TrendStack
Look

Algorithmic trading · US equities

Trade the trend. Cut what fails. Test everything.

TrendStack is a fully automated trading system for liquid US stocks and ETFs, built by a former Managing Director of Emerging Markets Trading. This site shows how it was built and tested, how it performs, and a weekly view on markets.

Read the research story About the founder

Ahead of buy and hold

$10,000 invested in January 2016, simulated after costs.

Growth of $10,000 from January 2016 to November 2026: TrendStack simulation against buying and holding the Nasdaq-100 and the S&P 500
Jan 2016 – Nov 2026 · indices with dividendsExplore the full record ↓

About

Cira Jovanovic

Twenty years in trading, sales, treasury and asset-liability management, more than fifteen of them leading teams across London, New York, Munich, Vienna and Central and Eastern Europe.

At UniCredit I ran global emerging-markets trading, institutional sales and, finally, the bank's whole CEE markets business from London. That meant managing real risk every day, sitting on ALCO and credit committees, and working directly with the most senior executives, from CEOs and CFOs to finance ministers, central bank governors and debt management offices.

Since 2022 I have traded my own capital across FX, rates, credit and equities, and designed and built TrendStack from the ground up. It is a fully automated system: it reads market data through the trading day, decides what to buy and sell across a universe of around 100 liquid US stocks and ETFs, places the orders with the broker, sets a protective stop on every position and checks its own books against the broker's at the end of each day. No trade is entered by hand.

Behind it sits a research pipeline built with modern AI tools: ten years of minute-level market data, a test harness that runs every idea under the same strict rules, and a live engine that runs the very same code as the research. It brings a trading desk's risk discipline to quantitative research and automated execution.

Markets and risk leadership

Ran a trading desk in London, New York and Munich across FX, rates, credit and structured products, with risk taken and managed every day.

Building and transforming businesses

Led major change projects, greenfield banking operations, system integrations and new product launches, and reshaped businesses for profitability.

Applied AI and quantitative research

Designed and built TrendStack end to end: data, research pipeline, statistical testing and live automated execution through a broker API, using AI tools throughout.

Senior client and board presence

Partnered with the most senior executives, from CEOs and CFOs to ministers of finance and central bank governors, on financing and hedging, and presented to boards and committees.

The system

Follow established trends, step aside quickly when they fail

TrendStack is an algorithmic trading system: computer code decides every trade and sends it to the broker. It trades liquid US stocks and ETFs, long only, with no human in the loop. Positions last from hours to a few weeks. The exact rules stay private; the principles below are what they are built on.

Trend, not prediction

The system does not forecast turning points. It joins moves that are already under way and stays while they last.

Losses defined at entry

Every position carries a protective stop from the moment it is filled. Most trades are small losses; a minority of longer trades carries the result.

No leverage

The portfolio never holds more than its own value. Position sizes are equal and capped.

Risk adapts to the market

When the broad market weakens, the system becomes more selective about what it buys.

Every rule earns its place

A rule goes live only after passing tests written down before the results were seen. Most ideas fail and are dropped.

One code base

The live system runs the same strategy code as the research. What was tested is what trades.

What to expect

It often trails in calm rising markets and earns its lead in the hard ones.

Bull markets

When the index rises smoothly, simply holding it is hard to beat, and TrendStack usually lags. In 2019, 2021 and 2024 it made 14.6%, 23.0% and 13.2% while the S&P 500 made 31.1%, 28.7% and 24.9%. When a rally comes with strong, sustained trends it can lead instead, as in 2020 (+63.9%) and 2025 (+32.9%).

Falls and bear markets

When markets turn, it cuts losing positions quickly and protects the gains. In 2022 it made +4.9% while the S&P 500 lost 18.2% and the Nasdaq-100 lost 32.4%. In 2018 it made +3.0% against −4.9%.

Over a full cycle

Avoiding the big falls is what moves it ahead. Its deepest drop was about half the indices' (16.7% against 33.8% and 35.0%), so it starts each recovery from a higher base. Over 2016–2026, $10,000 grew to $88,594 against $43,050 in the S&P 500 (simulated).

The research

How TrendStack was built, and how hard I tried to break it

Most published trading results are the best of many tries. This is the story of the tries, including the ones that failed.

  1. Chapter 1
    The question

    Can discipline be automated?

    On a trading desk the edge rarely comes from clever predictions. It comes from process: cutting losses early, sizing positions sensibly and not overriding the rules when it hurts. I wanted to know whether that process could run on its own, without leverage, and do better than simply holding the index.

  2. Chapter 2
    The first attempt

    Dozens of strategies, and a hard lesson

    The first version traded dozens of intraday strategies at once: gap fades, VWAP plays, breakouts, momentum and mean reversion. Backtests looked good and some ran live. Tested honestly, most of the edges turned out to be luck, hindsight or trading costs.

    The lesson: the test is more important than the strategy.

  3. Chapter 3
    Starting again

    Rules for the rules

    In autumn 2026 I started again from zero, with every idea tested since the beginning written up and re-run under one strict process. Ten years of minute-level data on around 100 liquid US stocks and ETFs. Pass rules written down before any result was run. A development period split into two halves, 2016–2020 and 2021–2024, where an idea had to win in both. Twenty randomised orderings of the signals instead of one lucky sequence. And January 2025 onwards sealed, opened only once a rule was fixed.

  4. Chapter 4
    What survived

    A robust core, refined layer by layer

    Sector rotation, market-breadth timing, sentiment gauges, daily swing systems and intraday patterns all went through the same process. The design that came through was a trend-following engine that held up in both halves of the development period and in the sealed years.

    On top of that core, four further layers earned their place, each tested on its own against a pre-written bar: a market-condition filter that tightens what the system buys when the broad market weakens, a volatility screen on which stocks it trades, a calibrated stop width, and position sizing that adapts to stretched markets. Dozens of other refinements were tested the same way and thrown out. Many looked excellent in one period and fell apart in the next; without the sealed years, some of them would have gone live.

  5. Chapter 5
    Trying to break it

    Stress-testing my own result

    A backtest on today's stock list carries hidden hindsight: the list contains stocks we now know did well. So I re-ran the rules on a list chosen each day using only what was known at the time. Returns fell to 14.5% a year over 2016–2024 and 2.9% a year over 2025–2026. That is the honest lower bound, and the reason the headline figures below should be read as an upper one.

    The system also stayed profitable with trading costs doubled, and its entry timing beat random entries with the same holding times in the development years, by a smaller margin since 2025.

  6. Chapter 6
    Going live

    Real money, same code, a written exit

    TrendStack has traded real capital since November 2026, running exactly the code that was tested. Live fills are checked against the model. There is also a written rule for when to stop: a full review at 10% below the live high, and no new positions at 15%.

Simulated results

What $10,000 would have become

Growth of $10,000 from January 2016, after trading costs, against simply buying and holding the Nasdaq-100 or the S&P 500 over the same days. Pick a range or a date to see the value at any point in history.

TrendStack (simulated)Nasdaq-100S&P 500
Daily value of $10,000 in the TrendStack simulation, the Nasdaq-100 and the S&P 500
Date—hover, tap or pick a date
TrendStack—
Nasdaq-100—
S&P 500—

By calendar year

Year Trend­Stack Nasdaq S&P

PortfolioReturn a yearDeepest drop$10,000 became

Return against the deepest fall

Annual return against deepest drop for each portfolio
—a year, 2016–2020
—a year, 2021–2024
—worst calendar year
—years ahead of the S&P 500

How it was simulated

One ordinary account holding real shares, 30 equal positions, no leverage. Commissions and slippage are deducted from every trade; no interest is earned on idle cash. The simulation was run at a size where costs per trade are realistic and the results are shown per $10,000; a very small account would hold fewer positions and pay proportionally more in commissions.

The test was repeated 20 times with small variations in which trades were taken. All 20 made between 21.1% and 23.5% a year.

The S&P 500 and Nasdaq-100 are the SPY and QQQ funds bought on 4 January 2016 and held, with dividends reinvested, measured on the same days with the same method. Data to 5 November 2026. Returns a year are annualised on actual time elapsed.

Market views

The weekly view

Short notes on what moved, what the trend says, and what I am watching next. Personal views, not advice.

LATEST · 12 NOV 2026

A strong bounce, not yet a new uptrend

Equities
Markets rebounded hard from last Thursday's sell-off, led by semiconductors and tech. SPY and QQQ are still in volatile short-term downtrends. Breadth is mixed: the equal-weight S&P 500 is near its highs, but leadership keeps rotating. Nvidia's results on 17 November are the next big test for the AI trade.
Rates
Hotter inflation and hawkish Fed comments are pushing yields up again. The 10-year Treasury is approaching 5%, TLT has broken below its range, and higher borrowing costs weigh on equities.
Dollar & commodities
A firmer dollar and rising yields pressure gold, now testing $4,000. Crude holds in the mid-to-high $80s on hopes for Strait of Hormuz flows; geopolitical risk remains.
Crypto & volatility
Bitcoin broke key support and is in a downtrend. The VIX eased to the high teens: less panic, but no clear direction.
Takeaway
The recovery is incomplete: indices need to clear recent resistance before this counts as a new uptrend. Tech leads again, but fast rotation leaves the market leaning on catalysts like Nvidia, and rising inflation, yields and the dollar remain the main headwind for risk assets.
  1. 5 Nov

    One of the ugliest days since the 2025 tariff sell-off

    Major indices down 4–6% on the day and 10–15% below their highs. The AI trade hit hardest: chip and memory stocks down 10–15% today and 40–70% from their peaks. Days like this are why discipline matters: step aside when the trend breaks, keep capital intact, and be ready to buy quality at lower prices once the market settles.

  2. 3 Nov

    Waiting for a new trend

    Out of SPY and QQQ for several weeks and in interest-bearing cash. The long-term S&P 500 trend is still up, but short-term conditions are weak. Large tech and semiconductors face heavy institutional selling while money rotates into dividend payers and defensives. Treasury yields eased slightly, a short-term lift for bonds. No new risk until the market stabilises and a fresh trend is confirmed.

  3. 29 Oct

    Fed on hold, pressure stays

    After the FOMC held rates, equities remain under pressure from inflation and doubts about AI profitability. Until a clear trend forms I stay on the sidelines, earning yield on cash. The aim is to follow confirmed trends, not to forecast them.

  4. 15 Oct

    Hot CPI: consolidation, not reversal

    September CPI came in above expectations (3.7% year on year, 0.3% month on month) and markets sold off from recent highs. Treasury yields spiked, then reversed by the close, a sign bond-market pressure may be easing. Crude oil is still trending up, an inflation risk, while a strong dollar and high yields weigh on gold. By the evening: a sharp oversold bounce, but not out of the woods yet.

  5. Oct

    AI valuations under the spotlight

    A tech-led sell-off followed reports that OpenAI's annualised revenue is about $20 billion below earlier estimates. The Nasdaq fell over 1%, with AI and hardware stocks hit hardest. A reminder of how sensitive this market is to AI valuations.

  6. 2 Oct

    New highs, tech leading again

    After months of sideways trading, the Nasdaq and S&P 500 broke to new highs, and tech leadership is strengthening: semiconductors and many of the Magnificent Seven show bullish patterns that could fuel the next leg. Rising rates and record leverage are real risks, but not reasons to stand in front of a rising trend. Participate while the tide rises, protect capital when the trend changes, and let price, not headlines, decide.

Community

Join the conversation

For people who care about how markets work and how to test an idea properly: traders, investors, quants and the curious.

Opens at launch

The weekly view by email

Every new note in your inbox, plus the monthly live-versus-model update once there are enough live trades to report.

Follow on LinkedIn

Shorter takes during the week, research notes and discussion with other market practitioners.

Follow Cira on LinkedIn

Ask a question

Send a question about markets, testing or building a system. A selection is answered in the weekly note, without names.

Send a question
Ideas and methods, not tipsNo trade signals or paid groupsHonest about what failedRespectful disagreement welcome
TrendStack

Contact

Get in touch

For senior roles, speaking, collaboration or questions about the research.

Emailhello@trendstack.uk
LinkedInlinkedin.com/in/cira-jovanovic-1252b17b