How feasible is algorithmic trading?

toli said:

Yeah, the key word there is "overoptimization", and I know what that means, and why not to do it.

But in general, how feasible do you guys think this feedback loop is? Take a Pine strategy, tune and tweak it until the past data looks good, run it and make money? To me, it looks good, provided:

  • You keep the strategy simple (that is, no overfit, performs in most market conditions).
  • Well-defined (trade only these timeframes, only those days, only long trades).
  • Not be too greedy (if your goal is 10% a month, do 10% a month and don't try to byte more).
  • Allow for a good percentage of losing trades as long as the winning trades cover it up, e.g. 1:3 risk-reward ratio and 50% winning trades.
  • Never goes down. In the video, you see at some point the strategy stops making money and starts losing money. If that happens, to me that's a sign that it's not robust enough and changing market conditions will topple it. Better less money but more robust to market changes.

What is your experience guys with pine script and algorithmic trading strategies? Do they work for your goals? Do you tweak them often? What do you think of the above thoughts?
More...

Algorithm trading has made many, many billions and is feasible and effective, just not for most ET posters.
 
As VegasDesert said, the markets change and the algos have to be constantly updated via programmers, I know people who designed algorithmic trading, it worked and then degraded, took 6mths to rebuild and degraded again, at that point they decided better to have a 99% system and make the last 1% discretionary.

It was used to setup a crypto fund but they expired it, so far haven't come across anyone who was interested in running a fork due to the algo maintenance problem, a complete drop-in fund launch in a week but someone would have to maintain the algos.
 
nah.


this question is all wrong.


and as is necessary, the answers are all over the place.


for starters, it is important to define what an algorithm is. and then, algorithmic trading has existed for decades and in all probability generates better results for beginners than discretionary trading. that is because when one knows how to effectively evaluate a strategy one knows then and there whether it will work or not. and people developing algorithms will not trade them if they know they are useles. on the other hand, manual trading will never afford such dramatic knowledge.


strategies can be divided into trend following, mean reversion and high frequency trading. they all can be developed and operated with success. all of these strategies are parametric and the first two are heavily dependent on indicators.


now, creating programs that will trade based off of geometrical patterns on charts is another different matter altogether. such kind of algorithms could exist but i have never found a way to make any progress in such kind of programs. and they could be nicely profitable but i have no idea how would anyone create such code and which platforms could execute it.
 
You must log in or register to reply here.
Back
Top