|
Re: Zorro and hibernation mode
[Re: jcl]
#425555
07/08/13 13:16
07/08/13 13:16
|
acidburn
Unregistered
|
acidburn
Unregistered
|
Any Zorro restart at weekend can reduce the performance in the following week, for various reasons.
This contradicts previous information we had on Z strategies, and doesn't seem robust enough. Don't you think that documenting "various reasons" would be a good idea? I mean even FXCM shuts down it's servers on weekend, and if we can't do the same we can't even make a proper backup?! So far we only know that phantom trade logic is reset on every restart (can live with that). What else is impacted by weekend restart?
|
|
|
Re: Zorro and hibernation mode
[Re: ]
#425565
07/08/13 14:19
07/08/13 14:19
|
Joined: Jul 2000
Posts: 27,986 Frankfurt
jcl
Chief Engineer
|
Chief Engineer
Joined: Jul 2000
Posts: 27,986
Frankfurt
|
The composition of the initial bar. You start normally not at a bar boundary, but in the middle of a bar. Thus the bar will partially consist of historic prices from the server, partially of new prices received later. There can be a gap between the latest historic prices and the current price, especially when you start trading while the server is offline or not up to date, as on weekends. This gap makes the bar different to the same bar when Zorro were not restarted.
Aside from that, if your script does not use phantom trades or store other trade relevant information in internal variables, you can normally restart Zorro without affecting the performance.
The initial bar problem is less relevant with the MT4 interface because the MT4 server is normally always online. But there's another problem: MT4 stores not enough price data for long lookback periods such as used by the Z systems. We have not found a noticeable performance difference, but it is possible, as the MT4 history size is slightly below the recommended periods for the used filters.
|
|
|
Re: Zorro and hibernation mode
[Re: jcl]
#425584
07/08/13 19:05
07/08/13 19:05
|
acidburn
Unregistered
|
acidburn
Unregistered
|
The composition of the initial bar. You start normally not at a bar boundary, but in the middle of a bar. Thus the bar will partially consist of historic prices from the server, partially of new prices received later. There can be a gap between the latest historic prices and the current price, especially when you start trading while the server is offline or not up to date, as on weekends. This gap makes the bar different to the same bar when Zorro were not restarted.
OK, this is not that bad. Now finally I understand fully why you were always against restarting, even if it looks benign to the uninitiated (if done over weekend). The above paragraph should find its way into the documentation, IMHO. So, only one bar is affected. The same one that in many cases also includes big weekend gap. Ugly. It would be great if the last bar from previous week ended there, and new week opened with a new bar. But, I'm sure that is easy to imagine, but much more complex to properly implement, thanks to various timeframes, baroffsets, timezones, DST and whatnot. Actually, for a long time I wanted to crosscheck my tick data with fxcm data from zorro, to validate both of them, but I know it will take hours because some of that stuff can really get complicated. Back to our restart issue, just to make sure I got everything right. If weekends are bad because of data gaps and server outages, would than actually make more sense to do quick restarts on the weekdays, when the market is calm? Of course I understand the danger of missing some price data, especially if the strategy was about to open/close a position. But, at least the historic data should be available from the servers, without any gaps. Also, I noticed that fxcm servers are completely unavailable on Saturdays, but come back on Sundays (market still closed, of course). Is there any difference if (re)starting Zorro on Saturday or Sunday? Maybe it would be best, even in this case, to start late Sunday, after the market opens, but before Zorro resumes trading? Please give advice. Thanks.
|
|
|
Re: Zorro and hibernation mode
[Re: ]
#425603
07/09/13 10:07
07/09/13 10:07
|
Joined: Jul 2000
Posts: 27,986 Frankfurt
jcl
Chief Engineer
|
Chief Engineer
Joined: Jul 2000
Posts: 27,986
Frankfurt
|
We found that FXCM servers, although online on Sundays, were sometimes no up to date, so some price data was missing. This seemed not to have happened anymore recently, but FXCM does not guarantee the consistency of their price servers during weekends and holidays.
So my answer is: It's theoretically possible that restarting at weekends does not affect performance, but it's not guaranteed.
To tell the truth, for our own trading we've never experimented with restarts so far. In fact none of us even ever got the idea to intentionally stop and restart a live trading software. We have some Zorros here running for more than a year, still in version 0.9. So all what I told here is rather theoretical, we have no practical experience of restarting effects and side effects.
|
|
|
Re: Zorro and hibernation mode
[Re: jcl]
#425606
07/09/13 10:11
07/09/13 10:11
|
acidburn
Unregistered
|
acidburn
Unregistered
|
Thanks jcl for the nice explanation.
|
|
|
Re: Zorro and hibernation mode
[Re: ]
#425828
07/11/13 12:44
07/11/13 12:44
|
Joined: Jun 2013
Posts: 1,609
DdlV
Serious User
|
Serious User
Joined: Jun 2013
Posts: 1,609
|
Thanks also, jcl. However, I'm still a unclear as to the bottom line. It seems clear enough that if there's no reason to, don't stop Zorro. However, what if there is a (at least potential) reason? For example, my machine says it has "important" Windows updates to do. (Could be antivirus or whatever else...) Is there a recommended procedure? a) Ignore Microsoft & don't do them ever. b) Do them over the weekend after Zorro's logged out. c) Do them over some night when trading's quiet. d) Do them whenever's most convenient for you. e) Other? Thanks.
|
|
|
|