How long before you were fully automated?
- Thread starter travis
- Start date
The above website you mentioned states:
"This ultimately leads us to be able to test the hypotheses with specific data -- a confirmation (or not) of our original theories."
This is false. A hypothesis can be in most cases only falsified. It can never be confirmed. Can you ever confirm that "men are mortal? You will have to test all existing and future men for mortality.
Furthermore, a theory that generates a hypothesis can never be confirmed buy confirming the hypothesis for the same reason. The theory can be only falsified.
By the way, this leads to the doctrine of pessimistic meta-induction. If all previous theories were falsified, all future theories will be falsified. This can appply to trading systems. If many trading systems failed in the past, all trading system will fail in the future with high probability.
The website makes a fundamental error.
http://search.barnesandnoble.com/Quantitative-Trading/Ernie-Chan/e/9780470284889
Chapter 8 of Quantitative Trading: How to Build Your Own Algorithmic Trading Business by Ernie Chan just raised and answered a question I had always been wondering about: how does an independent trader with insignificant equity and minimal infrastructure trade with high Sharpe ratio while firms with all-star teams fail spectacularly?
[...]
...The key, it turns out, is capacity, a concept I introduced at the end of Chapter 2. (To recap: Capacity is the amount of equity a strategy can generate good returns on.) It is far, far easier to generate a high Sharpe ratio trading a $100,000 account than a $100 million account...
He also mentions a few more potential causes:
1) we're trading our money while they aren't
2) they have more restrictions, regulations, etc.
3) their boss may be incompetent and may force them, among other things, to stop their trading after a drawdown and increase the investment after a win.
---
Another important point he makes about "capacity", at the end of the same chapter, is the following:
Using the Kelly formula, you can indeed achieve exponential growth of your equity, but only up to the total capacity of your strategies. After that, the source of growth has to come from increasing the number of strategies.
This guy is a scientist, yet he talks clearly and with simple concepts and words and gets to the point. I really appreciate the way he writes. Thank you, Ernie.
http://www.epchan.com/bio.html
Chapter 8 of Quantitative Trading: How to Build Your Own Algorithmic Trading Business by Ernie Chan just raised and answered a question I had always been wondering about: how does an independent trader with insignificant equity and minimal infrastructure trade with high Sharpe ratio while firms with all-star teams fail spectacularly?
[...]
...The key, it turns out, is capacity, a concept I introduced at the end of Chapter 2. (To recap: Capacity is the amount of equity a strategy can generate good returns on.) It is far, far easier to generate a high Sharpe ratio trading a $100,000 account than a $100 million account...
He also mentions a few more potential causes:
1) we're trading our money while they aren't
2) they have more restrictions, regulations, etc.
3) their boss may be incompetent and may force them, among other things, to stop their trading after a drawdown and increase the investment after a win.
---
Another important point he makes about "capacity", at the end of the same chapter, is the following:
Using the Kelly formula, you can indeed achieve exponential growth of your equity, but only up to the total capacity of your strategies. After that, the source of growth has to come from increasing the number of strategies.
This guy is a scientist, yet he talks clearly and with simple concepts and words and gets to the point. I really appreciate the way he writes. Thank you, Ernie.
http://www.epchan.com/bio.html
I have no idea what you are talking about... I'm only an Amish trader who makes money through impractical reasoning.
And of course, I am now a fundamental error.
So... I'm an Fundamentally flawed Amish who trades...
Very nice!!!!
Capacity can be handled in terms of market capacity.
Handling Kelly capacity is done by the use of deductive coding of the ATS.
Attached is the market capacity ball park.
To move out of the Kelly ball park, the thing to do is review how to build the "carrier" of the ATS coding. Theories are used.
SCT is limited by several things: it is intraday and limited to 10,000 contracts per account, you are limited to 500 dollars a contract and partial fills are limited at 500 contracts, only 20 to 40 trades a day can be done using hold/ reversal.
Position tradig stocks is limited too: it is interday and only 12 to 15 streams of capital are prectical, 100,000 share is the limit and partial fills are at 3 to 5000 shares, a turn takes 3 to 5 days.
Stock Sector Rotation is unlimited in streams but is limited to 10 million shares and is interweek oriented, partial fills are at 500K shares a day and a turn is 4 to 5 weeks.
Sweeping is done weekly as shown and building accounts is nottes by > and >>. > is the normal adding of profits and >> is a dwell level required to aclimated to the rate of change of capital of profit taking.
Eliminating the Kelly phenomena is done by switching from induction to deduction. Probability has to be taken out of the picture.
Think of your coding of an ATS as beginning with two streams of operations: one is data handling and shaping and the other is building the "carrier" of the programming deductive bells and whistles.
One stream follows TSGannGalt steps of processing data into a holding area. Don't go too far into the display since the display has no signals generated from it. Just code up what you want interms of panels that show the market "tells".
Start the other stream with a theory based set of coding. A Paradigm emerges. It becomes the "carrier" of the deductive element outputs which need to be processed by the paradigm.
regarding your Kelly comment and the requirement suggested that all hypothesis must be null hypothesis, we have the contributed beginning point.
I chose two given to me by the market structure and process. these two hypothesis cover the waterfront. At any time one is true and the other false. This is because of their parametric measure: VELOCITY. To hone the measure it is just the sign of the velocity.
This means that I am processing, deductively, information from the hypothesis that is binary, mutually exclisive and in the form of vectors for the market variables.
What this means is that I have no probability involved and, as a consequence, Kelly has dropped out of the picture.
But I do need three ATS's for the capacity of the markets reasons shown in the illustration.
For me nothing is going on most of the time. That is there are no coding requirements to deal with as time passes. This is the interval of price change where profit segments are accumulating capital. The carrier conveys this to me and to the execution platform. Speaking in time terms, bar after bar, nothing happens that causes any action on the execution platform.
I have two variables that each have two parametric measures. This, in coding, is a matrix that has four different values to consider. Further, one variable determines the other variable.
If conditions change, then I will have to take an action to take profits and begin another profit taking segment.
The coding of an ATS comes down to making the appropriate timing of actions appear on the execution platform. Some of these functions are as simple as doing a series of partial fills in a very timely manner: it is a tattoo often seen on the T&S so I know it is familiar to all.
If anyone reading this is a programmer, I know that you can write the above in an hour or so. Or you can go and do a drag and drop on Worden Bros or use Multicharts to knock it out.
Another similar example is roughtrader doing ES. He is using the top box. Starting with one contract on 25MAR09 and up to last Friday (a month) he has added 20,000 dollars profit and is adding 1 contract every 5,000 dollars of profit. By 25JUN09 he will have added 100,000 dollars for a quarter of a year's work using an ATS. In his forum you see several bells and whistles added to the carrier. The next quarter he will go to a million dollars on 25SEP09. this will be lessened somewhat by filing estimated taxes periodically and peeling off some cash. Most of his money, however, is always idle.
Remember the carrier rule is simple and it says: "stay on the right side of the market" Stay on the right side of the market is defined by two variables V>>>P, where the measured parameter is the sign of the velocity of V giving P by the sign of the velocity and by two mutually exclusive non probalistic hypothesis.
Adding about 15 bells and whistles achieves one thing. More effectivenss and efficiency. I deduced these. And when added they lead to other deductions in real time as time passes.
I magine a manual trader. He sees a condition on his screen. If that condition is there, "What Must Come Next? Two hypothesis tell him by their parametric measure of velocity.
Look at the whistle called Pro Rata Volume (PRV). It enables the manual trader to see the sign of the velocity of V in 12 seconds after a new bar begins. Slipping this inot an ATS is fun to do once you have the carrier working; it adds "anticipation" to coding an ATS. It is nice to know a change is coming up in the trend or "sentiment" stated by the sign of the P velocity. Or a different "sentiment" that expreesses "continuation" or "change" of a price trend vector.
What I am giving you is the answers to three important trading questions that all "complete" ATS's answer:
1. Where is price in its cycle,
2. What comes next in the cycle, and
3. How fast is the cycle changing?
To see a lot of bells and whistles go to the thread here on the Hershey coding of things. You can also go to Fidelity's coding site where many many scripts and snippets under my name appear.
This is actually where most stop, is at the worst possible time, just as with any losing trader. If it has positive expectancy, and happens to go through a drawdown, does that mean the system has stopped working? The answer is no in the situation where there is high statistical economic significance and expectancy. Kudos, businessman.
jack,
Hrrrmmm... Now, I have a few things that differ from you.
1. I don't display what the market "tells". I display what the market "told" the models. I deal with the market during off hours on a seperated platform.
Why would you need to see the market, if it doesn't give you the bells and whistle, considering that it is coming from the models already?
2. Your PRV applies under a theory that the market is under a single cycle. Understandable and accepted. Considering the theory being based on cycles, you mention about including an anticipated periodicity and amplitude "detector". That too is understandable and accepted. You example is fits perfectly for a theoretical point of view. But then... before the theory is implemented, it needs to be observed and confirmed that it is true.
So now, I have 50+ theories about the market which I have tested and confirmed to be true, thus having 50+ models. Within those 50+ theories that I have, only 23(as of this writing) of the theories are actively profitable from tested observation.
You have mentioned, "he has just thee problems he does not know whre he is, what is next of how fast things are changing. the reason is he uses induction of raw data the same unprocessed data and all the time. The same old six degrees of freedom with the same types of detectors that do not change any degrees of freedom. No steer and no focus."
If I didn't know where I am at, then I wouldn't have more than 1/2 of the models inactive. They are inactive due to the theory that models have a cyclical nature where I am only trading them when the phase of monitered tendency is in a particular state.
If I didn't know how frequent a phase would change, I wouldn't waste valuable computer power to follow them. I anticipate a change in the phase so it's being monitored. This can be said with what will be next.
Finally, in terms of the degree of freedom. They all apply different theories of the market. On top of the models generating whistles and bells, I have 3 portfolio models (which I have a theory and tested) that moniters how the models should be trading.
3. You use a static capital allocation. And I am sure that it will eliminate the exponential growth and risks involved. You did, in fact mention, that statistics is not considered. Though, how does one change the picture when the market changes and the theories are within a non-profitable phase? Deductively, position sizing, risk allocation and portfolio management will be condensed and merged, if you have had some experience trading automated models.
4. SCT and PRV are based on subjective probability. They are deductive from a pseudo-scientific sense of mind. I'm a sole believer of "TEST EVERYTHING" protocol.
Please don't take my text as a proposition because I'm not going to bother with it because I like to keep things readible... but that's my take on the discussed issue so far. Surprisingly, I've gotten accustomed to jack's writing. In terms of general approach of models and development, I agree... I still have a lot of studying and catching up to do. I appreciate his thoughts for it. Though, when it comes to practical / applied "trading", I disagree. But he can always correct me. So no bad feelings. Fixing a bad thing is much easier than improving on a good thing.