When the Best Trade Today Is the Wrong First Move

A trading decision can look optimal over the next interval and still be the wrong way to begin a longer strategy. The reason is subtle: today's position determines how the trading account will line up with tomorrow's opportunities.

I’ve been researching this question over the summer and am continuing that work. In my research, I study this effect in a simple two-asset model. The result is a concrete reversal. An asset that wins a one-period comparison can lose once the holder has another opportunity to trade. The analysis also explains where this reversal comes from and how the boundary between the two asset choices moves.

A passport option gives its holder the right to trade an account and receive its positive value at maturity. If the account finishes below zero, the option pays zero. Mathematically, the payoff is

This asymmetry makes the choice of trading strategy part of the valuation problem. The holder chooses positions to maximize the expected option payoff under the model's risk-neutral measure.

The research considers two stocks following correlated geometric Brownian motions, with zero interest rates and dividends. Trading takes place at predetermined dates. The total absolute number of shares held is bounded by one:

This is a share constraint, so stock prices and share units matter. It is different from limiting the dollar amount invested. Convexity allows the optimization to be reduced to four candidate positions: long or short one share of either stock.

With only one trading interval remaining, the optimal sign on each stock is contrarian to the account's current value. A positive account favors a short position; a negative account favors a long position. Choosing which stock to use is a separate question, determined by comparing the corresponding option values.

The tempting next step is to repeat that one-period comparison at every trading date. My results show why this can fail.

Consider two independent stocks with initial prices of 1 and 0.38, volatilities of 20% and 50%, and an initial account value of 0.10. There are two trading intervals, each of length one in the model's time units. The comparison is:

| Initial position | Value with one interval | Value with two intervals and optimal rebalancing |

| Short stock 1 | 0.142920 | 0.189698 |

| Short stock 2 | 0.143750 | 0.187144 |

The one-period calculation prefers stock 2. With a second trading opportunity, stock 1 becomes the uniquely optimal initial holding, including when compared with the long positions and every other feasible allocation.

The displayed values are numerical approximations. The research separately establishes rigorous enclosing bounds, proving that the two-period advantage of shorting stock 1 exceeds 0.002. The reversal is therefore supported by a computer-assisted certificate with controlled numerical errors.

Both competing positions are short. In this example, the reversal concerns the choice of asset, while the contrarian sign remains the same.

To understand the mechanism, imagine that the second trading interval is very short. That last opportunity is especially valuable when the account arrives close to zero. The point where the payoff changes from flat to increasing. Far above zero, the payoff is locally linear. Far below zero, a short interval usually provides little opportunity to reach a positive terminal value. Near zero, small price movements can make a substantial difference relative to the time remaining.

The first position determines which stock-price outcomes bring the account to that threshold. Shorting stock 1 makes a zero account correspond to a particular price of stock 1; shorting stock 2 selects a different condition. Those conditions imply different distributions of the opportunities available at the next date.

Even independent stocks can produce different continuation values. Independence does not make the relationship between the account and the stock used to generate it disappear.

The main analytical result makes this intuition precise. Keep the first interval fixed and let the final interval have length $\varepsilon$. For each initial candidate position $q$, its value with optimal subsequent trading satisfies

Here, $A_q$ is the one-period value. The coefficient $B_q$ measures the leading benefit of the additional trading opportunity. It combines the density of reaching a zero account with the conditional expectation of the larger available dollar variance. The next coefficient, $C_q$, refines that correction and also accounts for the percentage volatility of the asset providing the larger dollar variance.

The mathematical challenge is that both the payoff and the choice between assets introduce nonsmoothness. The proof controls the integrated expansion and its first derivative with respect to the second stock's initial price, uniformly over compact admissible ranges of that price and correlation.

That derivative control lets the analysis move from values to decisions. Suppose the two short positions tie in the one-period problem at a second-stock price $b_0$. For a sufficiently short final interval, there is a locally unique nearby price at which they tie in the two-period problem. The research calculates the displacement of this boundary through orders $\varepsilon$ and $\varepsilon^{3/2}$.

Between the original boundary and the shifted boundary, the one-period rule and the optimal two-period strategy can select different assets. The reversal therefore extends beyond a single numerical example. For a concrete parameter family, the research certifies a region of disagreement uniformly over correlations from zero to one-half, with an explicit sufficient bound on the final interval's length.

These results concern trading at fixed dates. They do not identify the optimal continuously rebalanced strategy or establish the global shape of the asset-selection boundary. The second-order boundary refinement is asymptotic.

What the analysis provides is a way to measure how future flexibility changes the first decision. In this model, evaluating an asset requires considering the account states it creates and the trading opportunities available in those states. The best first move depends on what the holder will be able to do next.