Who was Nicolas Darvas?
Darvas (1920–1977) was a Hungarian dancer, self-taught investor and author. He trained as an economist at the University of Budapest. In June 1943, at 23, he fled Hungary with a forged exit visa and fifty pounds sterling (>Wikipedia).
He later toured with his half-sister Julia as a dance team. On the road, he read some 200 books on markets and speculators. His main source for stock ideas was Barron's, usually a week-old copy, and he sent buy and sell stop orders to his New York broker by cable and telegram (>Wikipedia).
After Time magazine covered him in May 1959, he published How I Made $2,000,000 in the Stock Market in 1960.
The record also has a caveat. Time later reported that the New York Attorney General called his story "unqualifiedly false" and could find ascertainable profits of only $216,000. A court then blocked the probe as an "unwarranted invasion of the free press." Investigators also admitted they had not tracked down all his brokerage accounts (>Wikipedia). The exact size of his gains is still disputed.
The box theory in plain English
Wikipedia summarizes Darvas's idea well. "He considered a stock price wave as a series of boxes. When the stock price was confined in a box, he waited. He bought when the price rose out of the box. He simultaneously set a stop-loss just under this trade price."
He also had a filter. Darvas picked a stock "when it made a good advance on strong volume, with favourable fundamental company research" (>Wikipedia).
So the method has three parts:
- Strength first. Look only at stocks already making new highs.
- Wait for the pause. Let price settle into a range, the "box."
- Buy the break, protect it. Enter when price leaves the top of the box, and keep a stop just below.
The rules, step by step
Darvas described boxes in words. Pattern researcher Thomas Bulkowski turned them into testable rules (>Bulkowski, ThePatternSite):
| Step | Rule |
|---|---|
| 1. New high | Find a new 12-month high. |
| 2. Box top | The high of that day becomes the box top if the next three days all have lower highs. They do not need to fall in sequence. They just have to stay below the first day's high. |
| 3. Box bottom | After the top is set, find a low followed by three days of higher lows. That low is the box bottom. |
| 4. Entry | A close above the box top signals a buy. Buy at the next day's open. |
| 5. Exit | A close below the box bottom signals a sell. Exit at the next day's open. |
Two details matter in practice:
- Boxes stack. If price makes a new high but does not close above the box top, you start looking for a new, higher box. The stop can move up as new boxes form.
- Closes beat touches. Darvas triggered on a new high. Bulkowski found that waiting for a close worked better.
What independent testing found
Bulkowski tested these rules on 104 ETFs and 557 stocks from March 12, 2001 to October 1, 2010. He charged $10 commission each way and made no allowance for slippage (>Bulkowski).
The results depended heavily on the timeframe:
| Test | Win rate | Avg gain per trade | Trades |
|---|---|---|---|
| ETFs, weekly, 52-week high (his "Best!" row) | 49% | 10.5% | 262 |
| ETFs, daily, 365-day high | 38% | 0.1% | 540 |
| Stocks, daily, 365-day high | 35% | 0.0% | 3,378 |
| Stocks, weekly, 52-week high | 42% | 6.9% | 2,410 |
His summary is blunt. The setup "fails miserably on the daily scale but works best using exchange traded funds on the weekly scale." On the best row, he reports an average gain of $1,047 on a $10,000 trade, with a hold time of about 10 months. That checks out: $10,000 × 10.5% = $1,050, and the small gap likely comes from rounding in his table.
Drawdowns were real. He reports a maximum drawdown for stocks of 48.5%, and an average per-trade maximum drawdown of 13.7% on the best ETF row.
Three lessons for anyone trading boxes today:
- Win rates are low. Even the best row lost more often than it won. The edge comes from the size of the winners.
- Timeframe matters. On daily stocks, Bulkowski found the entry often came "just before price peaks."
- Costs and slippage were light. His test used $10 commissions and no slippage, so real results would likely be worse.
These are Bulkowski's numbers for his version of the classic rules. They are not results for Loomi Box.
How Loomi Box finds breakouts
Loomi Box is one of four algorithms Loomi.Ai lists alongside VIX Gate, The Affair and Ascent. Loomi's page labels it a "Darvas breakout" and sums it up as "Waits for the box. Buys the break."
Here is Loomi's full description: "A ten-line filter run on the whole universe every close: tight box of highs, volume on the break, regime in favor. It predicts nothing; it reports which names earned a box tonight."
That maps cleanly onto Darvas's own ideas:
| Darvas idea | Loomi Box check |
|---|---|
| Wait while price is "confined in a box" | Tight box of highs |
| A "good advance on strong volume" | Volume on the break |
| He stayed out of bear markets | Regime in favor |
| Buy when price rises out of the box | Reports names that "earned a box" at the close |
Loomi does not publish the ten lines themselves. It also does not publish performance statistics for Loomi Box. There is no win rate, profit factor or certificate for it on the site, so we quote none here.
How to use it on the Loomi.Ai Desk
Loomi's page says the four algorithms "live on the Desk today." Here is a simple routine built on that:
- Check the list after the close. Loomi Box reports which names earned a box that session. Treat each one as a candidate to research.
- Size through VIX Gate. Loomi says VIX Gate shows "the live VIX band, the regime it puts the desk in, and the position size that band allows."
- Cross-check the trend. Ascent sorts stocks into four stages by weekly moving averages and buys only Stage 2. A box breakout inside a Stage 2 trend is closer to the "straight-line run" Bulkowski says the method needs.
- Decide your stop before entry. Darvas set his stop just under the entry. Write down yours before you act.
The Desk costs $8/mo intro, then $20/mo. To run Loomi Box yourself, you can own the algorithm as a Python script for $149 one time, or add Loomi Full Auto for $79 per month (up to 5 active strategies, then $19.99 per month for each strategy after that) and let the bots place the trades.
Put the research to work on your Desk
Explore the tools behind the research, compare the available access, or open your Desk if you already have an account.
FAQ
Does the Darvas box strategy still work? In Bulkowski's 2001–2010 test, it did poorly on daily charts and best on weekly ETF charts, with a 49% win rate and a 10.5% average gain per trade over 262 trades. Results depend on timeframe, costs and market regime.
What timeframe is best for Darvas boxes? In Bulkowski's test, weekly beat daily by a wide margin for both ETFs and stocks.
Where do you put the stop loss in a Darvas box? Darvas set it just under his buy price. Bulkowski's rules exit on a close below the box bottom.
Is the $2 million Darvas story true? It is disputed. New York's Attorney General said it could trace only $216,000 in profits. A court blocked the probe, and investigators admitted they had not found all his accounts.
Loomi.Ai provides educational market research, not financial advice. Loomi publishes no performance statistics for Loomi Box, and the historical tests cited here are hypothetical.
Sources
- >Nicolas Darvas, Wikipedia
- >Bulkowski on the Darvas Box Technique, ThePatternSite (Thomas N. Bulkowski)
- >Loomi.Ai pricing, Loomi.Ai
- >Pricing, Loomi.Ai