Automatic Trading - Is There anything like that ?

trady1

Junior member
Messages
15
Likes
0
Automatic Day Trading - Is there anything like that ? - part A
By: Buky Carmeli TradySupport @ Trady1.com)

Preface

Professional day traders would probably agree that the most important factor in day trading is the ability to ignore what others may refer to as "human been emotions". In other words, if you decide to buy and sell stock, particullary if you decide to do so on a daily basis, then you better adopt a good "mechanical" strategy and track that one, rather than rely on your personal feelings. Well... that's very easy to say, but... what is exactly a "good trading strategy" ? Which strategy may be the one that may cause a normal human been to ignore his basic emotions ? Is there any thing like that ?

Choosing a trading strategy

As we start it is important to understand that technical stock trading (and particullary day trading) is based on a statistical analysis of past behaviour of buyers and sellers. In other words, the said analysis may provide a good oversee of what most buyers and sellers used to do in the past, but it cannot provide any guarantee that buyers and sellers will keep doing the same actions, and present a similar behaviour in the future. That will lead us into a first conclusion that any trading strategy cannot provide 100% success (not even close to, by the way...). Understanding that will lead us to one of the most critical questions a professional day trader shall explore: is the adopted strategy good enough ?

Let's examine the following situation once a day trading strategy is adopted by a trader. In the first day the strategy is handled by the trader and yields a profitable day (i.e., stock is purchased, tracked, held and sold by the exact definitions of the strategy). In the next day, the same strategy - same definitions, same trader - yields mix results, and in the subsequent day the trader loses money. Then, the said trader may decide to adjust his strategy, or even to replace it by a new one. And once he does it he needs more trading days to check that new - or adjusted - strategy and so on. Clearly, being a "jumpy" trader won't provide any good decision, however in addition to time and patient factors, we bring up a very important key to use while choosing a trading strategy: expected results. If you expect to double your money (i.e., 100% profit in about 250 trading days) then a very aggressive model is required. If you just expect to reach an annual profit rate of 25% then a different model may be used. So, to the question "is that strategy good enough ?" the answer is "depends...". And by the way, it is more easy to identify and decline a bad or not good enough strategy then confirm otherwise.

So what is a day trading strategy ? What's behind this term ? Basically a strategy will be defined by 3 major components: (a) Buying decision (b) Holding methodology (one may call it "patient"...) and (c) Selling decision. As said before, a good enough strategy will be a one that provides the expected results over a period, however, in order to take this discussion into more practical aspects, it is necessary to "dis-assemble" the term "strategy" into its components (buy, hold and sell) and examine each and every by itself. The following sections will propose how to examine a day trading strategy. Note that I'm not proposing a strategy here, this will be discussed later, I just bring up several important examination factors of how to judge a given trading strategy.

Buying decision

There are many different factors and reasons why buying a certain symbol at a certain time during a day. From a technical standpoint, the key is to check how the price of a purchased symbol was changed since a buy order was placed. Note that for LONG stock trading a desired price change must be positive (rise) and for SHORT trading it must be negative. The change examination will refer to the change in price since a buying decision was made and the time duration since that decision. Clearly, a change for example of 60 cents made during 30 seconds won't let a trader to enjoy that decision (from my experience, if that is the situation and you just bought this symbol - run out as fast as you can - 60 cents in 30 seconds ? that is probably a manipulative stream of deals). From the other hand, if a change of 15 cents is reached in 10 minutes, then during these 10 minutes you may buy and sell (and I guess reach about 10 cents profit out of the 15 cents change) or you may decide to hold your positions and make more profit (or lose all...). To conclude that point: A good buying decision will be post verified by the rate of change ("slope of price graph") in price since buying point till the daily price peak (either highest or lowest depends on type of trading).

Holding methodology

Here comes the hard part, the one where emotions become very intensive. Once you buy a symbol you have to remember that other traders (of all kinds, not just day traders) are watching that symbol and everyone has its own strategy, i.e., everyone wants to sell it at some point, either today or later. Also remember that trading is eventually a "zero-sum" game, for any buyer there is a seller, and vice versa. Therefore, the price per share may be changed up and down (above and below the buying price) and may bring you to re-consider your previous decision to buy this symbol. Panic is a very human been feeling and no one will blame you if you decide to sell your positions at some point because price change is unexpected. Greediness is also - like it or not - a human been common behaviour. You might want to maximise your profit - by extending the holding time - and you may get larger daily profit or lose all. The post examination of your holding strategy is the delta in price between the maximum daily price peak (or the minimum if you buy short and expect to cover), and the actual selling (covering) price at the time and point you eventually decided to place the order.


Selling strategy

Generally speaking, there is a strong connection between Holding and Selling strategies. Clearly, the point where you decide to sell your positions is the point where you stop holding it, however, selling may be executed in one order or in two or more splits. If you sell all your positions at once you close your position with this specific symbol and hopefully you make profit on it. However, you may want to sell your positions in two phases, one part now and another one later. You may also want to split your positions equally between phases. Once you sell a portion of your positions, you either "lock" your current profit or current lose. On the other hand, since every sell order cost you a flat commission, you may want to split your selling activity very carefully. The examination of a given selling strategy will be based on the delta between the actual profit/loss reached by the trader, versus the profit/loss that could be reached if you would take a sell-all-at-once action.


What next ?

Few words about the author and TRADY. I'm acting as a day trader for several years. I've a strong and extensive computing and software engineering background.

In 2004 I decided to combine my day trading and engineering experience and I started to develop a powerful multi-computing array for day trading. The name of this multi-computing array is TRADY. This multi-computing array consists of 20 servers computers and can be extended up to 80. It works in parallel, each computer is responsible to track 10 different symbols (tickers). Several computers inside this array are also responsible to find out the best symbols to watch at a time, so my daily portfolio is dynamic, and changed from day to day, and in most cases several times during the day as well.

Once a list of 1600-1700 symbols is built (those who are considered to be the best at a certain time), a narrow list of about 150-160 symbols is picked by the computers and watched very closely. Each and every symbol is sampled 10 times in a minute (!!). Per each and every symbol a complex ed calculation is made by computers, in order to determine whether this symbol price can trigger a buy order, or sell order or just ignore. Based on more than 35000 symbols, all were automatically selected by the system, each was sampled about 3200 samples per day, a statistical trading strategy was developed.

In the next part (part B for this article) which I plan to provide within two weeks, I will extend about TRADY, how it picks symbols, how it decides what and when to buy, and when to sell and so on. I will also review the trading performance of TRADY.

I plan to review the whole stuff in three parts. Your comments will be more than appreciated.
 
Go for it !!
sounds very interesting
what is it you are looking at statistical info or a form of pattern recognition


Andy
 
Hello and good morning

The quick answer is both, even a pattern is eventually a statistical model that describes buyers and sellers behaviour. The technological key behind the TRADY multi-computing system is to be able to open a wide scope (for hundreds and thousands of symbols at once) and to be able to provide enough computing power to each and every symbol separately.

It is possible to show that more than 35000 different symbols which were automatically picked by the system, (and each one is sampled about 3200 times over a trading day hours) can be divided into several pattern-based categories, each one can be mathematically described and managed. The rest of that will be placed in section B of my email.

Thank you for your reply.

Buky Carmeli

TradySupport @ Trady1.Com
 
cool
i tend to do it the old fashoned way, maybe time to upgrade the old brain
good luck on your venture


Andy
 
This multi-computing array consists of 20 servers computers and can be extended up to 80. It works in parallel, each computer is responsible to track 10 different symbols

It only samples 10 times/minute? what is wrong with realtime?
Good luck.
 
Looks like just what the organisers of the automated trading championship are looking for.
They are based in Kazan ( Russia ) wherever that is.
Comp starts 1/10/2006, bring your own camel and sandwiches.
The prize money looks good though. If Putin will let it go !
 
In 2004 I decided to combine my day trading and engineering experience and I started to develop a powerful multi-computing array for day trading. The name of this multi-computing array is TRADY. This multi-computing array consists of 20 servers computers and can be extended up to 80. It works in parallel, each computer is responsible to track 10 different symbols (tickers). Several computers inside this array are also responsible to find out the best symbols to watch at a time, so my daily portfolio is dynamic, and changed from day to day, and in most cases several times during the day as well.

Your setup sounds very inefficient. You should be able to track hundreds or even thousands of symbols in real-time, tick-by-tick on a single fast PC.

Just hope you generate enough profit to cover your electricity bill!
 
jmreeve said:
Your setup sounds very inefficient. You should be able to track hundreds or even thousands of symbols in real-time, tick-by-tick on a single fast PC.

Just hope you generate enough profit to cover your electricity bill!

What happens during a power cut?
 
Hi
System is protected against it by a chain (cascade) of two UPS (Uninterruptable Power Supply) units, each is capable of handling 3000 Watts. The software is also able to restore its last state in case a computer is "stuck" (not because of power necessarily)
Buky
 
jmreeve said:
Your setup sounds very inefficient. You should be able to track hundreds or even thousands of symbols in real-time, tick-by-tick on a single fast PC.

Just hope you generate enough profit to cover your electricity bill!

Reply from Trady:
On the profit side... last year I made about 120% profit on a base investment of above $150K, This year it seems to be about the same.
On the trading side...what matters are the following factors: (a) good algorithm which implements a good strategy, (b) wide scope and (c) enough budget to buy. In part B of my article I will explain that by choosing dynamically the right symbols and by tracking it tightly (does not matter if it is tick by tick or every few seconds) there's enough data generated to let computer make its own computation and decides whether this symbol is about to raise up or not, and statistically, is that a real price change or a temporary one.
By the way, electricity is not that expensive, communication bandwidth is !!!
Good Night, Have a great weekend
 
trady1 said:
Hi
System is protected against it by a chain (cascade) of two UPS (Uninterruptable Power Supply) units, each is capable of handling 3000 Watts. The software is also able to restore its last state in case a computer is "stuck" (not because of power necessarily)
Buky

I was under the impression that chaining UPS is not recommended because the second UPS presents a complex load to the the first that UPS are not designed to deal with.
 
trady1 said:
Reply from Trady:
By the way, electricity is not that expensive, communication bandwidth is !!!
Don't know about that. 20 MBit/sec ADSL is becoming widely available here in Australia. I assume we are not uniquely blessed in that respect. How much bandwidth do you need ?

I think we can safely assume that going into the future energy prices will rise much more than bandwidth costs which will probably fall.
 
jmreeve said:
Your setup sounds very inefficient. You should be able to track hundreds or even thousands of symbols in real-time, tick-by-tick on a single fast PC.

Just hope you generate enough profit to cover your electricity bill!

I have to (mostly) agree. I'm sure it's quite possible to soak up any amount of processing power, but I'd really want to know why and no indication is given as to why a cluster is needed. As you say tracking hundreds of instruments tick by tick is very possible on an ordinary PC.

If I was looking to set up some high performance system with a lot of crunch power, I would need to be seriously convinced of the need for a small cluster. With something like AMD's 4x4 setup becoming available and reportedly the ability plug in two quad core CPUs in the not too far distant future, a LOT of processing power is becoming available at 'retail' prices. Cheaper, simpler software environment, more reliable hardware environment, much cheaper to run and house.
 
Chain of UPS

dcraig1 said:
I was under the impression that chaining UPS is not recommended because the second UPS presents a complex load to the the first that UPS are not designed to deal with.

That's true
Basically a cheap ups (what you can buy in Office Depot in $200 US) generates a square wave signal that can be used by a single computer (or in some cases two). As you may probably know a square wave is a sum of multiple frequencies sine wave, one of them, the dominate one, is at the frequency of the electricity line (50/60 HZ), but if you load that UPS too much there will be a wave shape distortion and the 50/60 HZ frequency will become less dominate, therefore the power will fall down. This is in few words without diving into electronics engineering issues. However, if you use a real sine wave UPS, that by the way costs $2000 and above, per unit, then each UPS is feeds and provides the next item in line with the same wave, power and voltage as electricity company does. In many critical systems a chain of UPS (a real sine wave unit) is used.

Buky
 
bandwidth

dcraig1 said:
Don't know about that. 20 MBit/sec ADSL is becoming widely available here in Australia. I assume we are not uniquely blessed in that respect. How much bandwidth do you need ?

I think we can safely assume that going into the future energy prices will rise much more than bandwidth costs which will probably fall.


Hi
I dont use ADSL or Cables, but a direct connection of 1.5Mbps. The cost of energies and cost of communication is an issue, but it is a sub one relative to the real issue, which symbol to buy, when, how long to hold it and how to sell it, isn't it ? Anyhow, we got into that subject because we raised the issue of the sampling rate. In part B of my article I'm going to raise this subject and to explain what and why my system does what it does in order track a symbol and still be able to make the heavy calculations which are required in order to decide - on the fly - whether to buy a symbol or not. Thank you

Buky (Trady1)
 
*JDR* said:
Do you mean "human being" ?

Or are you suggesting that a human has been something else ?

JDR

I'm not sure what are you exactly reffering to but I guess - without having the context in front of me - that I meant to "Human Been", anyway, if my English is a problem we can switch to binary... kidding

Thanks

Trady1
 
Top