I have no idea what that guy is talking about with double moving averages and what not, but I'd bet that a few simple IF/THEN statements could have produced similar results over the same period of time. Say it hand done well, would you go an invest a few thousand dollars using the IF/THEN algorithm?<p>On a similar note, say this guy did lose his $3500. Do you think you'd be reading this post on how to automate day trading? Randomness and the survivorship bias play a huge, huge role here.<p>As far as the website ideas go, I'm pretty sure that some very smart people have demonstrated that using past performance to estimate future performance simply does not work; no amount of tweaking buy/sell algorithms will help you predict where the market is going tomorrow.
I'm the CTO of a startup named QuantRunner Software, where we write backtesting software, (beta coming this month) <a href="http://quantrunner.com" rel="nofollow">http://quantrunner.com</a> . I found the article interesting, but it leaves out some important details such as<p>1. Why did he decide to enter into this strategy at 10:45 am on October 15th?
2. What would the outcome of the same strategy have been if it was left running for the next month, for the past year, for the past 10 years.<p>I'm curious.<p>Questions about how the backtest was run.<p>1.How do you get your data?
2.you seem to have written an excel spreadsheet to run the backtest, what's going on with that? How did you design it?<p>For other people on HN, does anyone here use backtesting tools? What do you think of the existing software you use?
I've been thinking recently about how it would be cool to have a service which allowed you to try different trading algorithms.<p>For instance, it would have a few algorithms built in, you could choose which you want and with what parameters and then it would monitor how it performs. Also available would be the ability to see how this algorithm would have performed in the past (to see how it worked in bad markets, etc).<p>It could also have a generic API to allow you to run your own algorithm on your end and simply make calls to BUY and SELL certain things.<p>(In the future it could even allow _real_ trades, etc. One other cool way to make money off this would be to actually invest in the best performing strategies as they occur over time.)<p>I'm guessing this is all available to the big firms, but haven't ever heard of anything geared to the hackers who want to try things out.<p>Does anyone know if this sort of thing exists? Would anyone else think this is neat?
If you thought something were profitable in a market, and it was easy enough to implement that any hedge fund could, why would you publish it? They'd just do it too until it was no longer profitable.
So, the obvious question for me is where can I buy stocks for $2.40 per trade? Any day trading successes I might manage would be eaten up by transaction costs at my current broker (where trades are $9.95, I think). I'd probably be a far more active trader at $2.40 (which may not be a good thing, but cutting costs is always good).
People have been using computers to trade since the 80s. It is estimated that in some markets (Forex, Commodities), more than 50% of the orders are put in by machines.
Heh, so this has piqued my interest. I have absolutely no experience in online investing/day trading, although I'm pretty sure a lot of you are veterans.<p>Any books you'd recommend to a programmer with day trading aspirations?
And lastly:
I wrote a lot of that just as I was figuring things out, so I don't know how much of that I still stand by. But this morning, I wrote the one thing that I'd recommend to anyone who is interested in this sort of thing.<p><a href="http://fattyfatfat.com/2008/12/moving-average-crossover-theory/" rel="nofollow">http://fattyfatfat.com/2008/12/moving-average-crossover-theo...</a>
A year ago I embarked on a project to learn Common Lisp by building a trading system, to execute with the Interactive Brokers API.<p>I modeled the woodiescciclub.com trading strategies, primarily because the math is simple and the patterns are well defined. However since then I've seen other parties on the web talk about how the Woodies patterns actually don't make money consistently, although I didn't corroborate this.<p>I actually found CL to be a fantastic language for this. I was writing up the Woodies patterns using a DSL I had created:<p><pre><code> (defpattern zlr "Zero Line Reject" :with-trend
((count 1) (within 100 250) (trend :up :or-opposite-invert))
((count 1) (within 100 200) (trend :up :or-opposite-invert))
((count 0 10) (within -50 99) (trend :up :or-opposite-invert))
((count 1) (within -50 50) (trend :up :or-opposite-warn) (angle :down))
((count 1) (within 0 115) (trend :up) (angle :up) (away 8)))
(defpattern hfe "Hook From Extreme" :counter-trend
((count 1) (within 200 300) (trend :up))
((count 1) (within 0 199) (trend :up) (angle :down) (away 3)))
...etc
</code></pre>
The defpattern macro was creating, for example, #'is-zlr, #'is-zlr-up, #'is-zlr-down, which were then used by a trading engine that monitored 5-minute bars on a given futures contract and launched trades.<p>Another CL feature, restarts in the condition system, helped too. When the data feed broke or got corrupt, an exception thrown which could reload the data feed and restart back in the listening loop without forgetting any context was helpful.<p>Also just running in a Lisp image was great because I could swap functions during a live feed without stopping the program.<p>In the end, the programming worked great, but the trading system lost money. I had it working in backtesting over a few months and then papertrade mode for weeks, and then when I took it live, the volatility rose and it started hitting the stop all the time. I adjusted things like stops based on ATR, but I really just ran out of my little budget to play with it.<p>It did however teach me CL. ;-)<p>Modern trading platforms have built-in tools and programming languages and writing things from scratch in CL is really just an academic exercise.<p>I believe it's possible to do algorithmic daytrading, but quite a bit more difficult than it seems. There are way more market behavior modes than you can appreciate at first, and the keys to success lie more in counterintuitive aspects of your system like position sizing rather than intuitive aspects like entry strategy.<p>The best talk I heard from someone on the subject indicated that it takes years and the end result is a fairly sophisticated algorithm-switching engine, because the hard part is just to survive.
Like many hackers, I often pondered the idea. The closest I got was automated trading in a stock market game run by the company I was working for. I never got round to programming a sophisticated strategy, so basically it was just "sell and buy random new stock if price has risen or fallen more than x%", where I think x was something like 15%. Even with that the algorithm was in the middle of the field (with 10000 participants), but it still lost a little money. As it was 2001, losing money was to be expected, I guess.
um well, hello. so thats my website. quite the site here. and quite the surprise to see that i'm getting any attention at all. a lot of very intelligent commentary.
1. Not having a stop loss (other than the 3500 circuit breaker) is a terrible idea in black box trading.<p>2. Yes, it is that simple. Note that simple != easy.