Nicolas Darvas turned roughly $25,000 into about $2 million between 1957 and 1959, while touring as a professional ballroom dancer and reading price quotes by telegram from Tokyo, Calcutta and Paris. He wrote it up in How I Made $2,000,000 in the Stock Market (1960), and the mechanism he described, the Darvas box, is still one of the most precisely specified rules any discretionary trader has published. It is a breakout system with a built in trailing stop, drawn as a series of stacked rectangles up the chart.
This guide states the rules as he set them out, separates the part a scanner can compute from the part he did by judgement, and shows what the mechanical core looks like when you run it on real daily bars.
A note on integrity before we go further: InvestingPaths is an independent research site. Darvas died in 1977; nothing here is endorsed by him or his estate. Everything below is our reading of his published method. Trading involves substantial risk of loss, and nothing here is financial advice or a promise of any result. His result was achieved by one person, in one two-year window, in one of the strongest bull markets of the century, and it has never been independently audited.
Why the boxes exist at all
The boxes are usually taught as a chart pattern. They were not designed as one. Darvas was in Asia and Europe on a dance contract, receiving a single price quote per stock per day by telegram, with no chart, no news feed and no ability to react intraday. A method that needed him to watch the tape was not available to him.
So he built one that needed only a few numbers a day and produced an unambiguous instruction: a level to buy above, and a level to sell below. That constraint is the reason the rules are so clean, and it is worth keeping in mind, because most modern write ups quietly add discretion back in.
The rule, step by step
Darvas ran the same sequence on every candidate.
| Step | The rule |
|---|---|
| 1 | The stock must make a new 52-week high. Nothing else qualifies for attention. |
| 2 | Watch whether that high holds. If the stock fails to exceed it over the following sessions (commonly read as three), that high becomes the top of the box. |
| 3 | The floor of the box is the low the stock establishes underneath it without breaking down, in practice the level the advance launched from. |
| 4 | Buy on a break above the box top, on a clear expansion in volume. |
| 5 | Place the stop just under the box floor. That level, not a percentage, defines the risk. |
| 6 | As the stock advances it forms a new box higher up. Raise the stop to the new floor and, if you want, add to the position. |
| 7 | Exit when a box floor breaks. There is no profit target. |
Two things follow from this that are easy to miss.
The stop is structural, not arithmetic. A great deal of trading advice sets risk as a fixed percentage of the entry price. Darvas set it at the price where his reason for being in the trade would have been proven wrong. Those are different quantities, and only the second one adapts to how volatile the stock actually is.
There is no exit on the upside. The system holds until a floor gives way, which means a small number of large winners has to carry it. That is a defining property, not a detail: it is also why the equity curve of any trend-following rule is mostly flat with occasional steps.
The half of the method nobody quotes
Darvas called himself a techno-fundamentalist, and the technical box was only the trigger. His filter for what deserved a box in the first place was fundamental, and it was narrow: he wanted a company whose earnings power was in the process of improving, usually because of a new product or a new industry. His argument was that a stock does not go up for twenty months without a reason, and that the reason should be identifiable before you buy.
He was equally direct about what he ignored: tips, brokers' recommendations, dividends, and the price he personally paid for the stock.
Leaving this half out turns the method into "buy 52-week-high breakouts", which is a different and much weaker system. Any write up that gives you the boxes without the earnings filter has given you half a strategy.
What a scanner can and cannot compute
The box is often described as fully mechanical. It is not, and the ambiguity sits in step 2.
- How long must the high hold? Three sessions is the usual reading, but the book describes watching a stock behave, not counting bars. Different implementations pick different windows, and they produce different boxes on the same chart.
- Where exactly is the floor? The top is unambiguous, it is a specific high. The floor is "the level it did not break", which on a real chart is often a judgement between two or three candidates.
- How much volume is "heavy"? He was emphatic that breakouts needed volume expansion, and equally silent on a multiple.
Every automated Darvas indicator you can install resolves those three questions with someone's chosen defaults. That is fine, as long as you know it is a choice and not his rule. It also explains why two Darvas indicators on the same chart frequently disagree.
Building the scan in TC2000
You cannot draw a Darvas box with a formula, but you can build the candidate list the method starts from, which is the part that actually saves time. In TC2000 the pieces are:
| What you want | How to get it |
|---|---|
| New 52-week high | The built in 52-week high condition, or a Personal Criteria Formula comparing C to MAXH252 |
| Price still near that high | C / MAXH252 >= .98, so you catch consolidation under a high rather than only the day it prints |
| Volume expansion on the break | V > 2 * AVGV50 as a starting multiple, then tune it |
| Liquidity floor | V > 100000 and a minimum price, so the list is tradeable |
Custom formula conditions of this kind are a Premium feature; the built in 52-week high scan is not. Our TC2000 pricing breakdown covers which tier does what, and there is a student discount worth checking before you pay full price.
Two honest notes about doing it this way. The scan gives you stocks at highs; it does not draw the box, so the entry and stop levels are still read off the chart by eye. And a 52-week-high list in a strong market is long, which is exactly why Darvas had a fundamental filter sitting in front of it.
Testing the mechanical core
The part of Darvas that is testable is the skeleton: enter on a break above a recent range high, exit on a break below a recent range low, let the exit trail upward on its own. That is a Donchian channel breakout, and it is in our free backtester as the Donchian channel break rule, with a ready made version at Donchian breakout that you can run without an account.
Set the entry lookback wide enough to approximate a 52-week high, add Volume vs its own average for his volume condition, and Distance from the 52-week high or low to keep the list near highs. What you get is not a Darvas box, and we are not going to pretend otherwise. What you get is an honest test of the claim underneath it, which is that buying strength at new highs and trailing a stop under structure beat holding the same asset.
Two limits worth stating plainly before you read any number it gives you:
- The boxes are not reproduced. Our exits trail a channel low on a fixed lookback; his trailed a hand drawn floor that moved in steps. On the same stock these part company at exactly the moments that matter, which is the first pullback after a fast run.
- The fundamental half is not in the test at all. A pure breakout test tells you about breakouts. Darvas's own claim was that breakouts in companies with improving earnings behave differently from breakouts in general, and that is a separate question. Our value screener and the momentum screener come at the two halves from different sides.
An honest look at the record
The $2 million is the reason anybody reads Darvas, and it deserves more scrutiny than it usually gets.
- The window was two years, in a historic bull market. 1957 to 1959 was an extraordinary period for exactly the kind of speculative growth names he traded. A trend following method needs trends, and he had them.
- The account is his own. There is no audited record, and the book was published by a man who then sold a great many books. That does not make it false; it means it is testimony rather than evidence.
- He was concentrated. The returns came from very few positions taken very large. The same method run with the same rules and eight positions produces a completely different, and much less spectacular, distribution.
- The universe was smaller and slower. Fewer listed companies, wider spreads, no algorithmic participation, and a daily telegram instead of a live feed. The last of those was a constraint he turned into an advantage, and it is the one modern traders cannot easily reproduce, because nothing stops you from looking.
None of this argues that box theory does not work. It argues that the $2 million is not the evidence, and the only way to find out what the rules do is to run them on prices you did not choose in advance.
Frequently asked questions
What is the Darvas box in one sentence? A rectangle drawn between a stock's recent high and the level it has held above, which you buy a break above and sell a break below, with the rectangle redrawn higher as the stock advances.
Does the Darvas box strategy still work? The mechanical skeleton, buying breakouts to new highs and trailing a stop under structure, is a trend following rule and behaves like one: long flat periods punctuated by a few large gains, and worse results in choppy, range bound markets. You can test that half yourself on daily bars in our backtester. The discretionary half cannot be tested the same way.
What time frame did Darvas use? Daily closing prices, because that was all a telegram delivered. The method is usually run on daily bars for that reason; the boxes are not designed for intraday charts.
Is a Darvas box the same as a Donchian channel? No, but they are close relatives. A Donchian channel is the highest high and lowest low of a fixed lookback, recomputed every bar. A Darvas box is a specific high and a specific floor that stay fixed until a new box forms. The Donchian version is what makes the idea testable; the difference shows up on fast pullbacks.
Which screener do I need to find Darvas setups? Any scanner that can find new 52-week highs with a volume filter, which includes the built in tools in TC2000 and most competitors. The box itself is drawn on the chart afterwards, so no scanner finds it for you.
Run the mechanical half yourself
The rules above are worth more once you have seen what they do to an equity curve rather than to a diagram. The Donchian breakout template is the closest expressible relative, free and with no account needed, and the backtester lets you add the volume and 52-week-high conditions on top of it.