2 registered members (AndrewAMD, dr_panther),
1,282
guests, and 4
spiders. |
Key:
Admin,
Global Mod,
Mod
|
|
|
Compare Z2 results from different accounts
#427845
08/14/13 13:46
08/14/13 13:46
|
Joined: Jul 2013
Posts: 15
Dzz
OP
Newbie
|
OP
Newbie
Joined: Jul 2013
Posts: 15
|
Hi, Just looked for performance of Z2 from acidburn's demo and my own real account. Results are way too different, curves somewhere different somewhere similar and trades taken differently. 1. acidburn's FXCM demo2. my FXCM realFor me from first look it seems quite much difference to be sure that systems behave in comparable manner between different accounts (and according to backtesting performance????). Any good thoughts/ideas? Somebody can show more Z2 results on myfxbook?
|
|
|
Re: Compare Z2 results from different accounts
[Re: 3DCat]
#429569
09/14/13 12:43
09/14/13 12:43
|
acidburn
Unregistered
|
acidburn
Unregistered
|
Also check on my page have you been running the same version of Zorro in all the same periods. From what I understand, all Z strategies are reoptimized for every release and I upgrade often & fast. If we were running different versions in the past, I think the results could differ wildly. At first I thought I would run only stable versions, and after they've been proved stable. But due to lots of initial instability I changed my mind and I now run each and every release, beta or not, I don't mind. It's all paper money, anyway.
|
|
|
Re: Compare Z2 results from different accounts
[Re: Sundance]
#429958
09/20/13 18:52
09/20/13 18:52
|
Joined: Jul 2013
Posts: 15
Dzz
OP
Newbie
|
OP
Newbie
Joined: Jul 2013
Posts: 15
|
Yes, I know, and agree about:
1) different bars timing 2) small shifts w ACIDBURN in version timing (I checked), big differences coming trading same version 3) secret random factor (if it exists additionally to bar changes) ??? 4) one or two orders not taken due to margin requirments (according to logs) but only one-two
But this doesn't answer some fundamental design problems which I see:
1) if so small differences (bar shifts, etc) produce so big changes, that means that strategy is very unstable and trade-chain dependent, meaning you can be profitable or non-profitable depending when U start trading each trade and this is just random 2) if I backtest the same period that I was running real, I will get different performance, meaning I can't compare how realistic are backtests to reality and do they at all have something in common. On Mt4 this is common problem, may be less on FXCM API 3) if I can't compare backtest w reality whats the worth in further actions like optimizing, WFA, ??? Even strategy developement.....
I don't blame Zorro, I like its idea and the program itself - thanks a lot to developers. But looks like the concept is quite contradictory in its nature: random things and serious trading. Please don't kill me for this outlook... May be JCL or smb else has different conclusions from my points...
Lets see how it will go further. May be smb else can add links to more accounts to compare running same strategies?
|
|
|
|