Tony, A number of years ago I thought you had to have a mechanical system to implement it, but I don't anymore as I primarily trade only for myself, and it's a discretionary-based option & equities approach. One of the things I've been working on in recent years are robust ways of determining things. As a quick-n-dirty example, if Im trading just one item, I can take get pretty involved with the data to determine what I "should" trade (again, this is dependent on my criteria, but let;s assume for the moment that it IS expected growth maximization "in the long run"). So it;s the peak of the curve, as a % of my stake that I should be trading, and as just mentioned, there's some pretty involved ways to determine this from past data. But that's the past and I'm going headfirst into tomorrow. I need a "best-guess" estimate of where that peak will be. Without going into the why's, I can determine the best guess as p'/2, where p' is what I expect the probability of winning is in the future time window I will be solving my criterion for. So if I expect to win 50% of the time in the future window, then absent anything else, my best guess estimate for the peak is .5/2 = .25 or 25% as the peak of the curve in the future. It;s simple and robust, and I don't need a mechanical system to arrive at it, just an estimate of p' and what my worst-case loss might be in that future window.
Trading the equity curve is position sizing. Cut the loss fast and increase the size of winning trades. "The most fundamental point that you must understand in terms of money management is that in a negative expectation game, there is no money-management scheme that will make you a winner." (Ralph Vince) Trading the equity curve will not work if the strategy has no edge.
But expectation may NOT be what you think it is, what the generally-accepted means of determining it is (probability-weighted mean outcome). Rather, it is the median of sorted terminal nodes of the tree of outcomes to a horizon of Q trials. Only in the limit (Q->infinity) do the two converge.
No, trading the equity curve is a seperate additional step from position sizing. I'm talking about looking at the equity curve of a system which is already doing position sizing for eg volatility of positions, size of account curve and trend strength. And if you are cutting losses and letting winners run, then most likely adding this extra stage won't work. Trading the equity curve actually works better if your strategy is rubbish - in the sense it will reduce your losses. GAT
Very true I think. An example is doubling down on roulette, a truly losing proposition in the long run, until you have a winner. This is an extreme case, but illustrates the point. To my way of thinking, we are always trading the equity curve. If I think of that curve of the % of our stake risked on every trade, we have a horizontal axis that goes from "risk nothing" (0), to "risk the house" (1), whereupon we are on every trade, if I am always trading the same quantity, I am moving up the curve (in the direction towards 1) as equity decreases (just as our double-down roulette player was doing), down the curve as our equity increases. And I contend that even for a positive expectation game, there is a benefit to using this (various algorithmically-determined paths along this curve or surface) to try to achieve various criteria we are trying to achieve in trading. This, to me, is what equity curve trading consists of (and we are doing it whether we acknowledge it or not) and it;s not the same thing, I believe, to what others here are calling equity curve trading in terms of actually trading the cumulative equity line. It happens indirectly in the manner I am describing though, and is contingent on the trader having a defined criterion he seeks.