ATR Strategies to Test

Aug 10, 2007: 11:37 PM CST

Did you know you could develop a simple trading strategy that returns winning trades 80% of the time that uses random entries?! It actually CAN be done.

For those of you who are still reading, a second question that comes up is whether or not this strategy is profitable – the answer is “sometimes.”

This system works across all time frames and across all markets, including futures and stocks. Are you ready to learn what it is?

It’s the Average True Range Test Strategy. Forget moving averages, forget stochastics, forget Fibonacci lines, forget higher time frame confirmation, forget the TICK and TRIN, and forget everything that happens in the economy.

All you need is one indicator: The Average True Range. The ATR measures the open and close – which refers to daily range or volatility – averaged over a specific period, usually 14 periods. The resulting number gives us the likely price range that a security or futures contract is expected to move during one period (day, week, minute, etc). Many traders – myself included – use ATR calculations to place stops or determine position sizing, but that is not the focus of this article.

What if you could base a random-entry trading strategy simply using the ATR measurement? Can it be done?

Let’s take stock ATRR (not a real stock) that trades currently at $50 and shows an ATR reading of $1. This means that in a given day, we can expect ATRR to move $1 up or $1 down as an average move. I mentioned that some traders use the ATR value as a stop… meaning perhaps they will exit a position if a stock trades 2 times the ATR against them (meaning if we entered long at $50, we would take a stop if the price reached $48). Most traders use some other reason to set profit targets – very few of them use the ATR alone to set profit targets.

Let’s do just that and create a simple strategy:

Assume temporarily that we will only go long a given stock, and we will choose a random entry point (no technical analysis) and will hold the stock until it moves in our favor 2 ATR values, or we’ll exit if the price moves against us 2 ATRs. This means we’ll exit with a profit at $52, and exit with a loss at $48. Notice the risk/reward ratio is exactly 1:1. Random chance would say that there’s a 50/50 chance of either occurrence happening, and so over the course of 100 random trades in 100 random stocks, we could expect to break-even on our strategy.

There is a caveat: If we are experiencing a strong bullish environment, the system would be a bit more profitable than average chance because of the upside pressure (we’re more likely to hit 2 ATRs above price rather than get stopped out 2 ATRs below price). The reverse is true for a strongly bearish overall environment. So the overall market condition will shift the ‘breakeven’ probable balance to either the profitable or negative account balance at the end of these 100 trades.


Now let’s play a game.

Assume our profit target becomes 1 ATR above ($51) and 2 ATRs below ($48). Because price is likely to move a single “daily true range” more than it is to move two true ranges, our profitability of the system will rise to a level above 50%, regardless of general market conditions. In other words, we could expect the system now to produce winning trades 65% to 75% of the time (again, the actual value will be dependent on the overall market structure).

With a winning percentage near 70%, does this make for a winning system automatically? No, it does not, because the size of our losers are actually twice the size of our winners. This fact brings the overall profitability of the system back to “breakeven,” and again the eventual results will be dependent on a strong bullish or bearish environment. Either way, the system probably is not one you would like to trade.


Let’s step it up a bit and play for 1 ATR for a profit ($51) and 4 ATRs for a stop-loss ($46). With this system tested over 100 stocks, we may find a “winning percentage” upwards of 80%. Is it a profitable system? Again, no! Why not?

The average size of our losing trades is actually four times the size of our winners now. Let’s trade this system based on the expected outcome of 10 trades in relation to the size of winners or losers (assume we trade 1,000 shares):

80 winning trades bringing in $1,000 each at $80,000 total.
20 losing trades costing us $4,000 each at $80,000 total.

Again, the 8 winning trades bring in $8,000 to the account, while the 2 losing trades cost our account $8,000. This is the definition of “Breakeven”. Again, even though we ‘re right 8 times out of ten, this is not a profitable system.


What if the ATR value went to one half an ATR (0.5xATR) for the profit and 8 ATRs for a stop (a bit excessive). Let’s say the system returned a win percentage of 90%. Let’s run these numbers out of 100 trades:

90 winning trades bringing in 1/2xATR (1/2 an ATR in this example is $500) which totals $45,000.
10 losing trades costing us 8 ATRs, or $80,000.

At this point, our 90% correct system has cost us $45,000!!!

Let’s even run this scenario:

95 winning trades at .5 x ATR ($500 each) for a total of $47,500
5 losing trades at 8 x ATR (costing $8,000 each) costing us a total of $40,000.

The system would return our account $7,500 over 100 trades… but we would burn this all in commission and commission… making it a breakeven system at best and a losing one at the worst.


When developing a trading strategy, it is best to devise a benchmark to base comparisons for your results. Such benchmarks should not be arbitrary, but based on actual market data.

I encourage you to play around with the Average True Range function both for setting stops and for setting targets. The results you get – especially with random entries – should serve as your benchmark to beat.

Restated, if a random system can produce 65% winning trades with a reasonable win/loss ratio, then your system MUST outperform these results, or it’s not worth using your system.

Check back here within the next few days to learn how to add simple filters to the above ‘random entry’ strategies to turn them into profitable strategies.


5 Responses to “ATR Strategies to Test”

  1. jkw Says:

    I looked up the math for this. Assuming a random walk, the probability of going up by b before it goes down by a is a/(a+b). The expected winnings would be ab/(a+b). The expected losses would also be ab/(a+b). The total expected profit is ab/(a+b)-ab/(a+b)=0, regardless of what the values for a and b are. So whatever numbers are chosen, the expected profit before commissions and slippage comes out to 0.

    Are you still planning to say anything about adding filters to generate a positive expected profit?

  2. Corey Rosenbloom Says:

    I appreciate you providing the mathematics behind the logic. Absolutely – without filters or added strategies designed to add alpha, we can expect a zero positive outcome. In other words, due to random chance, it is difficult to conceive a system that produces any profit at all, given random targets and entries.

    I intended to pursue this topic further and absolutely discuss filters (such as trend filters, momentum filters, and volatility filters… as well as potential higher time frame filters) as added methods of generating alpha and creating positive expectancy.

    Unfortunately, the extreme volatility and downside market action we have experienced has forced me to shutter plans to discuss this topic further at the moment, but I will be addressing it more in the future. I felt it more timely to focus on recent market action than this at the moment in the public postings.

    Thank you for your comment and I encourage you to communicate more with additional ideas.

  3. Kal Says:

    Haven’t done the math on this, but intuitively, should using ATR as a loss limiting stop and allowing the the winning trades to increase without limit, but perhaps putting the a trailling limit on profits once established, greatly increase the profitability of a system based on on ATR, keep the downside limited and reduce commissions?

  4. Corey Rosenbloom Says:


    I believe so, yes. I haven’t tested this yet, but I will do so as time permits. Filters would be required to make the system more profitable.

    The above example is a simplified discussion on how traders should not necessarily focus too heavily on the %win portion of a trading system. I’m saying that you can generate random entries and produce an extremely high win ratio, but still have an unprofitable system. Next, I am addressing ideas to make this random entry structure more profitable – but anything we add will decrease the high win rate and increase overall system profitability.

    At the moment, I am only testing hard stops and not utilizing trailing stops. I will be doing so in the future.

  5. Swing Trading Says:

    Interesting post. I have made a twitter post about this. My friends will enjoy reading it also.