Backtesting is sooo confusing , anyone got a good guide to learn how to do one?
- Thread starter Wach80
- Start date
NinjaTrader has a "Strategy Wizard" that will allow you to create your strategies without any programming. You just enter the conditions you want and it creates the code for you.
Ok we have two types of backtesting , the algorithmic one and the manual one.
The manual backtest can analyze pretty much everything you can imagine and its also very straightforward and easy but it takes alot of time especially if you are gonna backtest huge past data. I understand this kind of backtest fairly well.
The algorithmic backtest is a huge pain to deal with , it requires programming to express your ideas and also requires understanding of parameter values and stuff like that. I really dont know where to begin with this kind of backtest , i dont know what ideas to backtest , how to express them , i dont know what values to use to express how i imagine the chart , i really know nothing and my brain gonna melt. How can i say for ex *volume should be that big* , *chart should be on a trend range* , *open interest should be like that* , *buy when price is on x level close when is on x level* , i dont even know if thats the best things to backtest for , really really troublesome.
The manual backtest can analyze pretty much everything you can imagine and its also very straightforward and easy but it takes alot of time especially if you are gonna backtest huge past data. I understand this kind of backtest fairly well.
The algorithmic backtest is a huge pain to deal with , it requires programming to express your ideas and also requires understanding of parameter values and stuff like that. I really dont know where to begin with this kind of backtest , i dont know what ideas to backtest , how to express them , i dont know what values to use to express how i imagine the chart , i really know nothing and my brain gonna melt. How can i say for ex *volume should be that big* , *chart should be on a trend range* , *open interest should be like that* , *buy when price is on x level close when is on x level* , i dont even know if thats the best things to backtest for , really really troublesome.
Its simple, concentrate on what you know and start backtesting. You'll then realize you're not confused at all.
In fact, look at the start date of this thread and then sit down a month from now to tell yourself how much "manual backtesting" have you done or how much time have you spent doing nothing but talk about things you don't fairly understand well.
Start your manual backtesting and start learning...stop wasting time.
Well said.
What I've done in the past for people that can't seem to reduce to writing what's in their heads is to do what the eye doctor does. Show them charts and ask them whether their criterion are met. By doing this with enough carefully chosen charts, a patient coder also skilled in trading should be able to ascertain the written criterion.
It is possible you can do this to yourself. Look at two charts, one that meets your definition of a trend, for example, and one that doesn't. Then, simply describe why one does, and why one doesn't. Eventually, you should have your written rule. You'll save a lot of money doing this yourself rather than paying someone to play 20 questions with you.
Well said.
What I've done in the past for people that can't seem to reduce to writing what's in their heads is to do what the eye doctor does. Show them charts and ask them whether their criterion are met. By doing this with enough carefully chosen charts, a patient coder also skilled in trading should be able to ascertain the written criterion.
It is possible you can do this to yourself. Look at two charts, one that meets your definition of a trend, for example, and one that doesn't. Then, simply describe why one does, and why one doesn't. Eventually, you should have your written rule. You'll save a lot of money doing this yourself rather than paying someone to play 20 questions with you.
I wonder why there isnt any software where you can pinpoint a pattern and buy/sell point on a chart and it would translate that into code automatically , i dunno if thats possible.
because some stuff are objective and some are subjective .there are lots of exceptions which itself generates another exceptions too .at the moment it's impossible .
Good point, but it can depend.
It can depend on the discretionary trader. Some may use the same clear rules; but just trade their mental algo manually. These traders could use such an app if they can't code.
Others may go on 'feelings' etc. sometimes.
Such an app would only work for traders with clear rules...but...such traders wouldn't need the app because they can simply reduce their clear rules to writing.
It seems there may be a market for those wanting to learn to backtest
:http://www.smbtraining.com/blog/systems-foundation
I see this is under the Currency/Forex category, but I assume this is applicable to everyone still...
Don't overanalyze things in backtesting/paper trading/studying the market;
Learn to somewhat...let loose, lose control -- yet still have an open, controlling mind
Don't overanalyze things in backtesting/paper trading/studying the market;
Learn to somewhat...let loose, lose control -- yet still have an open, controlling mind