He Lost 19 of 21 Trades. Why Didn’t He Change the System?

Peter Brandt has traded for 50 years. Two years ago, he went through a sequence in which 19 of 21 entries lost money.

He did not replace the system or rewrite the rules. In a recent Chat With Traders interview, Brandt said the sequence was worse than one standard deviation for his process, but still possible.

That can sound like a lesson in confidence. The more useful detail is his position size. Brandt says he normally risks 60 to 70 basis points of account capital on a trade.

At 0.65% fixed-fractional risk, ten full losses reduce equity by about 6.31%. Recovering requires a 6.74% gain. At 5% risk, the same ten losses produce a 40.13% drawdown and require a 67.02% recovery.

The losing streak is identical. The survival problem is not.

Win rate is incomplete

Brandt describes an informal poll he once conducted among floor traders. He asked whether they preferred a system that won 30% of the time or one that won 70%. Most chose the 30% system.

Low win rates are not inherently better. The point is that a profitable 30% system must usually earn much more on winners than it loses on losers. A system that depends on winning 70% of the time may have little room for error if its hit rate falls.

Consider two illustrative systems.

• System A wins 30% of trades, makes 4R on an average winner, and loses 1R on an average loser. Expectancy is 0.30 × 4 - 0.70 × 1 = +0.50R.
• System B wins 70% of trades, makes 0.5R on an average winner, and loses 1.5R on an average loser. Expectancy is 0.70 × 0.5 - 0.30 × 1.5 = -0.10R.

System B delivers more winning trades and still loses money over time. System A is wrong most of the time and can still have positive expectancy.

These are arithmetic examples, not performance forecasts. Real outcomes are not perfectly independent, payoff distributions change, and fees and slippage reduce results.

Brandt also says roughly 15% of his trades have historically produced about 85% of his profits. That is his own record, not a law that applies to every trader. It still raises a useful operating question: if a small set of large winners makes the year, how much are the other trades allowed to lose?

Keep the loss budget stable

Brandt says he uses roughly the same dollar risk whether he is long gold, short corn, trading currencies, or occasionally trading stocks. The number of contracts or shares changes, but a full stop is designed to cost the account a similar amount.

The stable variable is the loss budget, not the lot size.

A wider stop requires fewer units. A tighter stop permits more units. Liquidity, contract specifications, gaps, correlated positions, and execution constraints can still make the realized loss larger than the planned amount.

This structure changes the question from “Will this trade work?” to “How much capital remains if many trades do not work?” It also preserves the capacity to act when one of the unusually large winners finally appears.

Normal variance or a broken edge?

Surviving a drawdown is only useful if the strategy still has an edge. A 19-loss sequence could be normal variance, evidence of poor execution, or evidence that the market no longer rewards the rules.

Brandt does not diagnose that from one result. He tracks consistent rules and reviews performance across 50, 200, and longer trade windows. In the interview, he suggests following the same rules for at least 200 to 300 trades before deciding whether an edge exists.

That is his experience-based threshold, not a universal statistical minimum. A low-frequency strategy may not produce 200 observations for years, and the market can change before the sample is complete.

The principle is simpler: do not change the test while it is running.

Record the entry rule, stop rule, and risk budget before the trade. Track execution errors separately from strategy outcomes. Review expectancy and drawdown by a fixed strategy definition. Otherwise every losing streak can be explained away and every winner can be claimed as proof.

The Position Math Expectancy Calculator can combine measured win rate, average win, and average loss into expectancy per trade. It performs deterministic arithmetic; it cannot tell you whether the sample is large enough or whether the edge will persist.

Position Math is our own tool. It is provided for education and verification, not as trading advice or a promise of performance.

The lesson from Brandt’s long career is not that traders should tolerate unlimited losses. It is that each loss must be small enough to leave time and capital to determine what the streak means.

Control the cost of being wrong first. Then let the rare large winner matter.

Sources and method

The personal history, trade statistics, and risk rules in this article come from Peter Brandt’s August 5, 2026 interview with Chat With Traders. They were not independently audited. Drawdown calculations use fixed-fractional risk on current equity: drawdown = 1 - (1 - risk)^10.

The examples exclude fees, slippage, gaps, portfolio correlation, and changing trade size. Historical expectancy does not establish future performance. This article is educational content, not investment or trading advice.