TradeSim - How it works

As a trader you build systems and methods that give an edge. A probability advantage. At the outset of each trade the profit is unknown. What is known (or should be) is:

The existence of an edge - or statistical advantage - for a method of trading is demonstrated by past profits of the method. Assuming that the edge is good enough to be profitable, how can it be measured? How will it do in the future? These are questions answered by TradeSIm.

The outcomes of individual trades are varied. There is a range of profits and losses. Profits are what the market gave the profit-taking plan while losses are the result of stops.

If entries and exits are made randomly, the returns (profits and losses) from this method will likely be randomly distributed and on average profits will not outweigh losses. This is much like coin tossing - there may be runs of heads or tails - even long runs - but the average ratio of heads to total tosses will tend toward 50%. If this were a trading method there would not be much of a profit expected.

The purpose of methods for entry, stop loss and profit taking is to bias both the probability of a profitable trade and size of trade returns toward profits. A random entry and exit method produces a random distribution about zero. A good trading method chooses entries that have high probability of profit, cuts off losses with stops and lets profits run as far as the market allows. The distribution now looks Asymmetrical. The loss side is chopped off and the profit is large.

Analyzing Trades

Having made a series of trades (or produced hypothetical results by system testing on past data), how can the quality of the trading method be measured? By measuring the results.

The easiest result to measure is the total profit from a series of trades. This is certainly a useful money in the bank number but other measures of success are also important. The percent of winning trades is useful since this influences the probability of having runs of wins and losses. Drawdown is very important. If the method produces very good profits but has very high drawdowns it may be psychologically and practically unsuitable for trading.

A list of past trades contains all this information. Equity is known after each trade. Drawdown and percent profitable can be calculated. The problem is that the past series of trades will not repeat with the exact same returns.  If we were sure that in the next series of trades that this sample of past results would repeat exactly and in the same order then the future trades could be taken with confidence that the exact measures of profit and drawdown would repeat. If we knew this we would even try to skip the unprofitable trades. No such assurances exist in trading. Even if the returns in the series of trades were statistically equivalent to the past series of trades, the order in which they occur will not likely be the same. They will combine in random order producing longer and shorter strings of wins and losses.

TradeSim uses a method of analyzing the range of future outcomes to combine the historic returns in a random way.  Put the historic returns in a hat and draw them one at a time.  Record the value of each draw then replace it in the hat.  As drawing continues, the order of the trades is scrambled randomly.  Some trades may be included more than once while others are not included at all.  This is representative of what will happen when the method is traded in the future.

Suppose there are 50 historic returns and the draws are repeated 50 times.  At the end of these 50 trades  note the ending equity, the drawdown, the longest run of losses, the lowest equity and any other measure that may be of interest.  These 50 trades are one trial.  Now repeat the trial 1,000 times.  At the end of the trials, lists of ending equity, drawdown and etc are sorted into order so we can see what percentile of the trials was profitable and what percentile of trials have drawdowns greater than some value.  Better yet plot all the measures of trading results against their percentile ranking for a visual picture of how good the method is.


Home

Copyright 2002, Larry Sanders

Last update 2002.06.15