Just wondering what others feel about setting Calculate on Bar Close to false in their strategies. I’ve been testing it now and then over the last month or so, but I’m not sold on it yet and wondering if it’s a waste of time.
The thing is, I’m not set up yet the way I want in order to test it effectively. I need a second/third system and I also need to upgrade to the zenfire data feed. But, from what I see right now, it’s almost too much noise to be useful – especially on entries. My testing has been restricted to short timeframes like 3-5 range bar, so maybe its better suited for longer timeframe strategies.
I can imagine if you have it set up just right it could give you a nice edge on your buys and sells, but it would be nice to get some feedback that supports this. If you have thoughts or opinions on how to best use this setting, I’d appreciate hearing them.
do you use only for adjusting your stop loss and/or profit targets?
do you use it only on 5min + timeframe strategies?
do you use it exclusively on all strategies and would never run one without it?
I never use calculate on bar close set to false. First on backtesting it doesn't work. Second on live charts any possible gain from a few seconds earlier entry is given back ten fold overtime due to the extra false signals or entries where you thought a signal occurred but it turns out it didn't once the bar actually closed.
I forgot to say that in strategies I use a tiny time frame like a 1 bar Range period for adjusting stops, still no calc on bar close. even on my big charts where a bar can take 5 to 10 minutes to form I find no value in cobc false.
Good to know caprica. I'm going to give the 1 bar Period a try, that sounds interesting. I was adding in the Period tick for a while, but I've eliminated that too since it seemed to hurt more than help.
I guess it depends on what setups/strategies you like, and it either works for you or it doesn't.
Writing a COBC = false strategy is more difficult however due to the issues you mention, mainly 'false signals' which arise temporarily and then revert before the bar close.
Essentially the way to handle that is to monitor for certain events which you want to act on in the middle of a bar, generally you would have some kind of threshold to ensure that only significant events would be acted upon. All other 'normal' event processing is handled on BarClose (FirstTickOfBar == true).
Here are a few situations where you would want to consider using COBC = false:
Monitor volume and DOM in real time
Monitoring Bid/Ask spread (very important if trading spot forex IMHO)
Advanced use of IOrder interface (ie modifying limit orders, scaling in or out, chasing targets etc)
MTF or MultiInstrument strategies
Pattern classification techniques (monitoring divergence formation, CCI patterns)
Thanks for sharing that. What you describe is the direction I was going in with my tests, so that’s encouraging. I was also coding it so that I could run with cobc set to either true or false, so that I could at least use the backtest for most of the strategy – which seemed to work pretty well. But, I think I’ll put it aside for now until I have a better platform for testing and more time to spend on it.
I don’t trade the forex, but monitoring volume and bid/ask spreads sounds like something I’d like to try at some point. There seems to be lots of games played with bid/ask volume when you watch the way it behaves. Some nice fakeouts.
There’s just so many cool things to try that it’s a challenge to stay focused on the ones that have potential to add the most value. I think I need to hire some assistants!
BTW I wouldn't put much faith in the NT backtester, even if you are using COBC = true. NT Market replay is better but really you have to run it live before you know what will happen.
I have mentioned before on NT forums that my most profitable strategy fails horribly in backtesting. It fares slightly better in market replay, usually breaking even or making a small profit at the end of the day. Luckily I had enough faith in the idea to test it live despite the poor backtest/replay results because it works quite well in real trading =)
One thing to note about market replay, I have observed that the results are much more accurate if you run it at low speed, ie 2x-8x speed or thereabouts. But as you said there are way too many interesting ideas to try and not enough time, so I normally run it on 200-500x speed just to get a feel for the strategy then if the results are close to my expectations I will run it in live simulation for a few days.
That's funny. I hadn't come across that tidbit on the forums, but somehow I'm not surprised.
I have a love/hate relationship with the backtester, but right now we're on good terms. I found a couple things I was doing that were causing it to be flakey on me. Not all the time, but just enough to drive you crazy. So, knock on wood, I'm going to continue with it for the time being.
But there's no substitue for the real thing I guess. I like testing in sim mode, but I just wish there was a way to get the output into a log file instead of that stupid window. I'll start using market replay more once I have my second system built, but right it just isn't practical with my setup.
It's nice to hear how people like yourself that are having success approach things, and how you utilize the tools.