Mark, Others, on the objectives of optimisation:
I remark point-wise to your response:
1) Agreed that 30 days could be too short for model building, but it depends, I think, strongly on the objectives(extend and accuracy) that are set. I think each method would require its own minimum period for making a meaningful model;
2) I think the frequency of re-balancing is more a question of personal preference, and possibly a matter of simple expediency, rather than a question of functionality. It would seem obvious that a re-balance resulting in a small shift of the cash/stock ratio is not worth the trouble of executing. I would certainly agree that some practical frequency should be chosen rather than each price update. In fact, I see this re-balancing as a separate subroutine A-La-The-Idiot-Wave-Mechanism that one can invoke as one pleases.
3) The cash addition mechanism to AIM has never been a problem to me as I saw this as a natural thing to do if the investment performed well. So, if a) re-balancing is advised I would simply add cash to the Reserve and update the PC with X-percent of the extra cash(X could be a variable that was optimised, if I was using the PC), and b) I was using the PV as Control Set Point then the extra cash would automatically be accounted for in the PV.
My preferred way of adding extra cash is my TurboVest Strategy as I would already be inclined to be fully invested most of the time.
(TurboVest = Controlled Margin Investing in times of Falling Stock Prices). I know that this stratgy needs to be applied with skill, but any investment method requires understanding of one's actions(in order to avoid surprises)... for example, my baseline ironclad criterion is that stocks must be worth buying a the time I buy them! Stocks that are worth buying could be going down or going up when I buy them. In this line of thinking it makes sense to buy a good stock, even if it has bee rising in value for a long time.
For as far as adding extra cash to buy new stock it is therefore necessary to ascertain quality rather than asking if the price is going up or down. The next point is, of course, that adding cash should be done on the basis of a plan: "How much am I willing to lose?" If one is willing to lose all he has then he will risk all he has one way or another. That is not an investment question. Although there are ineffective ways to risk all and effective ways to lose all! Think of the Kelly Risk Calculations for betting on gambling odds. If one risks all foolishly then the probability of winning is low. If one risks all wisely then the probability of winning is high.
Now, developing an effective optimisation technique is quite something different than thinking of the objectives in the first place, as you have already discovered. Having said this, I want to emphasise that manual parameter optimization for a specific stock price profile is easy...Many of us have alreay done it by tweaking all the parameters we had, and this results in an optimised Yield very quickly. It should therefore be a cinch to automate this optimization process.
My problems, as far as optimisation techniques go(the other ones I do not mention), are that I have no specific experience/knowhow of these techniques other than basic exposure to them from my engineering education(and interest in these matters). So, I feel unable to implement my ideas as my skills for doing so are minimal(and I have a strong aversion to statistical theories). So, for the moment, I am 2 eons away form doing any formulating of optimisation techniques myself, and I will die before 1 eon has gone by. I will have to be satisfied with chewing the fat on this and leave the formulating and testing to you and others. In this process it will be decided which objectives have a chance of succeeding and which do not.
Regards,
Conrad
AI
Market Data 
Markets 
