The only return that matters is after-tax
Every optimizer so far has worked in pre-tax space. But for a taxable investor, pre-tax optimality is a fiction — the only number that funds a retirement or a goal is what's left after the IRS. An "optimal" rebalance that realizes a pile of short-term gains can easily be worse, net of tax, than doing nothing; a 2% pre-tax improvement can be entirely erased by the tax on getting there. Tax-aware optimization treats the tax bill as a first-class cost in the objective, not an afterthought for the accountant.
This installment covers the two places optimization meets the real world most sharply: taxes and the short book. Both produce reviewed research that routes through the gallery's gates — and the tax-sensitive results in particular are designed to flow into tax review before anything is executed.
Tax-budget optimization
Tax-budget optimization lets you optimize subject to a cap on realized tax cost: "rebalance toward the target, but don't realize more than $X of net gains this year." The optimizer trades off the benefit of moving toward the ideal portfolio against the tax drag of getting there, and stops at the efficient frontier of after-tax outcomes. This is how you capture most of a rebalance's value while staying inside a tax budget the client can live with — often the most valuable constraint in the whole gallery for a taxable account.
The companion view is the after-tax comparison: the engine shows you the pre-tax "ideal" alongside the tax-constrained portfolio, so the give-up is explicit and the conversation with the client is honest.
Lot-level harvesting and wash sales
Tax intelligence lives at the lot level, not the position level. The engine reasons about individual tax lots — their cost basis, holding period, and unrealized gain or loss — to do two things well:
- Tax-loss harvesting. Identify lots sitting at a loss, realize them to bank the deduction, and replace them with a suitable substitute that keeps the portfolio's risk and factor exposure intact. The optimization picks which lots to sell to maximize harvested loss per unit of tracking error introduced.
- Wash-sale handling. The wash-sale rule disallows a harvested loss if you rebuy a substantially identical security within 30 days. The engine is wash-sale aware — it won't propose a harvest that a repurchase would invalidate, and it tracks the windows so the realized benefit is genuine, not illusory.
This is the difference between "sell your losers" (naive, often self-defeating) and a real after-tax optimization that preserves the portfolio while capturing the deduction.
Long-short and leveraged construction
For active and institutional mandates, the short book is not an afterthought — it's half the alpha. The gallery constructs both sides:
- Long-short / 130-30 style portfolios allow controlled shorting under a gross-leverage policy: go 130% long and 30% short for 100% net exposure, or whatever the mandate specifies. The optimizer builds the long and short books jointly, so the shorts hedge and complement the longs rather than being a bolted-on overlay.
- Gross and net exposure controls (the exposure constraints from Part 8) govern the leverage envelope — how much total gross exposure, how much directional net — so the strategy stays inside its risk and margin limits.