Many algorithmic structures have specific roles to be filled by indicators. For example, the No Nonsense Forex algorithmic structure includes: the ATR (for SL/TP placement), a Baseline (MA), 1st Confirmation, 2nd Confirmation, Volume/Volatility filter and an Exit indicator. With this example and others in mind, I have three question prompts: 1. What kinds of algorithmic structures benefit the most from walk forward analysis? Context: SL/TP placement metrics, indicator infrastructure and time frame interplay/differences. 2. How can we structure our own algorithms to maximize the benefit from walk forward analysis? Context: Think "trim the fat, keep the meat." 3. In 5.1 - 5.3 of your Algo Trading for a Living series, you describe your algo creation process in terms of the AGILE methodology. Please expand more on a successful algorithmic structure, from step 1 - entry and exit - and onward. I would like to know more about how you decide what to use and why. Thank you so much for this series. It has opened my eyes to the world of quant.
IM from NNFX too.......how do you know if your parameters are for a trend following or mean reversion? with just the indicators alone....since the baseline is the signal for trending up or down
I'm interested to know how the settings from multiple walk forward tests can be applied into one set a parameters/rules for an algorithm software to follow when my understanding is one optimization phase finds the best settings to be tested for each walk forward is one set. If combining multiple walkforward results in a set to get the best statistical outcome it would mean different sets from each of the optimization stages.
I would like to ask one point. I am watching your videos and I woul like to congrats you and the Darwinex. I wtched the video that you mentioned about the commissions and teached how to adjust. Do you have a video teaching about how to set or adjust a correct spread in order to realize the backtest? Thansk in advance. Paulo from Brazil.
How do you optimize systems which have multiple parameters to optimize. Also when should one use monte carlo based methods for optimization vs walk forward optimization. Would like to hear you thoughts on it.
Hey, thanks for this series, very useful information. But I am lost here with the WFA. Let us say I have 5 stages, in stage 1 the best parameters value are X and Y. How to benefit from this information when launching stage 2 ? If I launched stage 2 as standalone stage what if the best parameters of stage 2 are now E and F,... and at the end the pre-live opt is the one giving me the parameters to trade with in Live. What is the connection between all those stages and how to benefit from the parameters optimized in one stage for the other one? Thanks in advance.
hi, to answer your question : I'd likes some insight on how to measure alpha decay ; is it the same thing as measuring the CAGR/maxDD, Sortino, or Profit factor ?
Sometimes when designing a system that has two or more indicators for different purposes (e.g. a trend decider, a filter, and a stop loss), I usually make the lookback period for all of them to be the same, to reduce the number of parameters I have to optimize. Please let me know whether this approach is correct or not?
What would you do if the stategy involves more than the 3 parameters that you recommend to optimize at once? I suppose if you find the best common parameterset fo the first 3 parameters over all stages and with this fixed set move on to only vary the next 3 parameters you loose part of the advantage of the approach to adapt to market regime? Would you instead fix the best parameterset from the first 3 parameters for each stage individually and then optimize the next 3 parameteters only with them? Or reduce only the range of the first 3 parameters to the "successful area" and then optimize 6 parameters next - 3 of them with broad range and 3 of them with narrow range? Or what else?
