Profit factor cannot translate to actual profits. It tells nothing about profitability. Do you think that someone who had 5 bucks of winners and 1 buck of losers is better than someone who has 1M of winners and 0.5M of losers? The former has a PF of 5 and the latter a PF of 2. Edge, the way I defined it in my post, translates to profits directly by multiplying it by the number of trades. Then you divide by your capital and you get your rate of return. Max drawdown by itsef is not an edge. Otherwise, T-Bills would be the best edge around. After so many posts you have, I supposed you had learned something
Few years ago, I changed what I sought from my trading as I use to seek highest winning percentages. Now I seek to keep my losing trades below a certain percentage based on weekly performance. My goal is normally 15% or lower losing trades, so I will have a combined total of 85% of winning and breakeven trades. I take far less trades cause of my goal, less commissions, and after so many losing trades in a row, which is few, I can double up contracts. So my edge is a well backtested method.
Depends on what you trade, but for stocks I suggest a Monte Carlo simulation comparing the result from your system with a large (>10000) number of random "systems" that just randomly selects symbols and dates. The random trades should be the same number and length as the trades from your system. We refer to this test as the "monkey test", with reference to articles on monkeys throwing darts can beat Wall Street gurus (of course they can - if they're lucky). Then calculate the return/sharpe ratio/drawdown for each of the equity curves of the random systems. The proportion of random systems that are better than yours give some indication if your results are from luck or not. We use a walk-forward data set for this type of test (and not the back test data set to avoid data snooping). Often you will be amazed how good results a dart-throwing monkey could get if it's lucky...
Quote from intradaybill: Profit factor cannot translate to actual profits. It tells nothing about profitability. Profit factor already includes all trades. Profit Factor is the sum of all profits divided by the sum of all losses. It tells you everything about the edge. How much your total profit is, is very unreliable. Do you think that someone who had 5 bucks of winners and 1 buck of losers is better than someone who has 1M of winners and 0.5M of losers? The former has a PF of 5 and the latter a PF of 2. This example really does not prove anything. If someone had $111 billion of profit and $110 billion of losses, would he have made a billion? It would be eaten up in commissions, fees and other costs. Edge, the way I defined it in my post, translates to profits directly by multiplying it by the number of trades. Then you divide by your capital and you get your rate of return. The number of trades is practically irrelevant. After getting something like Profit Factor, you must then calculate slippage, commissions, fees, risk, errors in judgement and a lot of other things. You then need to subtract all this and see if you are still reasonably above a PF of 1.0. Max drawdown by itsef is not an edge. Otherwise, T-Bills would be the best edge around. max DD, Sharpe, and other stats are measures of the reliability of an "edge" not an edge themselves. True Profitability depends on what kind of advantage you have in the way of wins as compared to losses, after all trading costs and risk is removed. Again, PF, max DD, Sharpe and other stats serve this purpose well.
We coded the one we use, so I'm not familiar with the capabilities of standard software tools, like TS, Ninja, etc. Maybe someone on ET knows more about the support of portfolio level testing in these tools. If not, then this would be nice feature to implement as part of a portfolio test function.
It is best to build your own, that way you can customize it to meet your specific needs. This is snapshot of what I use:
e.g. http://www.adaptrade.com/ http://www.compuvision.com.au/ http://www.portfolioriskanalysis.com/ http://www.zentrader.de/html/monte_carlo_simulator1.html Compare the features and conditions and you'll find what you need... bye, Volker
How to determine the edge? One option is to comparing your back test result to random buy/sell results. I found the buying at pivot support level and selling +/- 50% retracement from the buying has ~2-5% higher probability to win against the random buying.