A few Alpha Algo examples and some lateral thinking ideas from the Autoscalp Team

A venture capital screening tool

SeedStartup is a Dubai Based Startup Accelerator and Seed Venture Fund for Startups From Around the World.

Through AlphaMetrix... the world first hedge fund marketplace:

• Fund Managers: Access Resources to Build their Business
• Investors: Build More Robust Portfolios

What is an alpha algo :

An Alpha is the very core of a speculative idea.
Before it becomes a trading system,
before it becomes a strategy.
Alpha is what was used to be called "animal spirit" in a culture "dragon breath" in another.
Matteo Gandola, autoscalp creator says <<Alpha is where know how lives>>.

Why our alphas are different from competitors' ones ?

<<Because our Alphas are based on market experience , not just on coding experience>>.
One might think that just a few HFT skills are required to develop alphas but
you will be surprised to know that  most C.V. we trash at autoscalp are  :

• "quant analyst"
• "experts in trading system coding"
• "low latency expert"
• "c++ , mt4, java , R language trading system coder..."
• "PhD... in math... statistics..."

Market experience is the very core of a speculative model, not coding.
<<first you must have a clear speculative idea in mind , then look for a coder>>

An alpha algo creation example :

data set :

1. assume you have 3 seconds lag to interbank price

2. assume you have a spread widened 2 pips

3. assume you feel confident because your broker gives you "guaranteed stops"

then you feel like developing a fast system to scalp market inefficiencies...
(there are many ways: from triangular arbitrage, trend following, next price forecasting...)

but after you code your "wonderful theoretically profitable trading system"
with a perfect and profitable back testing you end up wasting
a ton of money and even don't get the real reason why you lost all those money.

Ok that algo failed, now: was it bad coding ... or lack of experience?

If the above 3 assumptions were true  real problem is that
you were not trading in real time.
Giving such combination of factors above, you are delayed
at least 3-4 minutes without knowing.

If you did such mistake, you miss experience: 
even if price appears to you in real time , it is not.
Spread widening and slowed execution is a deadly combination likely
to make all your profits disappear.

The core of this example then is:
if you are away from real market price , is useless you waste time in coding.
Experience also says that one of the most dramatic issues you can face is
that most of time coders and algo developers tend to stick to their code
even if is not profitable at all.  Removing an algo is more expensive than creating it.

Conclusion:
an alpha is not necessary a trading system or strategy but it can be even
a decision filter, like the  one above,
to prevent wasting time in allocating precious resources in useless coding .

Another alpha creation example : spotting an artificial liquidity injection pattern.
We took an eurusd ticks data sample and converted it until played it as wav sound as
we identified a recurrent humming noise in the background we discovered that was
a systematic artificial liquidity injection in the order flow from liquidity provides.
Then we have set a few models to wait for such injection before entering market.

Hear eurusd ticks play as a sound

CRACKING AN UNBREAKABLE CODE :

We stumbled upon the unbreakable Dorabella Cipher reading this:

reference: Daily Telegraph world top 10 uncracked codes.
See #7 Dorabella Cipher

After some think tank work... it comes out that to us it looks like a song,
not a text message.
This is why cryptographers failed to crack it for 200 years.

Dorabella Cipher was written to the daughter of a preacher from her admirer. Who was ,
yes a great math expert but a musician too. Today if you were him,  you would probably
send her your CD and he did exactly the same but he lived in 1800.
So guess what he sent her? Yes, obviously, a love song.

Dorabella Cipher is simply a song NOT a text message.
Symbols indicate fingers positions on a keyboard ,
where position of next note is relative to previously played note.

Conclusion: cryptographers failed to decrypt  Dorabella cipher for 200 years
because their assumption: -it must contain words- was simply wrong.
We cracked Dorabella cipher because we did not know that it should have contained
words rather than notes.The alpha algo was : always doubt the data set.

Exploitation: Now rethink same alpha algo (-always doubt the dataset-)
but  apply it to financial time series.  Historic prices covertly inherits the feed
provider's interpretation of such data which , as any interpretation can be correct
or... not. We often discover that biggest flaws come from
the input dataset not even from its analysis.

ITALIAN BANKRUPTCY COUNTDOWN

triggered on 1/3/2012 - 8 Raby` al-THaany
Sorry Italy: time is up !

The model:
we took 2 years history of the (forced) ECB buying of Italian bonds and rendered it
as sound file to filter background noise. That "humming sound" was quickly coming
to a fade out within mid april 2012. BUT For every day btp - bund spread stayed
above 5,00 we subtracted one day to the deadline.
Countdown then pointed to the date above.

Our conclusions:

Among EU countries at least Italy is already in technical default .

Since EU commited all its reserves to save Greece rather than keeping them for
bailing out the Italian bonds. Accepting junk collateral to mask it won't help.

A Public disclosure of Italian Bankruptcy date will be relative to an
Italian Bonds refinancing / expiry day, when "surprisingly"
EU will openly admit has no money left for that bail out.

Can read more about that issue on money morning website

EU CRISIS FOR DUMMIES :

Money printed from ECB stays in virtual pallets but does not get distributed .

This is one trillion cash in 1$ notes lying on its delivery pallet ready to be shipped.

1 trillion (2 years output of ECB)	100 Million for comparison

So you now know how much a tillion is !
Good, now also assume one billion is a single a4 paper sheet with a stamp on it .
That 1 billion dollar paper never gets out of the banks
which lend it back to the ECB and receive 1% in return.

How many trillions does The European Central Bank  print in one year ? 0,5
How many trillions are needed from the people to reboot European economy ? 2
How many trillions are received from the population through banks in one year ? 0,1

So, either you wait  20 years to reboot EU economy or banks are the problem.
What ? Wait ! where are the rest 1.9 trillion european people need to restart the economy ?
Stored in banks derivatives / safes. Those virtual pallets are stored, not distributed.

Until banks borrow for free from ECB and lend to governments at 1%,  economy won't restart.
No matter the lies politicians tell the people and whatever is their political color.

Until those money lie on pallets and don't get shipped,  EU economy won't restart.
Provided there is still time for a reset. Thing which we at autoscalp, strongly doubt.

THE DOLLAR SWITCH OFF :

in an unseen shift , world is exiting the dollar based economy.

Summary of a hot topic discussed during online workshop of the autoscalp network:

While every western eye is concentrated on guessing the eurozone crash consequences
all BRICS countries in less than one year declared they don't want to trade oil based
in dollars anymore and even banned dollar from internal trade settlements .

What if the recent the fall of the oil price is NOT a correlation issue to the exchanges
but a shift in balance of powers ?

It happens every 120-150 years:
and is called  base currency shift or reserve currency status change.



Autoscalp think tank analysis results are:

1) more resilient is the base currency to its shift,  higher the costs to exit the crisis

2) superimposition between the old base currency and the new one lasts a 50 years in which most of time coincides with a major war (such migration has never been painless).

3) currency circulation has never been so fast in history and is also accelerating

4) HFT market making models could boost the dollar switch-off or even totally derail it.

Our Conclusion:
the switch might happen in a "snap" it could be ongoing now since all it takes today
is some synched server farms and a few HFT engines.
It could be so fast swift and silent than media could even fail to report it.

Update:  half Africa BRICS and now Japan use Yuan as base currency practically the last ones left to use USD  base currency are Britain, Europe and USA. Just in case you missed that news. That's when Japan started usin Yuan too.

HFT EXPONENTIAL GROWTH

year by year:
HFT Market growth and implications in markets decentralization :

In our opinion HFT market cant be stopped by Tobin Tax or anti-hft laws as today the liquidity stream is simply so wide it will simply find more efficient way to be traded , outside centralized markets.

The more government apply pressure on liquidity such as anti-hft laws or volume compressing taxes such as tobin tax,  the faster liquidity will flow out of centralized markets.

Look at the chart below. Today HFT is a flow so huge that trying to put brackets to that pressure will just make liquidity look for faster, more efficient places to flow , such as Singapore or the hyper technological Mongolia stock Exchange ...what Mongolia ???
Yes. Mongolia stock exchange did not almost  exist a couple of years ago , now being one of the highest technologically advanced , being it one of the most recently built. That means Mongolia is one of the fastest around ,so maybe that's the point... you didn't notice.
Now look again at the chart below and ask yourself if still has no meaning...

THE END OF RATING AGENCIES :

Privately Owned Rating Agencies should be immediately replaced with a permanent commission of the United Nations who should be granted Exclusive Competence on Country Rating.

When rating agencies started their job of giving marks to government, simply put there was no United Nations and banks needed a reference to track their investment in foreign , uncooperative countries. At that time Rating Agencies were the only options. Today they are obsolete and create more and more mistakes which you see on the every day news.

Historians will likely look back in a few years and identify such raw management of Country Rating in private hands as a con-cause of the ongoing European collapse and responsible of a couple of crisis in the US. Since we are too close to such events most of us can't focus it yet.

Those agencies are more or less privately owned companies. Not super-partes and there is a person somewhere who receives a monthly salary in the thousands range but mangages the downgrade worth Billions if not Trillions. This alone speaks "flawed".

Also that person is not supported by an international community but just some collegues.
Who can choose...to agree on such mark or either be fired as in every private company.
Then the person on top  has a power beyond comprehension, which  I try to summarize below:

A small team of people, receiving
10.000 salary is to decide if a govermnent is worth
15.000.000.000.000  if rating is A
12.000.000.000.000  if rating downgrades to B

The real question until rating is in private hands is :
What is the corruption cost of the whole structure ?
Dimes....compared to the shock wave on a government when a rating downgrade happens.

Our view: You can't privatize transparency.
Country Rating must become an U.N. job.

N.Y. SKYFALL

THE NEXT FLASH CRASH : 

we think will happen between mid may and early november 2013.
At the end of the SKYFALL day N.Y. will have lost between -10% and -15%.

HFT market faliure already happened... just scale it up.




"BATS Market IPO crash." From birth to death in 750 milliseconds.  Image Source: Nanex.

Here is how the Autoscalp think thank sees the next flash crash will happen:

It will start quietly , an ordinary day: no particular releases or expected liquidity injections,
no breaking-news, but 30 minutes ahead of NY OPEN... hell happens.

The 2013 Skyfall Flash Crash step by step :

It will start quietly , an ordinary day: no particular releases or expected liquidity injections,
no breaking-news, but 30 minutes ahead of NY OPEN... hell happens.

1)
A remote decentralized pool abruptly posts that is "out of shorts" and halts trading.
One of the brand new ultra low latency decentralized markets in a faraway country,
probably trading an index related ETF .
In less than 2 millisec  such situation is sniffed by most of other market algos
either the sentry ones (the ones that prevent crashes) and the speculative ones .
Things which  should prevent a disaster to propagate but
end up working exactly the opposite:
as every market starts cutting liquidity into their local pools "sell" order pipe list
gets washed out (executed) so many times till it keeps empty and price keeps falling.
A few millisec later each market starts  cutting liquidity into their local pools on
the shorts side  into every sub pool and all customers, not only retails, get cut out.

2)
Event which creates the the infamous signal : Panic please !
Since all that should have worked to close the gap and  fail safes were  triggered
in the same moment and all markets are doing same thing...
result is that it accelerates the liquidity crisis on the sell side.
So that  "halt all shorts" reaches the internal pool of each market maker too which
is perceived as "short more" .
Unfortunately again each market is populated by independent algos and  while
dumb algos start pipe lining a ton of sell orders , smart algos quit trading , in either
case another even stronger signal propagates now in any platform...

3)
"No match on short orders" is read in any market  and spike hunting propagates
at lightning speed. At that point any market and any algo practically say same thing:
"no order match on shorts" , the wave is ready to become a tsunami:
N.Y. market  goes "skyfall" and other markets follow.

4)
NY  falls -5%
Then the NY main fail safe kicks in and blocks trading.
Another mistake, this is not synced for all decentralized markets which continue the
free fall float of the index swallowing 100 billions a second.  After fail safe no-trade
period expires, NY reopens and goes straight down...

5)
NY falls -10%
At that time the disaster goes at the frantic speed of 500 billions a second and is
spreading worldwide on all exchanges ,triggering one of the major stock exchanges
crash in history quite similar to the "trillion dollar bet" that destroyed credibility of
the Black Scholes model.

At that point the ultimate  NY market fail safe kicks in again and halts trading for a
much longer period but as price is not bottoming ,  it can't rebuild floor.
NY market completely halts for hours.

6)
Just after a worldwide "halt trading"  is reached NY price finally stabilizes between
-10% and -15%. At that point major banks and brokers post a warning such as:
"for lack of collateral reasons we are not reopening until prices normalizes".
Which means they officilally won't know what happened for 48 hours (netting time)

Conclusion:

If some trillions evaporated in one trading day , where did all liquidity go ?
Our opinion:  such liquidity never existed .

Now... some pictures which inspired us some models and discussions,