Marcus's $600,000 Portfolio: Why Continuous Tax Loss Harvesting Beat the Annual Approach by $214,000

Most investors who do tax loss harvesting at all do it once a year. Sometime in December, they — or their advisor, or their robo-advisor's annual sweep — look at the portfolio, find whatever positions happen to be underwater at that moment, sell them to capture the loss, and buy a replacement. It feels responsible. It feels like tax planning. And compared to doing nothing, it is meaningfully better than nothing. But the December scan captures only a small fraction of the losses a portfolio actually produces over a year, because it sees the portfolio at exactly one moment in time — and that moment is rarely the moment when the most losses are available. The losses that appeared during a February drawdown and recovered by April are invisible to a December scan. The individual lots that dipped below their purchase price during a summer volatility spike and bounced back are invisible too. The annual harvester sees a single snapshot; the market produced a year's worth of moving pictures.

This is the difference between annual and continuous tax loss harvesting, and it is not a small one. This article follows Marcus, a software engineer with a $600,000 taxable portfolio, through a twenty-year comparison of the two approaches applied to the identical portfolio in the identical markets. The only variable that changes is when and how often the portfolio is scanned for harvestable losses. The annual approach captures $180,000 in losses over the two decades. The continuous approach captures $1,080,000 — six times as much. After tax, the difference is worth approximately $214,000 to Marcus. Same portfolio, same holdings, same market returns, same investor. The entire gap comes from the frequency of harvesting.

Meet Marcus

Marcus is 45, a senior software engineer in Austin, Texas. His taxable brokerage account holds $600,000, built over fifteen years of steady contributions and dividend reinvestment into a diversified mix of low-cost index funds and ETFs — U.S. total market, international developed, emerging markets, a small-cap tilt, a REIT fund, and a bond allocation. It's a sensible, boring, well-diversified portfolio of the kind that most disciplined investors hold. Because of years of monthly contributions and quarterly dividend reinvestments, the portfolio contains roughly 400 individual tax lots scattered across its dozen positions, each lot acquired at a different price on a different date.

Marcus's tax situation is straightforward. His income puts him comfortably above the Net Investment Income Tax threshold, so his federal long-term capital gains rate is 23.8% (the 20% top LTCG rate plus the 3.8% NIIT). Texas has no state income tax, so 23.8% is his all-in marginal rate on long-term gains — which makes the arithmetic in this case study clean. (For residents of high-tax states, the advantage documented here is proportionally larger; a Californian at a 37.1% combined rate would see the same harvesting difference translate into a much bigger dollar figure, as our state capital gains tax rates 2026 guide details. The 23.8% used throughout this article is therefore a conservative, floor-level estimate of the advantage.)

We're going to run Marcus's portfolio twice through the same twenty years of market history — once managed with annual December harvesting, once with continuous lot-level harvesting — and compare what each approach captures.

The Two Approaches, Defined

Annual harvesting is the December scan. Once a year, near year-end, the portfolio is reviewed at the position level. Any position showing an overall unrealized loss is harvested; the loss is booked and a correlated replacement is purchased to maintain market exposure. This is what the large robo-advisors did for years, what most financial advisors do for clients who ask, and what diligent DIY investors do on their own. It has two structural limitations. First, it happens once, so it can only capture losses that exist on that specific December day. Second, it usually operates at the position level, so a fund that is up overall gets skipped entirely — even if a third of its individual lots are underwater.

Continuous harvesting scans every position at the individual lot level, every market day, all year long. When any lot crosses below its purchase price by enough to make harvesting worthwhile — accounting for transaction costs and the wash sale rule — the loss is captured and a replacement purchased. This approach sees every drawdown when it happens, not just the one that happens to coincide with December, and it sees losses inside winning positions that the position-level view conceals entirely. The mechanics of why those hidden lot-level losses are so abundant are covered in detail in our unrealized losses hiding in your winners deep dive; the short version is that in any diversified portfolio, even in an up year, a meaningful share of individual lots are below their cost basis at any given moment.

The difference between the two is not philosophical. It's a difference in how many of the portfolio's real, harvestable losses actually get captured before they disappear.

Why the December Snapshot Misses So Much

Consider a single, ordinary up year — the market finishes +12%, a perfectly pleasant year. To the annual harvester scanning in December, this looks like a year with almost nothing to harvest; the portfolio is up, most positions are green, the December snapshot shows few losses. Captured: maybe $2,000, from one or two laggard positions.

But that +12% year was not a straight line. It included a 9% pullback in February that recovered by April, a choppy sideways summer with several 4-5% dips, and a sharp but brief autumn selloff before the year-end rally. During each of those episodes, dozens of individual lots across Marcus's portfolio dropped below their purchase prices. A continuous harvester captured those losses as they appeared — in February, in the summer, in autumn — and booked them before the recoveries erased them. By December, those lots were green again and invisible to the annual scan, but the losses had already been harvested and locked in months earlier. Captured by the continuous approach in that same +12% year: around $44,000.

That is the entire insight of this case study, compressed into one year. The losses were real. They existed. They were harvestable. The annual harvester simply wasn't looking at the moment they were available. This is why, as we noted in our mid-year tax planning 2026 analysis, a flat or modestly-up year with high intra-year volatility can be one of the most productive years for a continuous harvester and one of the least productive for an annual one — the same market, opposite outcomes, driven entirely by timing.

A Representative Seven-Year Window

Here is how the two approaches diverge across a representative seven-year stretch of Marcus's twenty-year horizon, spanning several different market environments:

| Year | Market environment | Annual (Dec scan) | Continuous (lot-level) | |------|--------------------|-------------------|------------------------| | 1 | Steady up, +14% | $2,000 | $38,000 | | 2 | Grinding up, +9% | $1,500 | $41,000 | | 3 | Bear market, −18% | $31,000 | $94,000 | | 4 | Sharp recovery, +21% | $4,000 | $49,000 | | 5 | Choppy, roughly flat | $7,000 | $62,000 | | 6 | Up, +12% | $2,500 | $44,000 | | 7 | Correction, −9% | $15,000 | $50,000 | | 7-yr average | | $9,000/yr | $54,000/yr |

Two patterns stand out. First, in the down years (Year 3's bear market, Year 7's correction), the annual approach finally captures something meaningful — because in those years the December snapshot happens to catch the portfolio while it's underwater. Even then, continuous harvesting captures far more, because it harvested throughout the decline rather than only at the December checkpoint. Second, and more importantly, in the up and flat years — which are the majority of years — the annual approach captures almost nothing while the continuous approach keeps capturing $40,000–$60,000 anyway. That gap, repeated across every benign market year, is where the bulk of the advantage accumulates.

The seven-year averages — $9,000 a year for annual versus $54,000 a year for continuous — are a 6x difference, squarely within the 5-8x range we cite throughout this site. This is not a cherry-picked window; it's representative of the full twenty-year horizon, which contains the usual mix of bull years, bear years, and choppy years in roughly historical proportions.

The Twenty-Year Result

Extending those long-run averages across the full horizon produces the headline comparison:

| Milestone | Annual cumulative | Continuous cumulative | Extra losses captured | |-----------|-------------------|------------------------|------------------------| | Year 5 | $45,000 | $270,000 | $225,000 | | Year 10 | $90,000 | $540,000 | $450,000 | | Year 15 | $135,000 | $810,000 | $675,000 | | Year 20 | $180,000 | $1,080,000 | $900,000 |

Over twenty years, continuous harvesting captures $1,080,000 in realized losses against the annual approach's $180,000 — a difference of $900,000 in additional harvested losses, all from the same portfolio in the same markets.

The value of those extra losses depends on Marcus's tax rate. At his 23.8% all-in rate, the $900,000 of additional harvested losses is worth:

$900,000 × 23.8% = $214,200

That is the $214,000 advantage: the additional federal tax that continuous harvesting lets Marcus avoid or defer over twenty years, relative to harvesting only once a year. And it is a conservative figure, for two reasons. It uses the floor-level 23.8% rate that applies in a no-income-tax state; in a high-tax state the same $900,000 of losses is worth far more. And it counts only the direct tax value — it does not count the compounding on the tax dollars that stay invested rather than going to the IRS, which we'll come to shortly.

"But Losses Are Only Worth Something If You Have Gains to Offset"

This is the correct objection, and it's worth addressing directly, because a harvested loss sitting unused is not worth its face value. Losses create value in three ways: they offset realized capital gains (saving tax at the gains rate), they offset up to $3,000 of ordinary income per year, and they carry forward indefinitely until used. A continuous harvester who captures $54,000 a year needs somewhere to deploy those losses, or they simply pile up as carryforwards.

For Marcus, the losses get deployed continuously, and there is more than enough demand for them. Over twenty years, his $600,000 portfolio grows to roughly $1.9 million at a 6% real return, and that growth generates a steady stream of realized gains that the harvested losses absorb:

• Rebalancing. Keeping the target allocation requires periodically selling appreciated positions, each sale realizing a gain that a harvested loss can offset.
• Fund distributions. Index funds and ETFs distribute capital gains; harvested losses neutralize them.
• Matched-pair basis raising. The most valuable use: deliberately realizing gains on low-basis lots and pairing them with harvested losses so the net tax is zero, permanently raising the portfolio's cost basis. This is the entire strategy described in our matched pairs deep dive and raising cost basis to $0 tax — and it is only executable at scale if you have a large supply of harvested losses to spend. The annual harvester, capturing $9,000 a year, can raise very little basis. The continuous harvester, capturing $54,000 a year, can run an aggressive basis-raising program that the annual harvester simply cannot fund.
• Eventual drawdown. When Marcus eventually sells to fund retirement, the harvested losses offset the gains on those sales.

In other words, the continuous harvester's extra $900,000 in losses is not stranded. It is the raw material that makes every other tax strategy on this site executable — the concentrated-position unwinding in our California couple's case study, the embedded-gain defusing in our buy-and-hold case study, and the retirement-bracket strategies in our retiree's zero-tax gain case study. All of them depend on a supply of harvested losses, and continuous harvesting produces roughly six times more of that supply than annual harvesting does.

The Compounding Layer

The $214,000 figure understates the true advantage because it ignores time value. Every dollar of tax that Marcus defers or avoids in Year 3 is a dollar that stays invested and compounds for the remaining seventeen years. The continuous approach front-loads far more tax savings into the early and middle years (because it's capturing losses every year, not just in occasional down years), and those earlier savings have more time to grow.

Modeling the reinvested tax savings at the same 6% return, the continuous approach's advantage grows from the $214,000 direct figure to roughly $330,000–$360,000 in present-value terms by Year 20. This is the "tax alpha" that academic studies of continuous harvesting have estimated at roughly 0.5%–1.0% of additional annual after-tax return — and on a portfolio the size of Marcus's, compounded over two decades, that fractional annual edge becomes a six-figure sum. The exact figure depends on assumptions, but the direction is unambiguous: the longer the horizon and the higher the tax rate, the more the continuous approach pulls ahead.

Why Almost Nobody Does This Manually

If continuous harvesting is so much better, why is annual harvesting still the norm? Because doing it by hand is genuinely impractical, for three reasons.

It requires daily lot-level attention. Capturing $54,000 a year across 400 lots means monitoring every lot, every day, and acting in the brief windows when losses appear. No human investor watches 400 lots daily, and no advisor charging a reasonable fee can justify doing it manually for a single client. This is the kind of high-frequency, low-stakes-per-decision work that software handles trivially and humans handle almost never.

It requires flawless wash sale compliance. Harvesting frequently dramatically increases the number of opportunities to accidentally trigger the wash sale rule — buying a substantially identical security within 30 days of harvesting a loss disallows that loss. A continuous program must track every replacement purchase, every dividend reinvestment, and every contribution across every account to stay clean. Get it wrong and the harvested loss evaporates, as our RSU wash sale trap case study shows happening to an investor who wasn't tracking across accounts. Continuous harvesting is only safe when wash sale tracking is automated and comprehensive.

It requires emotional indifference to activity. Harvesting through a February drawdown means selling into a falling market — exactly when human investors freeze or want to wait for "things to settle." The continuous approach captures the February losses precisely because it acts mechanically, without waiting for a bottom that can only be identified in hindsight. The discipline that makes the strategy work is easier to maintain in software than in a human who is watching the value of their account drop.

These three requirements — daily lot-level monitoring, comprehensive wash sale tracking, and mechanical discipline — are exactly what continuous, AI-driven harvesting software exists to provide. The strategy isn't new or proprietary. What's new is the ability to execute it consistently across hundreds of lots without the cost and unreliability of doing it by hand.

Who Gains the Most

The size of the continuous-versus-annual advantage scales with three factors, which together identify who should care most about this distinction:

• Tax rate. The advantage is the extra losses multiplied by the marginal rate. Marcus's 23.8% is the floor. A high earner in California, New York City, or another high-tax state, as covered in our state capital gains tax rates 2026 and NIIT 2026 explainers, sees a proportionally larger dollar advantage from the identical harvesting difference.
• Portfolio size and lot count. More lots mean more independent opportunities for individual lots to go underwater, which means more harvestable losses for a continuous scanner to find. Marcus's 400 lots are typical for a $600,000 portfolio built through regular contributions; larger portfolios produce even more.
• Time horizon. The advantage compounds, so a 45-year-old like Marcus with a multi-decade horizon captures far more cumulative value than someone five years from fully liquidating. The earlier a continuous program starts, the larger the eventual gap.

The Bottom Line

Marcus's two portfolios were identical in every respect that investors usually obsess over — same holdings, same asset allocation, same contributions, same market returns. The only difference was the frequency of tax loss harvesting: once a year versus continuously. That single difference produced $900,000 in additional harvested losses and a $214,000 direct tax advantage over twenty years, growing to well over $300,000 once the reinvested savings are allowed to compound.

The lesson is not that annual harvesting is worthless — it beats doing nothing. The lesson is that annual harvesting captures only a small fraction of the value that's actually available, because it looks at the portfolio once when the portfolio produces opportunities continuously. The gap between the December snapshot and the year-round reality is the single largest, most overlooked source of tax alpha in a normal diversified portfolio. For the rate brackets that determine how much that alpha is worth, see capital gains tax rates 2026. For the lot-level mechanics that make the continuous approach work, see unrealized losses hiding in your winners and optimal tax lot selection. And for what to actually do with the losses once you're capturing six times more of them, the matched pairs deep dive and raising cost basis to $0 tax are where the real money gets made.

Same portfolio. Same markets. Same investor. A $214,000 difference, decided entirely by how often you look.