope with NEWS DRIVEN DAYS

Click to expand...

Why do you think you can't write an algorithm to handle such 'news driven days'? The price action that characterizes it should be plain to see, right?

ahardy66 said:

Why do you think you can't write an algorithm to handle such 'news driven days'? The price action that characterizes it should be plain to see, right?

Click to expand...

Because you need a new kind of Mathematics. Conventional logic has difficulty modelling greed /fear in terms of yes or no.

We need more work on fuzzy logic . SO next time an algo told ya I LOVE YA tell the code where to go cos he does not mean it,,


Grey1

Grey1 said:

Because you need a new kind of Mathematics. Conventional logic has difficulty modelling greed /fear in terms of yes or no.

We need more work on fuzzy logic . SO next time an algo told ya I LOVE YA tell the code where to go cos he does not mean it,,


Grey1

Click to expand...

Makes sense !

Thats where I think rightbrainers have a clear advantage...

Far too much is not programmable, be it simple chart patterns that the eye can see but that can't be explained to a computer because they are never quite the same, or translating one of the main drivers of price, fear and greed, into maths, etc.

Again, not saying that algos can't make money, as very obviously they can, but I do believe that it is far easier for those who can make their decisions visually off of charts.

I also believe that the longevity of far vaguer patterns discernible only to the human eye far outlasts those of algo driven models that, being far more stringent, will sooner or later fall prey to competitors running the exact same model.

Backtest around long enough and you will come up with what Winton are doing, and, besides, their prime brokers won't exactly be fast asleep at the helm either.

;-)

Emulating a system based on strict, linear and logical systems like a fixed income arbitrage strategy - after sufficient computer backtests - is far easier than trading, just for the sake of argument, some classic visual chart pattern, eg a head and shoulders pattern, that will look different to everyone trading it, and everyone trading it will trade it differently.

But Goldman Sachs shouldn't really have any problems figuring out exactly what Wintons maths and hence strict and above all every-time-the-same-rules based systems do, and copy them, eventually reducing their profitability, a vicious cycle that accelerates when the other hedge funds catch up which they will, and sending Winton back to the drawing board for the next new idea.

There's 2 different arguments here. Grey1's saying you can't model a news-driven day, which I don't agree with, and Markus's saying any algo model is going to kill the pattern it exploits, assuming other people eventually see it and come in to exploit it too - which I agree with.

I don't think the patterns a human eye can see and the patterns that a computer can see are 2 separate subsets. If you can find a programmer who is an excellent communicator, I'm sure s/he can wheedle the exact specification of a pattern out of anyone successfully trading these news-driven days.

I bet this has already happened and that algo traders do exploit news-driven days, and to such an extent that it almost isn't profitable.

Logic dictates that as soon as it does become unprofitable to trade any particular pattern, those algo traders will stop and then the pattern will re-emerge.