A Signature Method Tames Hedging When Trading Itself Moves Prices

If your own trading can nudge the market, hedging an option stops being a clean formula and becomes a balancing act. That is exactly the problem this paper tackles for path-dependent options, including contracts whose payoff depends on the full price history, such as Asian, barrier, or look-back options. The authors introduce a signature approach that turns this nonlinear, non-Markovian control problem into a tractable form under a mean-quadratic variation criterion. In their framework, the hedge can be written as a linear feedback rule in the time-augmented signature of the control variables, with coefficients governed by infinite-dimensional Riccati equations on the extended tensor algebra. Their numerical experiments show that these signature-based strategies work well for pricing and hedging general path-dependent payoffs when instantaneous and permanent market impact are present. A striking takeaway is that market impact itself smooths the optimal trading strategy, so low-truncated signature approximations become highly accurate and robust in frictional markets, unlike the frictionless case.

A large hedge order can move the price you are trying to protect against. If you have watched a thin stock jump after heavy trading, you have seen the issue. The problem gets harder for path-dependent options. Their payoff depends on the full path, not just today's quote. Asian, barrier, and look-back contracts all carry that memory. Once trades push prices, the hedge must balance two losses. It can miss the target, or it can trade too hard and stir the market. That is the knot this approach unties.

Why the usual hedge gets messy

This is a non-Markovian control problem. That means the past still matters. A signature is a compact summary of a path. It records the path's twists and turns in layers. That lets the hedge depend on the whole past without tracking the full past step by step. The result is a linear feedback rule in the time-augmented signature of the control variables. Linear feedback means each trade comes from a weighted mix of path features. The weights come from infinite-dimensional Riccati equations, which are big balance equations from control theory.

How the signature keeps history in play

The setup starts with a score that blends average miss and choppy trading. It cares about tracking error. It also cares about rough trading, which market impact makes costly. Instantaneous impact changes the execution price right away. Permanent impact changes the future market price too. The signature step turns the trading path into a feature list. A time-augmented signature keeps the clock and the control path together. Those weights live in an expanded bookkeeping space called the extended tensor algebra.

Low truncation means keeping only the first few layers of the signature. In frictionless markets, that shortcut can break down fast. Here, the frictions help instead of hurt. They smooth the trade path. That makes the short approximation more stable. The numerical tests show that this is not a fragile trick. It stays accurate and robust in frictional markets.

Why friction can make life easier

For quants, the gain is practical. The hedge no longer needs to stare at the full price history in raw form. It can work from a structured summary instead. That matters most when trades themselves shape the market. It also matters for contracts whose payoff depends on the path, such as Asian, barrier, and look-back options. The result is not that friction disappears. It is that friction can make the math kinder, not harsher.

What the surprise points to next

One concrete consequence stands out. In frictional markets, a short signature truncation may be enough to get a good hedge. That makes path-dependent trading less brittle. It also means the market's own pushback can simplify, not just complicate, the control rule. The surprise is the point. The force that seems to make hedging harder can also smooth it into something a lean model can handle.