Live Testing reports - page 38

 

AZBO

Seems you are here.

Had you receive my PM ????

 
kreeper_6:
Hello & Thanks, I don't see how bactesting over longer periods would be necessarily bad; wouldn't it give a better picture? Unless you mean that backtesting using history data isn't so use full.

Bit difficult in my bad english... Backtesting is usefull but you can't trust everything of the past to predict future..

You know what will happend tomorrow? You know ? I don't...

Ok backtest will give you an "idea"... Perhaps... coz world change every day and wil this "idea' of the past fit 100% of what will happend to morrow ? i don't think so

My 2c

 

Hello & Thanks,

I don't see how bactesting over longer periods would be necessarily bad; wouldn't it give a better picture? Unless you mean that backtesting using history data isn't so use full.

I know the past doesn't predict the future , phoenix isn't supposed to be predicting; phoenix should be making clear choices based on the facts of what is currently happening. I'm going to start trying optimization to get lower drawdown for some of the pairs.

 
bertbin:
Bit difficult in my bad english... Backtesting is usefull but you can't trust everything of the past to predict future..

You know what will happend tomorrow? You know ? I don't...

Ok backtest will give you an "idea"... Perhaps... coz world change every day and wil this "idea' of the past fit 100% of what will happend to morrow ? i don't think so

My 2c

But I'm saying that phoenix isn't predicting; it is making an educated guess using indicators to open positions. So backtesting should be an accurate picture of what kind of results to expect.

 

Kreeper_6, Phoenix 5 is a market dependent condensing detection counter trend system. It is market dependent, but not market aware. If the market shifts to a strong trend (like it did last summer) Phoenix should be limited to only follow the trend. You can do this manually with Phoenix 5, Phoenix 6 will have a filter system.

Phoenix 6 has the following types of trades:

Entry Signal (same as normal)

Confirmation Signal (bet more! These are jackpot trades with a high confidence)

Filter Signal (don't trade when this happens! time window trading and trend avoidance are two examples)

Exit Signal (If you see this happen, exit now instead of wait for a loss)

I don't know of any exit signals currently in use aside from the current "grace/force close system" which is also time dependent. The basic assumption is that if we were aiming for a quick trend reversal and it didn't happen, it probably won't happen. Rather than wait for a complete loss, take whatever loss or profit is currently available. A more intelligent variation may be fibonacci or elliot wave related.

 
kreeper_6:
I don't see how bactesting over longer periods would be necessarily bad; wouldn't it give a better picture?

let's say you do some backtesting for the past 4 month and you get your settings sorted out. now, the market behavior changes because the market goes in cycles (like life does). if you run another backtest with the changed market data, you will get different settings. you understand what i mean?

AZBOfin

 
bertbin:
AZBO

Seems you are here.

Had you receive my PM ????

i'm at work right now... yes i have, i'll contact you tonight.

ciao

AZBOfin

 
AZBOfin:
let's say you do some backtesting for the past 4 month and you get your settings sorted out. now, the market behavior changes because the market goes in cycles (like life does). if you run another backtest with the changed market data, you will get different settings. you understand what i mean? AZBOfin

Hmmm.......I'm thinking maybe I am getting the wrong Idea about these settings and optimization. Do we want to be optimized for "now" and then re-optimize when the market changes; instead of optimized for "ever".

 

Phoenix 5: "optimize for market"

Phoenix 6: "optimize forever"

 
daraknor:
Phoenix 5: "optimize for market" Phoenix 6: "optimize forever"

i guess that sums it up

AZBOfin