The system I trade

The system I trade

I do not pick stocks. I run a set of rules, every morning, on American stocks. The rules say what to buy and when to sell. I do what they say.

Since November 2026 the system is wider than it used to be. Instead of a couple of rule sets on the largest companies, it now runs several independent ones side by side, each working its own part of the US market — large companies, the Nasdaq names, and smaller companies too. They share one account, they rarely want the same thing on the same day, and that is the whole point.

It still takes about ten minutes a day. Orders go out before the market opens and that’s it. The cycle repeats tomorrow 🙂

Positions are short. A typical one is held about four days. Nothing is a bet on a company; it is a bet that patterns that have repeated for twenty years will repeat again a few thousand more times.

Track record

· long-only US equities · every number below is a simulation of the same rules I run today, on data as it stood on each day.

Growth of 100

Drag to zoom. The lower panel is the same account measured from its own high-water mark — how far below the previous peak it sat, day by day.

Year by year

Return for the calendar year, and the deepest point that year reached below its running peak. The two belong together: the bar is what you were paid, the marker is what you had to sit through to collect it.

Month by month

The deepest drawdowns

Time to the bottom is rarely the hard part. Time back to the previous high is.

What a single trade looks like

Distribution of the result of every closed trade, in per cent of the money put into it. Clipped at ±15% so the tails do not flatten the shape.

How this was measured.
  • Index membership as it stood on each day, including companies that later left the index or stopped trading — no survivorship bias.
  • Orders go out the morning after the signal and fill at the market's price, never at a price chosen with hindsight. Orders that never fill cost their place in the queue.
  • Commissions are charged per share at the rate a retail account pays. The broker's per-order minimum is not modelled: a growing account soon outgrows it, a small one does not. Whole shares only.
  • A limit order counts as filled when the day's price reaches it. In real trading a price can touch an order without filling it, more often on smaller companies.
  • Not included: interest on idle cash, the financing cost of leverage, and slippage beyond the fill rules above. The first understates the result; the second and third overstate it.

What is good about it

Every calendar year has been positive. Twenty-two years on the record, twenty-two up — counting 2026, which is not over yet. The weakest was 2018 at +4.0%. In 2008 the system made +10.2% while the S&P 500 lost 37%; in 2022, +6.8% against −18%.

2026-11-17T14:06:29.9 image/svg+xml Matplotlib v3.1.8, https://matplotlib.org/ 2005 2006 2007 2008 2009 2010 2011 2012 2013 2014 2015 2016 2017 2018 2019 2020 2021 2022 2023 2024 2025 2026* -40% -20% +0% +20% +40% +60% The system I trade S&P 500 with dividends Every calendar year since 2005 22 of 22 years positive. Worst +4.0% (2018); the S&P 500 lost money in 3 of them. * 2026 = to 16 Nov. Backtest of the rules I run · US stocks, index membership as it stood each day · orders sent before the open · commissions per share
Calendar-year returns, the system against the S&P 500 with dividends, 2005 to 2026 (2026 to 16 November)

Almost any twelve months would have been a good twelve months. Of the 252 twelve-month windows on the record, 249 ended higher. The worst lost 7.9%. The same exercise on the S&P 500 with dividends: 85% up, the worst −43%.

2026-11-17T14:06:30 image/svg+xml Matplotlib v3.1.8, https://matplotlib.org/ -40% -20% +0% +20% +40% +60% +80% +100% 0 10 20 30 40 50 number of 12-month windows S&P 500 with dividends: 85% of windows up The system I trade: 99% of windows up Any 12 months you could have picked Return over every 12-month window that ends on a month end, 252 windows. Worst: system -7.9%, S&P 500 -43.4%. Backtest of the rules I run · US stocks, index membership as it stood each day · orders sent before the open · commissions per share
Every 12-month window on the record: 99% up for the system, 85% for the S&P 500 with dividends

Four months in five are positive. 207 of 263. Four red months in a row has happened once, in the summer of 2011; three in a row twice.

The system I trade Combined monthly percent gains - green up, red down; the sign is printed too Jan Feb Mar Apr May Jun Jul Aug Sep Oct Nov Dec Year S&P 500 TR 2005 -2.1 +1.8 +1.1 +1.6 +2.9 +1.1 +5.6 +3.1 +1.3 +2.0 +2.1 +2.7 +25.6 +4.8 2006 +5.0 -0.8 +4.3 +1.0 -3.7 +1.6 +0.2 +0.9 +1.3 +4.4 +2.5 +1.1 +19.1 +15.8 2007 +1.8 +2.3 +0.8 +1.9 +2.0 +4.7 -1.7 +5.9 +0.2 +5.4 +0.7 +2.0 +29.0 +5.1 2008 -11.3 +3.3 +3.6 +3.2 +3.4 +0.4 +6.5 +2.6 -0.7 -9.3 +2.6 +7.1 +10.2 -36.8 2009 -1.0 -2.4 +2.8 +2.9 +8.7 +1.2 +0.8 +9.7 +6.5 +0.2 +6.0 +2.7 +44.6 +26.4 2010 -4.9 +8.2 +2.8 +2.9 -0.4 -1.5 +3.3 -0.1 +3.5 +3.9 +4.0 +1.8 +25.7 +15.1 2011 +5.9 +4.6 +3.9 +2.9 +3.1 -0.8 -1.0 -9.6 -2.3 +4.1 +2.8 +0.1 +13.5 +1.9 2012 +0.6 +1.1 +1.7 +1.9 -6.0 +4.2 +2.0 +4.3 +2.9 -0.5 +0.9 +1.5 +15.1 +16.0 2013 +2.8 +1.0 +2.9 +3.1 +1.2 +4.6 +2.7 +1.2 +2.6 +4.8 +3.9 +0.6 +36.2 +32.3 2014 +1.4 +3.8 +0.7 +6.8 +1.1 +2.7 +0.4 +2.7 -0.1 -6.5 +1.3 +2.4 +17.4 +13.5 2015 +3.6 +2.9 -1.2 +1.7 +4.0 -1.3 +0.5 -1.8 +0.2 +1.7 +0.9 +0.2 +11.7 +1.2 2016 -4.4 +0.8 +2.3 +1.1 +0.1 +4.0 +0.7 -0.7 +1.2 -2.1 +6.3 +0.5 +9.9 +12.0 2017 +1.6 +2.7 +2.9 +0.6 +1.2 +1.2 +1.3 +2.2 +2.9 0.0 +2.6 +3.7 +25.5 +21.7 2018 +1.0 -3.8 +0.8 +2.5 +2.1 -0.7 +1.7 +3.8 +3.6 -5.1 +4.5 -5.8 +4.0 -4.6 2019 +1.7 +1.4 +1.9 +3.5 -4.8 +5.7 +1.9 -4.2 +1.4 +3.6 +4.4 +1.9 +19.5 +31.2 2020 -2.0 +0.4 -3.6 +8.5 +4.7 +8.5 +8.4 +4.6 +1.9 -1.7 +14.5 +3.1 +56.8 +18.3 2021 +3.7 +2.7 +7.5 +1.6 +1.7 +2.0 +4.2 +3.3 -0.3 -0.9 -0.1 +4.7 +34.3 +28.7 2022 -1.3 +2.4 +5.6 -2.7 +1.5 -7.0 +2.3 +2.2 -2.1 +3.9 +3.3 -0.7 +6.8 -18.2 2023 +2.0 +0.2 +1.4 +1.4 +1.2 +4.3 +4.8 -3.6 -0.1 -2.4 +4.5 +2.2 +16.7 +26.2 2024 +1.9 +4.6 +3.3 -1.1 +3.1 -1.1 +3.5 -1.3 +2.3 +1.2 +4.7 -1.0 +21.6 +24.9 2025 +5.3 -2.9 -4.2 +4.4 +1.7 +2.8 +3.6 +5.2 +3.9 +5.2 +4.4 +3.2 +37.2 +17.7 2026 +1.8 +2.1 -2.0 +2.2 +6.2 +5.0 +7.6 +4.6 +1.8 +2.0 -4.6 +29.4 +9.4 US stocks, index membership as it stood each day - orders sent before the open - commissions per share - S&P 500 TR = the S&P 500 ETF with dividends
Month by month, 2005 to 2026, with the year and the S&P 500 with dividends at the right

The bad stretches are shorter than they used to be. The deepest fall, −21%, came in August 2011, and getting back to the previous high took about thirteen months. That is the only time the account has been below a high for more than a year. The system this page used to describe once spent 664 days — almost two years — below a previous high.

It does not depend on any one piece. Several rule sets, different stocks, different reasons to buy. When one of them has a bad year, the others usually do not. No single trade decides anything: a typical position is held four days, and about seven hundred trades a year carry the result.

It does not need me to be right about anything. No view on the economy, no opinion on a company, no forecast. The rules do not know what year it is.

Money is not locked up. On average a bit under half of the account is invested. The rest sits in cash, waiting for a setup worth taking.

2026-11-17T14:06:29.8 image/svg+xml Matplotlib v3.1.8, https://matplotlib.org/ 100 1,000 10,000 The system I trade (22.7% a year) S&P 500 with dividends (10.6% a year) 2004 2008 2012 2016 2020 2024 −40 −20 0 below the last high, % Growth of 100, log scale Top: the account indexed to 100 at the start of 2005. Bottom: how far each sat below its own previous high. Backtest of the rules I run · US stocks, index membership as it stood each day · orders sent before the open · commissions per share
Growth of 100 on a log scale against the S&P 500 with dividends, and how far each sat below its high
The system I trade Strategy stats - a backtest, not a forecast Strategy S&P 500 TR Period 2004-12-31 .. 2026-11-16 2004-12-31 .. 2026-11-16 Annual return (CAGR) 22.7% 10.6% Max drawdown -21.1% -55.2% Return / drawdown (MAR) 1.1 0.2 Sharpe (daily, no risk-free) 1.5 0.6 Sortino 1.8 0.8 Volatility (annualised) 14.4% 18.9% Best / worst month +14.5% / -11.3% +12.7% / -16.5% Positive months 79% 66% Longest time under water 392 days 1772 days Trades (per year) 15261 (698) Win rate 64% Profit factor 1.4 Average trade +0.9% Average win / loss +4.1% / -4.8% Average holding period 7 calendar days US stocks, index membership as it stood each day - orders sent before the open - commissions per share - S&P 500 TR = the S&P 500 ETF with dividends
The statistics side by side with the S&P 500 with dividends

What changed in November 2026

The version this page used to describe ran two rule sets on large companies with bigger positions. The new one keeps them and adds others beside them, with every position smaller, so the account is spread over more independent bets. Measured with the same simulator, on the same data and the same dates, the new book earned a little more per year and, more importantly, fell less deep and came back faster.

2026-11-17T14:06:30 image/svg+xml Matplotlib v3.1.8, https://matplotlib.org/ 100 1,000 10,000 New system (22.7% a year) Previous system (20.8% a year) 2004 2008 2012 2016 2020 2024 −20 −10 0 below the last high, % The new system against the one this page used to show Same simulator, same data, same dates. Top: growth of 100, log scale. Bottom: drawdowns of both. Backtest of the rules I run · US stocks, index membership as it stood each day · orders sent before the open · commissions per share
The new system against the one this page used to show: same simulator, same data, same dates

One more way to see it. Take the daily results, reshuffle them in blocks of about a month, and ask how deep the worst drawdown of each of 2,000 alternative histories would have been. The new system’s bad cases are noticeably less bad: one history in twenty fell further than 29%, against 43% for the previous one. Those are also the numbers I plan around — not the backtest’s own worst point, which is just the one history that happened.

2026-11-17T14:06:39.5 image/svg+xml Matplotlib v3.1.8, https://matplotlib.org/ -60% -50% -40% -30% -20% -10% 0% 0 100 200 300 400 500 600 700 800 simulated histories Previous system: 1 in 20 histories worse than -43% New system: 1 in 20 histories worse than -29% How deep could the worst drawdown have been? 2,000 alternative histories: the same daily results reshuffled in blocks of about a month. Deepest drawdown of each history. Backtest of the rules I run · US stocks, index membership as it stood each day · orders sent before the open · commissions per share
Deepest drawdown in each of 2,000 reshuffled histories, new system against the previous one

What is hard about it

It still has long, dull stretches. Thirteen months to recover from 2011. Five times the account has spent more than six months below a previous high. Most people quit somewhere in month five.

Losing months are normal. About one month in five is red, and some are ugly: −11% in January 2008, −9.6% in August 2011. In the middle of one it feels exactly like the system has stopped working.

You have to show up every single morning. Skipping the ugly days is how you end up with the losses and none of the wins — the good trades are not announced in advance.

Most orders do nothing. Roughly four out of five never fill. If you need action to feel productive, this is the wrong way to trade.

Smaller companies cost more to trade. The simulation counts a limit order as filled when the day’s price reaches it. On smaller companies the gap between buyers and sellers is wider, and a price can touch an order without filling it. A stricter version of the test — an order fills only if the price trades clearly through it — took about a point off the annual return. I watch how my real fills compare with the simulated ones every month.

Past results are exactly that. These numbers come from running today’s rules over old data. They describe what happened, under the conditions listed at the bottom of the report. They are not a promise, and markets are free to stop cooperating.

The honest line

This is a journal of how I trade my own money. It is not a signal service, not advice, and not an invitation to copy anything. If you take one thing from this page, take the thirteen months — not the annual number.

Disclaimer: All content on this site reflects the personal experience and opinions of the author and is provided for informational and educational purposes only. Nothing on this site constitutes financial, investment, or trading advice, and the author is not a licensed financial advisor. Trading stocks and other financial instruments involves substantial risk, including the possible loss of your entire invested capital. Past performance is not indicative of future results. You are solely responsible for your own trading and investment decisions; the author accepts no liability for any losses or profits resulting from the use of this content. Always do your own research and consider consulting a licensed financial professional before making any investment decision.