Is backtesting on software strictly necessary to trade?

[asimpleplan]

I've got a basic idea for a strategy using price action and the MACD.

Is it strictly necessary to backtest this strategy using software?

Or is it enough to 'paper trade' for several hundred trades and figure out entry/exit that way?

 

[DrSafari]

I would say option 2 is enough, just much slower. But best is to backtest and then paper trade and then trade small amounts for real and gradually increase your amounts.

 

[forker]

Use forex tester 2 to back test

 

[tomorton]

I only use back-testing to generate the possible market twists and turns that can affect selection of target, entry signals, stop-loss levels etc. I don't put faith in its statistical evidence as to win rate, probable gains etc. More important I suggest is reasoning through the logic behind a set-up, then the implications of modifications to the original criteria.

 

[DrSafari]

I don't put faith in its statistical evidence as to win rate, probable gains etc.

Just out of interest, why not?

 

[dbphoenix]

I've got a basic idea for a strategy using price action and the MACD.

Is it strictly necessary to backtest this strategy using software?

Or is it enough to 'paper trade' for several hundred trades and figure out entry/exit that way?

Computerized backtests are for all practical purposes a waste of time and effort as their results are based entirely on hindsight. But you won't be trading in hindsight; you'll be trading in real time.

Paper trade in real time or delayed or via replay. If you use indicators, you may be surprised how they shift and wiggle in real time.

 

[tomorton]

Just out of interest, why not?

Nothing ever works as well in real time.

 

[piphoe]

if you can define ALL conditions of your setup, AND code them..then YES backtesting is extremely worthwhile.

But thats the rub, those implicit assumptions that we make subconsciously of market behaviour and our setup we are unaware of ..

and those implicit assumptions made but not included in the code can screw up the testing

AND, assuming all that overcome, then WALK IT FORWARD first with paper!

then, walk it forward some more.

 

[forker]

Computerized backtests are for all practical purposes a waste of time and effort as their results are based entirely on hindsight. But you won't be trading in hindsight; you'll be trading in real time.

Paper trade in real time or delayed or via replay. If you use indicators, you may be surprised how they shift and wiggle in real time.

It depends on the quality of data being used. Tick data is always going to give accurate results (outside of spreads). In a good backtester, candles are drawn as though it's happening. You get the standard indicator behavior and unless you repeat the times being tested, you will get a realistic randomness you get on live. To say it's hindsight would only apply to historical charts.

 

[dbphoenix]

It depends on the quality of data being used. Tick data is always going to give accurate results (outside of spreads). In a good backtester, candles are drawn as though it's happening. You get the standard indicator behavior and unless you repeat the times being tested, you will get a realistic randomness you get on live. To say it's hindsight would only apply to historical charts.

There are several assumptions here that require discussion that is somewhat off-topic and in any case has been discussed before in other threads.

Assuming that the question has to do with intraday, candles are inappropriate as there is no close intraday.

As for randomness, price movement is not random.

As for hindsight, that trade that took place a second ago is hindsight.

But, as I said, all this has been discussed before. If the OP wants to know what's going to happen in real time, I suggest that he test his protocols in real time.

Everyone have a nice day

 

[piphoe]

There are several assumptions here that require discussion that is somewhat off-topic and in any case has been discussed before in other threads.

Assuming that the question has to do with intraday, candles are inappropriate as there is no close intraday.

As for randomness, price movement is not random.

As for hindsight, that trade that took place a second ago is hindsight.

But, as I said, all this has been discussed before. If the OP wants to know what's going to happen in real time, I suggest that he test his protocols in real time.

Everyone have a nice day

what you suggest is that candle just before this one thats forming is not as valid as the one that hasn't completed yet? and all those before aren't valid either..

then how do you justify using those "errant" candles (as you call them) to draw your TL's????:whistling

 

[dbphoenix]

what you suggest is that candle just before this one thats forming is not as valid as the one that hasn't completed yet? and all those before aren't valid either..

then how do you justify using those "errant" candles (as you call them) to draw your TL's????:whistling

Price movement is continuous throughout the session. There are, again, no closes. The question of validity doesn't arise.

I don't recall ever using the word "errant" so I don't know what an "errant" candle is. As for the TLs I sometimes use, those are drawn using swing points, i.e., the crests and troughs of waves. If you're interested in how all this works, you're welcome to study my journal.

 

[piphoe]

Price movement is continuous throughout the session. There are, again, no closes. The question of validity doesn't arise.

I don't recall ever using the word "errant" so I don't know what an "errant" candle is. As for the TLs I sometimes use, those are drawn using swing points, i.e., the crests and troughs of waves. If you're interested in how all this works, you're welcome to study my journal.

you don't draw your TL's on candle "opens" & "closes"?

where are they drawn?

 

[forker]

Nobody said their was an intra day close. Just because a specific time frame paints a new bar at specified periods doesn't imply there is a close. It merely means the scope of analysis requires a new bar to be painted. This belief that candle charts in a historical sense doesn't provide meaningful test data is nonsensical. Let me twist this in a different direction to see where it takes this discussion.

If you are trading off a 1 hour chart how does watching it real-time differ from historical prints?

If you are implying that there is useful decision making information such as velocity and time, which is the only additional data that doesn't exist in completed candles, then I agree to an extent of looking at a single time frame.

Velocity and time have an inverse relationship. The less time spent at a price point the greater the volatility. Conversely, the greater the time spent at a price point the less volatility. If looking at an hourly chart you are effectively looking at the measured volatility over an hour.

If this is the only information missing, then looking at a more granularity time frame would give you a clear view of the volatility data you don't get. this leads me my conundrum of the belief that historical charts don't give real-time effect.

 

[forker]

I didn't finish it, still editing please wait until I complete editing to read.


Editing complete.

 

[piphoe]

PA is a continuum, its how you capture it that matters

 

[dbphoenix]

Nobody said their was an intra day close. Just because a specific time frame paints a new bar at specified periods doesn't imply there is a close. It merely means the scope of analysis requires a new bar to be painted. This belief that candle charts in a historical sense doesn't provide meaningful test data is nonsensical. Let me twist this in a different direction to see where it takes this discussion.

If you are trading off a 1 hour chart how does watching it real-time differ from historical prints?

If you are implying that there is useful decision making information such as velocity and time, which is the only additional data that doesn't exist in completed candles, then I agree to an extent of looking at a single time frame.

Velocity and time have an inverse relationship. The less time spent at a price point the greater the volatility. Conversely, the greater the time spent at a price point the less volatility. If looking at an hourly chart you are effectively looking at the measured volatility over an hour.

If this is the only information missing, then looking at a more granularity time frame would give you a clear view of the volatility data you don't get. this leads me my conundrum of the belief that historical charts don't give real-time effect.

This may help.

 

[kalott]

I've got a basic idea for a strategy using price action and the MACD.

Is it strictly necessary to backtest this strategy using software?

Or is it enough to 'paper trade' for several hundred trades and figure out entry/exit that way?

uptrend?

parabolic uptrend?

uptrend with no pullbacks?

downtrend low volatility?

downtrend high volatility?

sideways chop?

do you have all this in "several hundred trades"?

 

[forker]

This may help.

Your thread doesn't answer my question however, it does help me understand why it can't answer it.

 

[darktrade]

Paper trading has exactly the same problems with backtesting since market dynamics can change immediately after you are convinced that you have a sound strategy. This paper in ssrn has in it two examples that demonstrate this problem. http://papers.ssrn.com/sol3/papers.cfm?abstract_id=2810170