darthtrader2.0

Blowfish, any chance you can do a review of this book? Maybe point to parts that you found interesting? I'm not sure what it is about this book that it defeats me without mercy.

Nucleaphynance has some good stuff too, less students asking questions than Wilmott but people are a bit more terse in their responses on there. The real problem though is I don't think your going to find what your looking for exactly. If there is an advantage to auto trading its removing uncertainty from the strategy...because of that I pretty much operate from the perspective that if what is being talked about is very specific to the point you can impliment the idea in code, then it probly doesn't work. The pairs trading book in the review section here is still the best quant intro i've found...keep in mind that the only reason it was probly published is because its 101 stuff that doesn't work anymore so it was safe to publish. It at least though gives a glimpse into the style of that kind of trading though with specifics.

No its just a crystalin powder that disolves easily in water. 1 gram is only 1/4th a teaspoon so its not hard to take a few grams a day. You probly could gel cap it but I would compare it to coffee or beer taste wise. The first time I tasted coffee and beer I thought they were the worst things I ever tasted. Actually, mixing it with alcohol would be a really really bad idea. I've read a few reports that people that have drank after taking it had a massive hang over and it magnifies the effects of alcohol with more blood going to your brain.

I'm not sure I totally agree with this. First, by "elaborate" aren't you simply refering to volatility if your only looking at price? I guess there is another factor in there that I do agree with as far as the participants motivations for trading the instrument. Q's have more volatility because of the arbitrage players, ect...I'm not sure I agree though that this correlates much with liquidity. I think you have to take into consideration that the purpose of a market is to price an instrument forward and the real reason that Q's have more volatility than ebay is because its fundamentally harder to price Q's forward than ebay. The real problem though is if you compare this idea between corn futures, eur/usd in forex and a small cap equity X on the AMEX. eur/usd is far more deeply liquid but you don't see crazy volatility since the variables trying to be priced are so huge and the rate of information that effects the participants trading is slower. Are corn futures harder to trade than equity X? How would you order these 3?

I've been taking this stuff for a month and can't give it enough praise. http://en.wikipedia.org/wiki/Piracetam It use to be prescription only but now can be purchased as a supplement. I got mine at bulknutrition.com. Its a nooptropic/smart drug that basically increases the blood flow to your brain and sopposedly improves communication between hemispheres of the brain. Qualitatively it feels like a good cup of coffee but without the stimulant effects, but with more concentration as opposed to fast scatter brain thoughts. The taste is pretty terrible when you start off but I've already aquired the taste that I look forward to it. Oxiracetam is a more powerfull , long lasting version that I'm going to switch to when the Piracetam starts running low.

To me this question is simple in that you can give someone a profitable system but they might not have the proper mental framework in order to trade it profitably themselves. On the flip side you can give Sigmond Freud an unprofitable system and it matters not at all what mental framework is being used, if the system is not profitable its not going to make money. The benefits of a simple system are that its easier to test and its easier to trade in the exact same fashion over and over. The downside is that its not going to be as robust as something more complex. I wouldn't doubt alot of intermediate traders atribute "pyschological problems" to their simple system when it starts to break down during a market regime change when in reality the real problem is that the system is starting to lose its edge. I personally believe the whole concept of the "holy grail" is a very poor one when it comes to trading. My personal goal is to find as many non random exploitable market setups/conditions as possible in order to add robustness to the overall portfolio of strategies. To me the idea of the "quest for the holy grail" makes one intentionally diminish robustness so that you can't be accused of searching for the "holy grail", it makes no sense. The biggest problem that I've seen with new traders is simple lack of patience at all levels. No patience to do hard work, no patience to stick around to see a few market cycles, no patience to sift through all the garbage the trading community produces, no patience to learn to program..everyone wants to jump right to proving they are the next Tudor Jones and never take the time to even give themselves a chance.

I'm a TA converted fundamental guy. To me there are two paths you can take...get really deep into aswath damodaran stuff or stick with old Ben Graham...The problem with Damodaran type stuff is that your going to have to make big speculations on interest rates and expected returns in your model. The bigger problem is that it would be a huge project to be able to automate this valuation process. You have to assume that the market is efficient enough that if you randomly pick a stock it will not be undervalued. Even if you find "gold" most the time it will be fools gold because your directly betting that your analysis is better than well paid teams of analyst who's job it is to understand all the variables. The most likely bet is that your missing a very important variable. Basically, I think fundamental analysis is the most difficult type of analysis you can do yet in the retail literature there is this air of it being easy. I think Buffet does a huge disservice to the investor community with his ideas. Its like if Tiger Woods talked about how to golf by saying "I just hit the ball". While thats totally true the big hidden variable there is "you aint Tiger Woods".

Also, not to lead people astray with posting this webinar... If you are not interested in whats listed in the financial toolbox, the various time series analysis tools, various data mining algorithms, ect...then Matlab might just be total overkill for the job. If you just want to do crazy stuff with indicators you would probly be far better off with something like Neoticker or Ninja... For the guys that are interested in Matlab, I would be interested to know what you are doing/looking to do with it that other retail trading software can not? For me this is a 10 year+ project with the eventual goal of creating a fully bayesian auto trader..data mining high frequency data with SVM and bayesian networks and then using bayesian inference to place bets. I decided this week that I want my setup to eventually evolve to Matlab, R, HDF5 and Reuters as the datafeed....I'll figure out a broker when the time comes. The big hurdle is understanding bayesian inference and SVM for which im just sticking for the next few years with R in order to learn. ("Bayesian Computation with R is a sick sick text)

I agree, I don't really get it either. While I voted yes, I don't really play poker but it would be interesting. I guess I still think though it would be better to have a non trading speculation section to keep some interesting things from getting lost in the "generalness" of the general section.

Assuming you want to stay away from your datafeeds API, you might be able to get this going if your datafeed can do DDE and assuming matlab can write to mysql, then you could probly do something like what Hlm posted above. I think the bigger question though is if you need to access the database to trade live or just to backtest ideas. If just for backtesting the easiest thing to do might just be to spit out text files of the data from ninja, then parse the text with python or something and shove it into the database at the end of the day. I would think that would be pretty simple to figure out. From searching around this morning I think thats what I'm going to start off with. The more complex issues of datafeed -> database -> matlab -> trade signal/order, well I'll cross that bridge when I come to it.

This to me is why its a shame that no software has a paper trader that randomly switches you between live trading and the simulator at some user specified %, but you don't know until after the trade if it was real or not. To me that would be the ideal way to do things starting off because there is the dichotomy between getting the practice/sample size and your account blowing out starting out. Maybe you could try a strict regimine of 1 live trade then 1 in the simulator until the gap closes between the two? Or maybe try taking your average number of trades over a certain time period and divide by 2 or 4 and that is the number of live trades you have to make and your done, abandon the simulator.

Well I should say I know virtually nothing about database design and what not. Is that the whole issue with SQL vs a time series database, single file/table size? While I could see the problems that would arrise if you dumped years of tick data into a single table that would obviously be a really dumb way to go about things considering that with financial time series we have the benefit of sectioning things off by the day. I'll have to try MySql out, I soppose the big advantage would be the learning material available. Maybe at some point we can try a database design thread on here.

Thanks Hlm, great info. I will have to try to get the DDE stuff going with dtn and see how that works, should be a nice learning experience in and of itself. For database stuff have you ever checked out HDF5? I know thats some heavy duty stuff for large files and matlab can interface directly with it: http://www.mathworks.com/access/helpdesk/help/techdoc/index.html?/access/helpdesk/help/techdoc/ref/hdf5.html I read on one of the quant boards about a guy who mentioned he was mining around 1 terrabyte datasets with it. My only concern with Mysql is that things might get overwhelming size wise rather quickly if i want to store tick data across multiple instruments. My interest here is in a very long term project for high frequency stuff that will take years to get sorted out. HDF5 might be total over kill though, i don't know. MidKnight, I kind of figured neoticker could do multiple strategies...I guess I've really never read something where neoticker couldn't do "X". There is just something about the program that bugs me though, this visual basic program for win95 feel to it.

Ahh great info Hlm. Could you describe your setup and process in more detail? Do you analize and store tick data? If you are storing things what DB are you using? Do you use multiple strategies? Thats one thing that seems nice with matlab and absurd that so much retail autotraders can't handle more than one strategy at once.

I found this too http://www.matlabtrader.com/ It has stuff to get matlab hooked up to interactive brokers. Note the full matlab version is $5000 with the toolboxes you would want. The student version is 99 bucks. Pretty sweet deal to get your feet wet it would seem. Here is the datafeed toolbox, pretty lame that it jumps from yahoo to bloomberg, lol. http://www.mathworks.com/products/datafeed/description3.html anyone ever mess with the esignal or dtn API?

I just watched this yesterday: http://www.mathworks.com/company/events/webinars/wbnr30376.html?id=30376&p1=50647&p2=50649 I was always curious as to why someone would use Matlab over a more specific trading software but after watching that video I have to say I'm completely floored. They use a simple moving average system as an example and the visualization tools are light years beyond anything else I've ever seen. The one backtest shows this color coded 3d plot of how the different MA parameters effect a sharpe ratio...just amazing. The only downside I can see with matlab is the price of the software and the datafeeds that connect to it are somewhat limited to the higher end of things(reuters, bloomberg). note:you do have to fill out a brief questionaire, it would probly be worth it to just create an account at the site as they have a few other webinars for finance that look pretty interesting.

Anyone else listen to any podcasts related to these subjects? I'm not much for TV or radio so when I'm doing stuff around the house I usually have some podcast going. The 3 I listen to on a regular basis is: http://www.traderinterviews.com Pretty short interviews with mostly retail guys, mostly trading forex. http://www.financialsense.com pretty doom and gloom gold bug focused but the its 3 hours long and they get alot of cool interviews in the second hour. http://www.econtalk.org Pretty straight economics/philosophy talk, can be pretty dull but its good to have on while cleaning...

Actually, could you explain how you would do this? I don't quite understand the mechanics of how that would be possible. While I mostly sit in the camp that the order book should be treated as noise and ignored there is an interesting thread on elite trader about accumulating all 5 levels of the DOM on each side, building an indicator from it and using it for entry. While it wouldn't be the meat of the trade or occur all the time I would think there would be some utility at a certain level of imbalance as far as getting better trade location on entry or picking up a few ticks at the exit.

I have to agree with MC. While I would be quite interested in a discussion its hard to not have some skepticism with what you posted about Time Series Analysis, argueably the most studied and well documented aspect of quantitative finance. Time series analysis is hardly rocket science, pun intended. Of course gotta have some neural networks, traders love those neural networks even though they have been largely abandon for more robust techniques in the data mining world like kernel methods and Bayesian networks. But hey, it makes for good marketting since everyone has a neural net in their head therefore they believe they can naturally understand the algorithm.If you used a support vector machine no one would buy it because they would inherently know they know nothing about support vector machines. The bigger problem with your post is if you have developed your variations on known techniques then what possible discussion can we have? Ignoring Blowfish's questions doesn't really remove some of the shadyness.

Cool information Blowfish... The only thing I can really see with the book that seems usefull is accumulated depth of market, just to make sure your not entering in on the wrong side of a stacked book. I haven't done much there but I would think because of liquidity its part of the markets nature to trade to size at that level. Even there though, I don't think we are fast enough at this level to deal with the book using anything other than tape reading techniques on the DOM. Even then, thats a whole lot of information/patterns to have to train your brain on and might do more harm than good for a very long time.

What about the poker section being under another header of something like "general uncertainty that is not specific to trading"? Then make the general section essentially a place for posting things off topic. Stuff thats just for socialization/community chat. Just an example, I found this book that sounds great http://www.amazon.com/Cambridge-Handbook-Expertise-Expert-Performance/dp/0521600812/ref=sr_1_1?ie=UTF8&s=books&qid=1215604490&sr=1-1 and this article http://www.scienceblog.com/cms/mysterious-neural-noise-actually-primes-brain-for-peak-performance-11980.html Right now, those would probly go in the general section but quickly get lost for future reference with the sports/pets/car talk.

I don't see how the magnitude of the pnl of a system is all that meaningfull, its somewhat of a number without context. Since the system seems to be working nice over the past year it obviously enjoys volatility. I would test it against highly correlated instruments over the past year just to make sure that you didn't just get lucky with the data you picked to test on and that you are truely capturing something that profits from volatility. Then test it against data from a time period with lower volatility. Even if the system doesn't work as well you can just add filters to make sure it only trades when there is enough volatility for the system to work. VIX > X is a nice simple tool for this. The most dangerous thing I would think you want to protect against is that everything looks good in the rear view mirror, volatility then dries up and your system is then trading market conditions that it simply can't perform in.

The South Sea bubble is kind of cool because there are some charts of it. Its a pretty boring story though compared to the Tulip mania. One interesting part is that Isaac Newton got caught up in it..I believe he sold early with a profit but then jumped back in and got killed. "Isaac Newton lost over 20,000 pounds of his fortune. As a result of this crisis, he stated “I can calculate the motions of heavenly bodies, but not the madness of people”."

I do think the problem is that Dr. Brett is speaking to a few different auidences in his writings. Obviously, finding an edge and excuting a legit edge for profit are two very different things. To me the best excerise you can possibly do to gain confidence in your edge is to try to automate what exactly it is you are doing. Even if you have no interest in auto trading it will sharpen your thoughts and clear out alot of muddled thinking in your trading. The setup is the easy thing to define, the hard part is entry, exits, trade management....If you normally only get in after a pullback, then how do you precisely define what a pullback is? If you don't have that defined then your introducing randomness into your trading. I think this is another cause of pyschological problems because then your probly not doing the exact same thing on each trade.

Are you talking automated or discretionary though? For discretionary, I would think the key is to simply keep very good metrics on your performance and also judge your trading against some measure of volatility like the VIX. Who knows, maybe there is nothing wrong with you method but you either get scared with a jump in volatility or the method fails from lack of volatility. How exactly you would do that though I'm not sure. For automated I just don't see anyway around monte carlo simulations against either detrended and/or "de-volatilized" data. If your just doing historic backtest with optimizations I don't see how you can get around the fact that you will never know if you have an edge or if your simply curve fitting to the trend/lack of trend, volatility/lack of volatility that is present in the data your testing against. Even if your not using optimization, you can't know if your system has a real edge or if you have just got lucky as far as the situation goes with the data your testing against. You need something to test against thats not the data itself(ie..10,000 random trades from a monte carlo engine)