Poll: How often do you reoptimize ?

[newwurldmn]

i almost completely reinvent my trading strategy every 8-10 months.

I'm not sure that's really optimisation, we're talking about some kind of parameter fitting process I think.

I run a daily optimisation to determine what trades I should do, but the parameters that control the forecasts which dictate those trades are never optimised. So I answered 'never'. The parameters are based on 50 years of data. Having another year of data is irrelevant.

Most people optimise too much and too often. Even HFT guys only optimise about every 3 - 6 months.

A lot of this is to do with people not designing strategies that automatically cope with changes to eg volatility, price levels and so on. If you don't scale your forecasts according to vol estimates, you'd have to reoptimise every time volatility changes.

GAT

 

[stepandfetchit]

IMHO: It should depend upon characteristics of your strategy. For my primary strategy, I continually (in back of my mind) evaluate what I am doing and why, and cross-reference with what I think I should be doing. Two (at least) factors at play (1) the market is not constant, (2) Is my present implementations for my strategy the best for me.
So far, after finding a satisfactory (risk/return/effort), It took 8 months till I replaced the implementation with an improved one, then 5 months till overlying am improved implementation, and then last round, about 3 months till overlying that with an improved implementation. (new system runs coincident with one I expect to replace until it is ready for prime time). I suspect the decreasing time frames are not related to anything other than the ability to focus on the relevant factors. (the changes were more technical than market morphing so far)
Take with grain of salt, as I expect no two of us are really that similar.

 

[BlackPhoenix]

It would be an interesting test, to see what parameter optimization strategy would give the best results.

 

[stepandfetchit]

It would be an interesting test, to see what parameter optimization strategy would give the best results.

FWIW: If your strategy uses simple "cannon launch trades" (no adjustments), instead of optimizing, simply solve for best (based on trader's criteria ) -- Just compute power and memory. -- This may uncover structures that you may would never consider, otherwise.

 

[Sergio123]

I just started trading on a real platform this year, but I optimize the toolsets to keep up with the latest technology and codebase. Stocks will go up and go down. The RSI, MACD, Bollinger Bands, etc. have been around for decades. Your knowledge, experience and tools that you use to trade should constantly evolve.