The Random Entry Experiment: Design and Setup
The experiment described in the study notes begins with a deliberately crude decision rule: for every trade in each of ten markets, the direction — long or short — is decided by a coin flip. No indicator, price pattern, or piece of news influences the choice. The system then stays in the market continuously, so when one trade closes, another coin flip determines the direction of the next position. Two further components complete the design. Exits are handled by a trailing stop placed at three times the current market volatility. Position size is set by a 1% risk rule, which sizes each trade so that being stopped out costs approximately 1% of account equity. The purpose of the exercise is not to recommend random trading. It is a controlled demonstration that isolates how much of a system's performance comes from exits and money management rather than from the quality of entry signals. The reported results are striking. Without risk-based position sizing, the random system was profitable in roughly 80% of runs, with each run producing different results because the coin flips differ. After adding the simple 1% risk money-management rule, the system was profitable in 100% of the runs over the ten-year period across the ten markets. Returns were described as modest rather than spectacular, and trade reliability — the percentage of trades that ended profitably — was only about 38%, close to the average for trend-following systems. Even so, a positive outcome in every tested run, achieved with random direction selection and a requirement to remain continuously exposed, is the detail that makes the experiment worth examining.
Why a Coin-Flip Entry Can Still Produce Profits
A coin flip cannot forecast price. Yet in this experiment the random direction choice did not prevent the overall system from making money. The explanation lies in where the edge is located. The entry determines only when exposure begins and which side of the market is held. It does not control how a trade ends, and it does not control how much capital is committed. Those two functions are handled by the trailing stop and the position-sizing rule. The trailing stop makes the outcome distribution asymmetric. Losing trades are cut at a predefined distance, so their size is bounded. Winning trades are allowed to continue as long as the market keeps moving in their favor, so their size is not fixed in advance. In any market where prices occasionally trend, a random entry has a chance of being positioned during part of that trend. Because the sample contains many trades across ten markets and a long test window, the entry points form a broad, unbiased sample of market conditions. The result is that the profile of gains and losses is shaped mainly by the exit rule rather than by the entry timing. A necessary condition for this outcome is continuous participation. The system is always in the market, so it does not need to predict when a trend will start; it only needs to be present when one occurs. This does not prove that entries are worthless. Entry signals still affect trading frequency, initial risk, and how efficiently capital is used, and a good signal may well add value in a different design. It does show, however, that an entry rule must be measured against a random baseline before it is credited with creating an edge.
