Revy, I can answer that for you.
Yes it is very worth backtesting for many reasons. For one thing, i have found psychologically the more solid my backtesting (and forward testing) process is, the more confident I am in the model, and therefore the less likely I am to fiddle with trades.
Backtesting if done the right way will provide a very solid basis with which to go forward in trading a strategy, and there are many benefits apart from the already psychological one:
1. First and most important is that it validates whether you have a viable model, and also just as important, how robust is it and which markets does it work well on and which not.
2. Backtesting provides the baseline to compare live trading performance with. Having that baseline will tell you if you are getting results that you should be getting (after all, if you experience say 4 or more straight losses you will start to question your model, but if your backtest stats say you can expect up to say 8 straight losses you know its within the expected range), or whether something is not right and your assumptions need to be revisited. This gives you a basis for taking action when necessary but only when necessary.
3. It also gives you a basis to determine how good a particular model compared to other models you may have, and also in what market conditions does it work well, and when not (i.e. does it work well in trending markets but not consolidating markets, or is it better for reversals, etc). This enables to decide when you start developing additional models what type of models you should buid - if you already have 1 or 2 good reversal models you probably don't need a 3rd since they will all perform well at similar times, increasing the choppiness in your equity curve - ideally you want to develop new models that will help smooth the equity curve you have from your existing models.
As for how to backtest - I downloaded an excellent (i.e. easy to read) article from Investopedia a couple of years ago which explains the process very well. Unfortunately I can't find the link but I will email you a copy to the address I already have. It explains the context of backtesting in terms of in-sample vs out-of-sample periods, how to select the periods of most relevance for each type of sample, and how long those pepriods should be. It also covers forward testing as well, and the risks associated with over-optimsation.
I don't think it covers the issue of recency, which talks to your pint about using the most recent data. Bascially for the reasons I discussed in my previous post if last year's performance turns out to be anomolous, i.e. well away from the historical mean (for example Brexit year would have been an outlier for the GBP I suspect), then relying only on that period will give misleading results - hence the importance of the out of sample testing and having some reasoning for selecting your sample periods (my coach encourages me to use 2008-09 - the GFC period as an out of sample period and if a models is robust to that it should be robust to anything).
I hope the above helps.
Cheers, Sharks
Expand
