Here is the entire bot in one sentence:
"Once a month, buy the 10 US stocks that went up the most over the past 12 months. Sell anything that fell out of the top 10. Repeat."
That is it. No charts to stare at. No daily decisions. One question per month, ten names, one transaction round.
Over 8 years (2019-2026), this rule returned +742% while the S&P 500 returned +212% over the same period. +530 percentage points of extra return for doing essentially nothing 28 days out of 30.
This is The Alpha Hunter. Our seventh live bot. Paper-traded with $10,000 virtual capital. Running since 2026-04-17.
Why It Works (In One Sentence)
Stocks that have been going up tend to keep going up — for a while.
It sounds dumb. It feels too easy. It is the most documented anomaly in 100 years of stock-market research.
In 1993 two academics, Narasimhan Jegadeesh and Sheridan Titman, published a paper showing this works on 50 years of US stock data. Since then it has been replicated thousands of times — on US stocks, European stocks, Japanese stocks, emerging markets, commodities, currencies. Same answer everywhere: yesterday's winners outperform yesterday's losers for the next 3-12 months. Then they reverse.
Nobody fully agrees why. The leading theories: investors are slow to update beliefs after good news (you don't notice a company is great until 6 months in), institutions chase performance (mutual funds buy what already worked), and risk premium for going against the herd. Probably all three.
What matters: the effect is real, robust, and old. We just rent it.
What The Bot Actually Does
On the first trading day of each month, the bot does these four things in order:
- Look at every US stock above $100M market cap. That gives a universe of about 4,000 names.
- Calculate each one's 12-month return. Just price today divided by price 12 months ago, minus one. No fancy math.
- Sort the list. Take the top 10. These are this month's winners.
- Compare to last month's holdings.
- Names still in the top 10: keep them.
- Names that fell out: sell them.
- Names that newly entered: buy them with equal weight.
Then it does nothing for the rest of the month. Whatever happens to those 10 stocks in the next 30 days, the bot does not care. Up 20%, down 30%, doesn't matter. It will look again on the first of next month.
Most months you replace 2-4 stocks. Some months you replace zero (the leaders held). Some months you replace 7 or 8 (during regime shifts, the leaderboard reshuffles).
The Backtest, Honest Edition
Tested on 8 years of US stock data (2019-01-01 to 2026-04-17) using yfinance for prices. Realistic 0.05% trading costs per trade (Interactive Brokers retail rate). Monthly rebalance only — no intra-month adjustments.
| Metric | The Alpha Hunter | S&P 500 (HODL) |
|---|---|---|
| Total return | +742% | +212% |
| Annualized return | ~30%/year | ~16%/year |
| Max drawdown | -41% | -34% |
| Trades per year | ~120 (~10/month avg) | 0 |
| Win rate (per stock per holding period) | ~58% | n/a |
The drawdown is real. -41% in the 2022 bear market — slightly worse than the index, because momentum stocks fall harder than the index when the music stops. Anyone running this with real money needs to know: you will see ugly months.
The +30% annualized is also real, but it is the backtest annualized. Real-world deployment will probably shave 2-4 percentage points off (slippage on smaller stocks, tax friction, occasional missed rebalance). So expect more like +24-26%/year if the future looks like the past, which it might not.
Why This Is NOT Just Recency Bias In Disguise
Fair question — "stocks that went up keep going up" sounds like a textbook example of recency bias.
Two reasons it survives the bias check:
1. The lookback is intentional, not emotional. Recency bias is when YOU feel that recent winners are the future because you saw them on Bloomberg. Momentum trading is mechanical: a fixed 12-month lookback, a fixed cutoff at 10 names, a fixed rebalance schedule. No story, no narrative, no "I think semiconductors will keep winning."
2. It works precisely because most people don't do it. If everyone followed momentum, the trade would crowd out and stop working. It hasn't, because most retail investors do the opposite — they take profits on winners (loss aversion) and average into losers ("buying the dip"). That asymmetry leaves room for a mechanical strategy to harvest the persistence.
The strategy IS exposed to one specific failure mode: momentum crashes. After every major bear market, momentum stocks reverse violently. The 2009 reversal cost the strategy 30+ percentage points in 3 months. The 2020 COVID crash cost similar. Anyone running this needs to expect that and not panic-sell.
How It's Different From The Crypto Bots
This is the only bot in the lineup that touches US stocks. Everything else (Watchdog, Tactician, Rotator, Tri-Rotator, Contrarian, Hedge Hopper) trades crypto. The Alpha Hunter is here for one specific reason: diversification.
When BTC drops 50%, all the crypto bots are taking hits — even if they sit in cash, they're not making money. The Alpha Hunter doesn't care. It is a completely separate stream of returns, driven by a completely different market. In 2022, when crypto lost 70%+, US stock momentum was down "only" -25%. That is not great, but it is not catastrophic.
Run all the bots together and your portfolio's drawdown profile smooths out. That is the entire point.
Honest Caveats (Read These)
The friction is bigger than it looks. ~120 trades per year on small stocks means real slippage. Backtest assumes 0.05% cost per trade — in practice on $10k positions in mid-cap names you might pay 0.10-0.20% effective, which compounds away ~2 percentage points per year over time.
Swiss tax classification risk. ~120 trades/year crosses well into the territory where Swiss tax authorities can reclassify gains as professional trading income (taxed up to 40%+) instead of tax-free private wealth gains. If/when this graduates to real capital, you'd want to either accept the haircut, hold via a corporate structure, or limit the rebalance frequency.
Survivorship bias in the universe. I filter on "US stocks above $100M today." That is technically survivorship bias — I'm not testing on stocks that delisted between 2019 and now. For a momentum strategy this matters less than for a value strategy (delisted stocks were usually losers, which momentum would have dropped before they died), but it's worth naming honestly.
It works in plateaus, not always. Like every momentum strategy, this one has multi-year stretches where it underperforms the index. 2022 was one. The 2007-2009 stretch was another (we don't have backtest data that far back, but published academic results show it). If you can't tolerate 1-2 years of "why am I doing this," this isn't your bot.
What Happens At Each Monthly Check
The bot runs once per month, on the first US-market trading day, at 09:35 ET.
- Pull yfinance daily closes for the entire US stock universe (~4,000 tickers).
- For each ticker, compute return over the last 252 trading days (≈ 12 months).
- Sort. Take top 10. Equal-weight.
- Compare to current holdings.
- Liquidate names that fell out. Buy names that newly entered. Done.
- Save state. Send Telegram with the new top 10. Wait until next month.
Total runtime: about 4 minutes. Total decisions per year: 12.
What You Can Watch From Here
- /bots page: Live Alpha Hunter card with current 10 holdings, total return, equity curve. Updates monthly.
- Newsletter: First-of-the-month signal email with the new top 10 and the names entering/exiting.
- Post-mortem ledger: /post-mortems — if this bot underperforms its backtest by more than 30% over 24 months, I retire it publicly.
If after 12 live months the strategy is tracking the backtest direction (anything between +18% and +30% annualized would count as "working"), it graduates from paper to a small real-capital allocation. If it underperforms the S&P 500 over a 24-month live window, it joins Sharpshooter in the post-mortem cemetery.
Related reading:
- The Tactician — The crypto-side momentum sibling on BTC
- 200 Strategies Tested — The Reality Check — Why systematic beats discretionary
- Recency Bias — Why "winners keep winning" sounds wrong but isn't
- Beat HODL or Don't Bother — The benchmark every strategy here must clear
This bot in the post-mortem ledger: See /post-mortems →
Every retired strategy and failed walk-forward — documented publicly.
