bBickford Technologies

Research record · as of 12 October 2026

Can any strategy beat simply holding the S&P 500, after costs and tax?

Bickford Technologies tested 26 ideas against VOO, an S&P 500 index fund, bought and held. This is the record: what was tested, what failed, why, and the one rule that held up.

Predicting prices: no. Every timing and pattern idea tested so far lost to trading costs, tax or noise.

After tax: one rule held up. Harvesting losses into another S&P 500 fund was never behind in 195 ten-year windows since 2000. It is small, and it waits on one legal question.

26ideas tested against holding VOO, after costs and tax
0beat it by predicting prices
1tax rule never behind VOO in 195 ten-year windows since 2000

The bar

To pass, an idea has to beat VOO after costs and tax, on data it wasn’t chosen on, by more than luck explains.

  • Each idea is written down before it is tested: the rule, the data, and why it should work.
  • It is judged only on the years after it was found or published, with up to 100 years of daily US market data.
  • Trading costs come from the broker’s published execution figures; gains are taxed at real rates, short-term as income.
  • The luck bar rises with every version tried, and the bar to put money behind an idea is a t of about 4.
  • When an idea fails, the reason is split into parts (cost, tax, time out of the market, decay after publication, fund gap, noise) and written down, so the next idea is aimed at what killed the last one.

Scoreboard: 26 ideas tested

25 failed against holding VOO. One, a tax rule, held up and waits on a legal question. Open a family to see each idea and why it failed.

Calendar and timing5 ideas The good days are real but small, and shrank after they were published. Cash, short-term tax and noise take the rest.

  1. Turn of the month, cash otherwise Failed

    In the market only for the days around each turn of the month (1 day before to 3 after), in T-bills the rest of the time. Judged on 1988–2026.

    Sitting in T-bills gave up 7.85% a year of market return. The month-turn days still beat an average day by 2.45% a year, but that is less than half of what they did before publication.

  2. Turn of the month, 2x fund Failed

    Hold VOO always, and switch to a 2x S&P 500 fund for the days around each turn of the month.

    In a taxable account every monthly sale is taxed short-term. In a retirement account it came out ahead only by carrying 2x risk (worst drop −63% vs −55%); the timing alone was t 1.54, far under the bar.

  3. Fed announcement days, 2x fund Failed

    Hold VOO, and switch to a 2x fund on scheduled Federal Reserve announcement days. Judged on 2015–2026.

    The drift these days used to show fell from +2.67% a year of timing to +0.19% after it was published: about 2 basis points a day.

  4. Day before market holidays, 2x fund Failed

    Hold VOO, and switch to a 2x fund on the trading day before each market holiday. Judged on 1991–2026.

    The only calendar effect still faintly alive, but too small for its noise: short-term tax takes it in a taxable account, and in a retirement account its timing is t 1.09.

  5. Buy short dips on SPY (RSI-2) Failed

    A published mean-reversion rule: buy the S&P 500 after a sharp two-day drop, sell on the bounce, cash otherwise. Judged on 2009–2026.

    In the market only about one day in nine, so it gave up 12.5% a year of market return; the timing it kept (+2.87%) was about what its trades cost.

Trend following2 ideas Months in cash give up the market’s return, and the published timing edge faded after publication.

  1. 2x fund above the 200-day average Failed

    Hold a 2x S&P 500 fund while the index is above its 200-day average, T-bills below (a published rule). Judged on 2015–2026, after publication.

    It tied VOO after tax (13.13% vs 13.15% a year). Its whole gain was leverage; the timing itself went from +5.73% a year on the paper’s years to −3.63% since.

  2. 10-month trend on the market Failed

    Hold the market above its 10-month average, T-bills below. Judged on 2007–2026.

    The months in T-bills gave up 2.24% a year of market return; the timing that was supposed to make up for it fell to +0.09% a year.

Momentum and rotation2 ideas Rotation runs up short-term tax, and large-company momentum faded to about zero after it was published.

  1. Momentum rotation, run by hand Failed

    Rotate into the recent winners each month (stand-in: the top momentum tenth of US stocks). Judged on 1993–2026.

    The paper edge (+3.85% a year) is real, but short-term tax on a hand-run rotation (4.09% a year) eats all of it.

  2. Large-company momentum (MTUM) Failed

    Hold a large-company momentum fund, and its paper portfolio. Judged on 2013–2026.

    The paper edge fell from +3.88% a year before publication to about zero. The fund did a bit better than its paper portfolio, but at t 0.48 it is indistinguishable from luck.

Factor funds4 ideas The funds you can buy trailed their paper portfolios, and their large swings against VOO would need decades to prove anything.

  1. Profitability / quality funds Failed

    Hold a quality fund and never trade it: QUAL, SPHQ and DURPX, each on its own live record.

    The paper profitability portfolio was ahead, but all three buyable funds trailed VOO; they trailed the paper portfolio by 2.6–3.2% a year.

  2. Low volatility Failed

    Tilt to the least volatile stocks (paper portfolio; funds USMV, SPLV). Judged on 2001–2026.

    Its edge is per unit of risk only: in raw return it never beat the market, and the funds trailed the paper by 3.65% a year.

  3. Small value Failed

    Tilt to small, cheap companies (paper portfolio; funds DFSVX, IJS, VBR, AVUV). Judged on 1991–2026 and since 2010.

    The funds delivered it, but its 12% tracking error needs 40+ years to prove even at its pre-1991 strength, and since 2010 it has lost.

  4. Managed futures funds Failed

    Hold managed-futures funds (AQMIX, PQTIX, DBMF, KMLM), each on its record since 2010.

    They carry little stock market exposure, so against VOO they gave up about 15% a year of market return, and their 19–23% tracking error makes any edge unprovable.

Options3 ideas Selling options gives up market exposure and its premium turned negative; daily option trades showed no edge beyond luck.

  1. Covered calls Failed

    Own the index and sell call options on it (Cboe BXM index, no costs). Judged on 2012–2026.

    It gave up stock market exposure (−5.79% a year) and its option premium went negative after 2012.

  2. Put-writing Failed

    Sell put options on the index, backed by cash (Cboe PUT index, no costs). Judged on 2012–2026.

    Same story: −6.17% a year of market exposure given up, and the premium went negative after 2012.

  3. One option trade a day Failed

    Thousands of versions of a same-day option trade on SPY and QQQ, 2025–26.

    The best version reached t 1.36 against a luck bar of 5.40 for that many versions tried: no edge.

Intraday and chase4 ideas Minute-by-minute trading lost before costs, or looked good only on the sample it was picked on.

  1. Chase (red to green), one-minute Failed

    Buy stocks as they turn up from a fall, minute by minute, on 60 days of one-minute bars in 2026.

    Lost 63% in 60 days in a replay even at zero cost: the small moves it chased slightly reverse.

  2. Buy what fell, one-minute Failed

    Buy the day’s fallers on one-minute bars, the same 60 days.

    +27% in the first 40 days, then −6.8% in the next 20 while VOO rose 2.7%: it looked good only on the days it was picked on.

  3. Opening-range breakout Failed

    Buy a break above the first minutes’ range, sell on a break below, 60 days.

    Lost 57% at zero cost.

  4. Intraday momentum Failed

    Follow the morning’s direction into the afternoon, 60 days.

    Lost 10% at zero cost.

Tax-loss harvesting6 ideas Real but small. Each version fell behind for a mechanical reason, and fixing those reasons one at a time led to the one rule that held up.

  1. Harvest, oldest shares sold first Failed

    Harvest losses across 30 industry holdings at a broker that sells the oldest shares first. Judged on 1991–2026 against the same holdings never harvested.

    Selling oldest-first means a harvest must sell a whole position, which is almost never down enough: +0.018% a year, t 1.40. The account’s mechanics cap the edge.

  2. Harvest lot by lot Failed

    The same, picking the losing purchases one by one (needs a broker that allows it).

    +0.063% a year after tax, but the yearly gap swings about four times its size: t 1.35, about 300 years to prove. It was also behind in 2 of 26 ten-year windows.

  3. Harvest VOO into a different index, back after two months Failed

    When VOO is down, sell it and park in a fund on a different index (IWB, IWV, SPTM), then return after two months. Fresh 10-year accounts from 2000.

    Behind in 9–14 of 190–195 ten-year windows, and so was a perfect stand-in: selling the parking fund after the bounce realizes a short-term gain that cancels the loss.

  4. Same, never selling the parking fund at a gain Failed

    Hold the parking fund until it is back at cost instead of selling after the bounce.

    A perfect stand-in was never behind, but a different index drifts away from VOO over the years it is held: IWB 12, IWV 18, SPTM 18 of 195 windows behind.

  5. Same, back to VOO after a year Failed

    Return to VOO once the parking fund’s gain is long-term (also tried VV, SCHX).

    24–28 of 195 windows behind (a perfect stand-in 26): the long-term gain uses up the harvested loss at the lower long-term rate.

  6. Harvest VOO into another S&P 500 fund Held up · waits on a legal question

    The never-sell-at-a-gain rule, with IVV or SPY as the parking fund.

    Never behind in 195 ten-year windows, 2000–2026. Waits on one legal question. See “Held up”.

Screened out before testing

A further 29 ideas were screened out on paper in later rounds, because even their best case could not reach the bar: share lending, index additions and deletions, month-end flows, equal-weight and total-market funds, volatility management, overnight-only exposure, factors published since 2015 and others.

Results are on the years each idea was judged on, after trading costs and tax unless noted. t measures how far a result stands above luck.

The one that held up: harvest VOO losses into another S&P 500 fund

A tax saving, not a way to beat the market. It is the only idea of the 26 that met the test, and it waits on one legal question.

The ruleWhen VOO is below what you paid, sell it and buy another S&P 500 fund (IVV or SPY) the same day. Hold that fund until it is back at cost; never sell it at a gain.
The resultNever behind simply holding VOO in 195 ten-year windows, 2000–2026.
The sizeA small tax saving, a few hundred dollars a year at most. The tax code caps it.

Why it can work

A harvested loss offsets income taxed at a high rate now. The same dollars come back later as a long-term gain taxed at a lower rate, or never, if the shares are held for life. The saving is that rate gap plus the delay.

How it was tested

  • A fresh ten-year account was started every month from 2000, 195 windows in all, each compared with the same deposits simply held in VOO.
  • Four stresses, alone and together: a higher trading cost, a lower tax bracket, the 3.8% investment income tax, and strict wash-sale accounting.
  • With SPY as the parking fund it was never behind under any stress. With IVV, one window fell behind, by a few dollars, only with all four stresses at once.

Why the other versions failed

  • Selling the parking fund after the bounce realizes a short-term gain that cancels the loss.
  • Going back to VOO after a year uses the loss up at the lower long-term rate.
  • A parking fund on a different index drifts away from VOO over the years it is held.
  • Only a fund that tracks the same index as VOO stays close enough. That is why it is IVV or SPY.

The open question

The wash-sale rule (IRC 1091) disallows a loss if “substantially identical” shares are bought within 30 days before or after the sale. VOO and IVV come from different issuers but track the same index, and the IRS has never ruled on whether that makes them substantially identical.

If it does, the losses would be disallowed (moved into the new shares’ cost, deferred rather than lost), the rule would be a tie less a little trading cost, and losses already claimed would need amending.

StatusWaiting on the legal reading, for a tax professional to confirm. If it passes, a paper test at a broker comes next. No real money.

The formula: one equation for “what beats VOO”

Every test measures or corrects a term in one equation: per year, after tax, against holding VOO. The equation then says which test could still matter, so the next test is chosen by the numbers, not by hunch.

Edge = T + S − D + K − C − Tax + H − Q − R − W

Leverage (L) is shown beside it but never counted as edge. Then t = edge ÷ noise × √years.

The terms, in plain words

T

Timing edge × keep

What a timing rule earned on the years it was found, times the share that survives publication. Measured keep: 0.21 over 9 effects (the published figure is 0.42).

S

Tilt

A factor portfolio’s edge on the years it was found, times keep, minus the gap between the paper portfolio and the fund you can actually buy.

D

Cash drag

Share of days out of the market, times the market’s premium over cash (7.8% a year, 1926–2026).

L

Leverage

Shown, never counted as edge. More market exposure earns more on average because it is more risk.

K

Compounding

Bigger swings compound worse: minus half the difference in variance between the strategy and VOO.

C

Trading cost

Trades a year times cost per side: 0.6 basis points a side for funds as the base, 2.6 as the stress.

Tax

Tax

Short-term gains taxed each year at income rates, against VOO’s long-term gain taxed once at the end.

H

Harvest

The gap between the income tax rate and the long-term rate on losses that offset income, plus the delay.

Q

Rebound give-back

Gains realized while a harvested loss is still being used: a short-term gain cancels the loss; a long-term one uses it at the lower rate.

R

Replacement drift

How far the parking fund strays from VOO, times the share parked, over the years it is held.

W

Legal risk

The chance a same-index parking fund is ruled “substantially identical” under the wash-sale rule, which would turn H to about zero. Not measurable from data.

Noise, t

Noise and t

Noise is how much the yearly gap to VOO swings. t = edge ÷ noise × √years. The luck bar rises with every version tried (t 3.27 at 11 versions); Bickford’s own bar is about 4.

How each test changed it

Oct 9

First draft, from tests of 25 strategies

Terms T, S, D, L, K, C, Tax, H and R. Calibrated: the equation predicts all 25 within 2 standard errors. K was added after the 2x-held result.

Oct 12

Harvest into a different index, back after two months

Even a perfect stand-in fell behind in 14 of 195 windows. Added Q: selling after the bounce gives back the loss.

Oct 12

Never sell the parking fund at a gain

The perfect stand-in was never behind; real funds were. So R, drift while held, is the term that binds.

Oct 12

Back to VOO after a year

Even the perfect stand-in fell behind in 26 of 195. Added Q long-term: the parking fund has to be held until the loss is used up.

Oct 12

Park in another S&P 500 fund (IVV, SPY)

R is about zero, so the equation predicted it would hold, and it did: 0 of 195 behind. Only W, the legal question, is left.

Where it points now

  • Price patterns, timing and tilts: with every usable term measured cautiously (one standard error below its estimate), t stays under 1 against a bar of 3.27. The largest usable edges are 0.4–1.5% a year against 3–14% a year of swing; proving one would need about 10 to 100 times more years than exist.
  • So nothing gets tested there until a new term appears whose edge is more than about half its yearly noise.
  • Tax mechanics: the harvest holds up once the give-back and drift terms are removed. What is left is W, a legal question, not a data question.

Where winners’ edges come from

  • Lower cost: already at the floor for a single S&P 500 fund.
  • Tax: yes, through harvesting and using retirement accounts first.
  • Patience: holding through other people’s forced selling earns the market’s return itself, the main return.
  • Paid for risk: leverage earns more on average, but it is risk, not edge.
  • Small, illiquid corners and information or speed: not available to a one-person desk, or not testable honestly.

About this record, and what it is not

Not investment advice. It recommends nothing to anyone. What to hold is your decision, with your own adviser if you have one.

No performance claims. The numbers are the results of historical tests. They say nothing about what any strategy will do next.

No trading. Nothing here connects to any brokerage or account, and nothing here can place a trade or move money.

Not a tax opinion. The one open question in the record is a legal reading of the wash-sale rule, for a tax professional to answer.

Bickford Technologies is a one-founder software company in Philadelphia. It builds small apps that answer one real question each. This record comes from its own research desk; the tests ran on its own tools, on public market data and the broker’s own published figures. The record is updated when the research changes; this version is as of 12 October 2026.

Questions about the record, a number you think is wrong, or an idea you want to see tested: [email protected].

The same record as an iPhone app

Bickford Trader Research puts these pages on your phone: the question, the scoreboard, the rule that held up and the formula. Free. No account, no network connections, nothing stored, nothing collected. See its app page for today’s status and the store link once Apple approves it, and its privacy policy.

See all Bickford apps →