Hi – Maybe we shouldn’t even attempt to ensure a well-working system/strategy/etc. because, well, when the data / environment changes, the Code may (will !) change its profitability.
Personally I never found an endlessly back-tested Code to work out the same in reality (actually, reality depicts that I better stop writing Codes :-). And no, I don’t over-optimize.
An other matter could be that a strategy (supported Code for it) should be presented for explicitly one security. For example, never think that a Code for EUR/USD works out the same for GBP/USD, no matter how similar they look at first glance. The one may give you profit of 100K while the other gives you loss of 1M.
No Code can be made resilient to
Trump
Covid-19
Brexit
Huwaway (5G stuff)
Uber (et al)
Covid-19 post effects (think Zoom, TeamViewer related)
Etc. etc. etc.
Additionally, what may work more or less for IG, may not work at all for IB.
What works for IB, will not work optimally at all for IG.
(both assuming that there will be AutoTrading for IB as well, which today only can be tested in Paper / BackTesting)
This is because of the enormous difference in Broker Fees (Commission).
Hey, the data isn’t even the same for PRT/IG vs PRT/IB ! (the differennces will mostly be marginal, but they make or brake your code/profit).
While the above counts for strategies (Codes running those), there’s also :
The graphical support – say Indicators. These can be checked in advance, I’d say. You like them (buy them) or not.
Special could be the Screeners; They present you instruments which readily can be checked upon by means of their functional description by the writer/seller which will enable checking for the functionality against Graphs. It will be YOU obtaining the instruments; not the person writing the Screener.
Nicolas, from the above may follow that Codes for AutoTrading should not even be allowed in. And to be honest, would I sell my ever winning code, might it exist ? OK, I might, for 100K euros. Mind you, I am serious here. An ever winning code is worth an infinite amount of money.
But you will be too afraid to buy it.
A Code which may win some and lose some, costing 50 euros, … you won’t buy that either.
When Codes for AutoTrading are not allowed in, OR when they are fully on the buyer’s risk (hence e.g. you won’t even check those), the problem could be solved. Thus summarised :
- An Indicator (or graphing from it) is the most easy to check for its workings;
- A Screener is also easy to validate against its promises. They are harmless when using them.
Both the above are as exciting to design and will be even more exciting to actually use.
My further (of course personal) advice would be to just obtain an life lasting license for e.g. 10 euros (no support), instead of a limited subscription like 50 cts per month (probably also without support). This also incorporates the trouble of invoicing (nobody is waiting for making invoices each month of 50 cts – also not invoices of 1 euro each year – trust me !).
Best regards,
Peter
(due beta tester for PRT/IB AutoTrading)