How far back to optimise

What are peoples' thoughts on an ideal period of time to optimise an EA over?

Generally I think the shorter the timeframe being traded the shorter the optimisation period. The type of trades it makes (trend, range, breakout, etc) and when may need to be taken into account too.

I want to optimise for a trend finding EA traded on the 30M chart around the clock. What would be good: a day, week, month, year or two?

I remember reading in one of Ronald Rayguns' posts he optimised sme of his EAs at the end of the week over the previous four weeks, with the idea that whateer had happened in the last month would likely continue for at least another week.

Perhaps there are some thoughts on macro economic cycles? There are what, 3 or 4 cycles a year? Or are there thoughts on the uselessness of optimising as well?

I suppose it would be possible to run 1000s of backtests to find the ideal optimisation period for a set future period period of time (if any), but that would take a long time.

5-10 years of data used in backtesting will provide a very strong system, but after you've tested it long-term you need to test the short-term to make sure it has been working in the last year or less. An ideal way of doing this would be to export the data from the bars in the system and create an equity curve to see how the trade has performed over time to make sure the strategy was great for 7 years but hasn't worked once in the last 2.