One other insight was that money management (deciding how much dollars to buy each stock or how much to bet at one time on one thing) is much more important than (a) stock selection (b) entry price (c) sell rules. In other words, what we need from the ranking system is a small but very reliable statistical edge.
Even though portfolio123 makes money management very easy, it makes the implementation easy, not the design. For example, 3 stock systems and 4 stock systems can give huge returns for short periods but will decay sooner than later or they swing too much drawdowns that we stop following them.
Money management includes several other things, such as starting capital. For instance, a sim, which averages 70% annual returns when started with 100,000 portfolio, drops to 40% when started with a 10000 portfolio. Why? In the first year or two, commissions steal away most of the returns! In fact, if you start with $2000, this sim may never see the light of the day! Power of compounding working against us!
This actually led me to experiment for the first time with a fixed amount investment. For instance, given a 20000 dollars capital instead of specifying 5 stock sim (by choosing percentage 20%), choose fixed amount and specify $5000. First year the sim buys 4 stocks and as it makes more and more money, in five years, it makes enough money to buy 30 stocks spreading the risk over many stocks, limiting drawdowns.
Bottomline: While we must look for the holy grail ranking system (which means that its top rankers immediately turn profitable and grow the fastest in a short time and top rankers never ever lose money), there might not be such a measuring stick that works forever.
Instead, we might want to equally strive for a frictionless or lossless money managing strategy (which is deciding how many stocks for what amount and how to cut losses and how to let profits run and when to cut profits short) - as good as possible so that even with a suboptimal ranking system (as measured by % winners), the net results are fairly good and more importantly very very consistent. In order to do this a lot of experimentation has to go into % allocation, number of stocks in the portfolio and the sell rules.
What I am trying to emphasize is that this money management should be independent of the ranking, but rather, adjust itself to the results of the sim. For instance, a sim could watch its losers and winners over time and adjust its stop losses dynamically. It can see how much of its return came from the top 5 winners and adjust profit taking to ensure it does not get overheated. Over-optimization and curve-fitting is bad, but who said adjusting to live markets is bad? There goes my biggie feature request to Marco and team. It could be a feature where stops, position sizes, cash reserve (money allocation and sell rules) are all specified in terms of performance parameters measured over the last 50 trades, so the sim and the portfolio use those variables instead of hard numbers.
I am sure in the vast world of trading, people would have experimented with such ideas.
Ravi