Concept
Random-entry Baselines
Random-entry Baselines are Statistics concepts. First implementations are in the build queue: the write-up leads, the indicators follow.
Tharp-style control experiments
What are random-entry baselines?
A random-entry baseline is a control experiment for a trading system: replace the system's entry signal with a coin flip, keep everything else identical (the exits, the position sizing, the instruments, the costs), and compare. If the full system barely outperforms its own coin-flip twin, the entry logic is contributing little, and whatever profitability exists is coming from exit management, sizing, or the market's drift.
The idea was popularized by Van K. Tharp, who described a random-entry experiment, developed with trader Tom Basso, in which coin-flip entries paired with a wide volatility-based trailing stop and consistent position sizing reportedly produced positive expectancy across a diversified futures portfolio. The lesson drawn was pointed: much of what traders attribute to clever entries actually comes from exits, sizing, and being exposed to trends at all. A random baseline makes that attribution measurable for any specific system rather than a matter of folklore.
Statistically, this is a permutation-style control. Running the random twin many times with different seeds produces a distribution of outcomes for expectancy, win rate, drawdown, and other metrics. The real system's results can then be located within that distribution: a system whose performance sits deep in the right tail of its random-entry distribution has demonstrable entry skill; one sitting near the median has none, however good its equity curve looks in isolation.
How traders use it
- Isolating the entry's contribution: run several hundred random-entry replays with the system's exact exits, sizing, and costs, then check what percentile of that distribution the real system's expectancy falls in; this is the cleanest attribution test most retail traders can run.
- Auditing exit and sizing rules in reverse: if random entries with the system's exits are themselves profitable in the tested market, the exits and the market's drift are doing real work, which is worth knowing before crediting the signal.
- Benchmarking long-biased strategies honestly: in a market that rose through the test period, random long entries will be profitable too, so the baseline exposes how much of a strategy's return is market drift rather than skill, complementing randomization tests run on the signal itself.
- Quality scoring in context: metrics such as SQN or profit factor gain meaning when quoted relative to the random baseline's distribution instead of in absolute terms.
- Limitations: results are specific to the tested market and period, random replays inherit the same historical path (so they do not test robustness to different market histories the way Monte Carlo price paths attempt to), and cost assumptions matter since random systems tend to trade at least as frequently as the originals.
Random-entry baselines vs related concepts
Randomization Tests: Randomization tests shuffle or scramble data or signals to ask whether observed performance beats chance in general; a random-entry baseline is a targeted version that randomizes only the entry while preserving the rest of the system intact.
Monte Carlo Price Paths: Monte Carlo path work varies the market history a system faces; random-entry baselines keep the history fixed and vary the entries, answering an attribution question rather than a robustness question.
Resampling Tests: Resampling reshuffles a system's own trade results to estimate the variability of its metrics; the random baseline instead generates entirely new trades from chance entries to serve as a control group.
Related concepts · Simulation
Concept family
Statistics
46 concepts mapped · 46 in the Library
Random-entry Baselines FAQ
Did random entries really make money in Tharp's experiment?
As reported, yes: coin-flip entries with a wide volatility trailing stop and consistent sizing showed positive expectancy across a diversified futures set. The claim is about that configuration and era, not a general property of random trading, and costs and drift did some of the work.
How many random runs do I need for a meaningful baseline?
Enough to characterize the distribution of outcomes; a few hundred replays with different seeds is a common working range. A single random run tells you almost nothing, since its result is itself a random draw.
My system beats its random baseline. Am I done validating?
No. It clears one specific hurdle, entry attribution on one historical path. Overfitting, regime dependence, and cost realism still need their own checks, such as walk-forward analysis and out-of-sample testing.
What does it mean if my system performs worse than random entries?
It happens more often than traders expect: the entry logic is systematically mistimed, or it filters out the trades the exits were best at managing. That is a genuine finding, and it usually redirects effort from entries toward exits and sizing.
Should the random baseline trade the same number of times as my system?
Ideally the entry probability is calibrated so trade frequency roughly matches, otherwise differences in cost drag and exposure time contaminate the comparison.
Build Random-entry Baselines your way.
Quant writes, tests, and refines it with you — then it runs on LuxAlgo charting or ports to TradingView.