Hmmm... did anyone tried to backtest and optimize a system on 5 years of 5-min data with WL? How much time does it take, assuming that system is simple with ~2 - 5 parameters?
Waw, Interesting. How is their programming environment, meaning, how does the language used to describe the system look like? How does it "feel" in relation to TradeStation? Do they allow dll's to be created and called from amibroker? nitro
Sure! Such as disktrading, you can buy historical data from them, or sharing with someone. I can only get 6 months or less tick data from Qchart(using QCollector), if I begin to collect right now. This means I can't get last quarter's e-mini tick data.
What I know about Tick Data is... Any 3rd-party software doesn't have the ability to clearly test them. Also, due to the volatility of the sub-minute, 1 point in a fill is extremely critical. As for ES, the median of the swing in 1 tick data is 0.184... So what that means is the if you took market orders for both sides, you'll lose money on both sides. Then comes the part about using Volume and time for filters... They are very cyclical, it changes every 3-4 days eventually ending up Breaking Even. (Even with Magic Fills) Finally, when you deal with Tick Data, there are some statistical signficance in Volatility type spread stuff. But the edge itself is extremely small, Sub-tick significance, that it's merely impossible to implement... unless you get into a higher timeframe or use some large set of cumulative analysis.
I think I should clarify a few issues here. The quote above was surely referenced to a different situation. Asuming you run a system in 5 minutes time interval and you have an 20 bar RSI associated with the system. This 20 bar RSI will only be recalculated at the end of the 5 minute bar and not on every tick of this 5 minute bar. There are many good reason for doing that, one is that when backtesting you used (most of the time) the closing price of each bar - and so we wait until you know for sure that you have seen the closing price of the bar. I personally prefer this kind of behaviour. On the other site you could/can use this information for intra bar action. For example you could say: If the RSI (bar -1, #Close, 20) is smaller then 30 then buy next bar at open plus one ATR. This order could be placed automatically. So WLD3 is able to use intra day data and can be used for intra day trading. Of course I will be answering further questions here if needed.
OK, I'm confused... I have one guy saying that it can't be used for backtesting on intraday tick data and another saying that it can... which is it? For example, In backtesting, If a trade occurs on tick that crosses a specified price, can some action occur at that moment or does it have to wait for a bar to complete?
Again, it depends on what triggered the signal. Lets say you wish to buy at the highest high of the lat bar, then this order will be executed as soon as this price has been triggered. You DONT have to wait until the end of this bar. If you use an indicator that is using the CLOSE of the bar then you have to wait for the close of the bar. WLD3 does not recognize every tick that finally will transform into the complete bar as the "close" of the bar. Excample, you have a RSI system that is using five minute bars. The rule tells you to buy if the RSI(20) is below 40. I will now show a run down on the price pluse the RSI indicator during that 5 minute period: Price - RSI 110 - 45 111 - 46 109 - 43 107 - 41 105 - 39 107 - 41 109 - 43 Lets assume that this 5 minute bar had 7 ticks. The final 5 minute bar would look like this: Open - 110 High - 111 Low - 105 Close - 109 At the closing price the RSI was at 43. So we would have not bought the stock/future even thou during the transformation of the five minute bar the RSI was for one tick below 40. Usually people test with the closing/opening price. So why would /should you get in at some inter-bar action. From my research done years ago it seemed to give me more and false signal. Hope this explains the behaviour of WLD3.
Hmmmm, IMO, WLD "should" have a feature called "Tick Resolution." If this is turned on, even when forming n-minute bars, WL would internally have every tick that formed those bars. Now, every tick, even in a n-minute bar, would be able to trigger a signal _while_ the bar was forming. For those that don't want this behavior, they would simply turn off "Tick Resolution." Have your cake and eat it too. nitro