Daily MACD Death Cross Shorts on US Stocks|3 Exits Compared
ENTRIQ backtest of shorting from the day after a MACD(12, 26, 9) death cross on daily charts of 100 US stocks (5–56 years each), with 3 exit patterns.
These are the results of backtesting 100 stocks in ENTRIQ (5 to 56 years of data depending on the stock) (Dataset v1.0).
What is a MACD death cross?
MACD is a trend indicator made up of the difference between a short-term and a long-term exponential moving average (the MACD line) and an exponential moving average of that difference (the signal line). When the MACD line crosses the signal line from above (a death cross), it is used as a guide suggesting that market momentum may be turning downward.
In this article, we examine historical price movements after opening a short position on the next trading day after this death cross. The test of buying (going long) on a MACD golden cross is covered in a separate article. Crosses between moving averages of the price itself are covered in Golden Cross & Death Cross|Backtest Your MA Signals.
The direction of P/L in short selling: A short trade is profitable when you can buy to cover at a price lower than the entry price. The more the stock rises, the larger the loss, and because there is no theoretical upper limit to a price rise, a single trade's loss can far exceed -100%. A largest loss below -100% is not a calculation error. ENTRIQ's chart replay also lets you practice short selling, so you can check the direction of P/L hands-on.
Test conditions
- Universe: 100 US stocks selected from the current S&P 500 constituents, taking the largest by market cap with up to 15 stocks per sector
- Period: The full history available for each stock (varies by stock). Only stocks with at least 5 years of data are included
- Timeframe: Daily (split-adjusted)
- MACD setting: Short-term 12, long-term 26, signal 9
- Entry: Sell short at the open on the next trading day after the MACD line crosses below the signal line
- Concurrent positions: New signals in the same stock while a position is open are skipped (only one position per stock)
Exit patterns
| Name | Exit condition |
|---|---|
| A_signal | Buy to cover at the open on the next trading day after the MACD line crosses above the signal line (golden cross) (if it doesn't, buy to cover at the close on the 60th trading day). No stop loss |
| B_10bars | Buy to cover at the close on the 10th trading day after entry. No stop loss |
| C_stop5 | Stop loss (buy to cover) when the price rises 5% from the entry price. If the stop isn't hit, buy to cover at the close on the 10th trading day. If the stock gaps up at the open, the fill is at that open price, so a single trade's loss can exceed -5% |
About survivorship bias
This test uses current S&P 500 constituents, so companies that were delisted or removed from the index are not included. For a short-selling test, this bias may work in the opposite direction from a long test: stocks that fell sharply and later dropped out are excluded, so the short results shown here may look less favorable than they would if those stocks were included.
The market regimes are classified as follows.
- Uptrend: The close is above SMA200 and SMA200 is higher than it was 20 trading days earlier
- Downtrend: The opposite
- Other: Range
- High volatility: ATR(14) divided by the close is in the top 20% over the stock's full history
- Low volatility: The bottom 20%
- Near earnings: The 5 trading days before and after an earnings date (earnings date data only from 2008 onward)
Overall results
| Metric | A_signal | B_10bars | C_stop5 |
|---|---|---|---|
| Sample size | 38,009 | 32,940 | 34,560 |
| Share of trades closed in profit | 29.2% | 43.6% | 39.3% |
| Average P/L | -1.04% | -0.80% | -0.54% |
| Average gain on trades closed in profit | +6.18% | +4.53% | +4.58% |
| Average loss on trades closed at a loss | -4.07% | -5.02% | -3.92% |
| Profit factor | 0.64 | 0.71 | 0.77 |
| Largest loss (single trade) | -299.0% | -156.0% | -46.4% |
| Median MAE (maximum adverse excursion) | -3.08% | -3.58% | -3.62% |
| Longest losing streak (within a stock) | 30 | 19 | 23 |
| Average holding period | 12.1 trading days | 10.0 trading days | 8.2 trading days |
Note: Past sample (N=38,009, N=32,940, N=34,560); not indicative of future results.
What the comparison of the 3 patterns shows
The share of trades closed in profit was highest in pattern B, next in C and lowest in A; the pattern that most often closed in profit was pattern B, with its fixed 10-trading-day exit. On the other hand, average P/L and profit factor were highest in pattern C, next in B and lowest in A, so in terms of overall trade results, pattern C's losses were relatively smaller.
However, in all patterns average P/L was negative and profit factor was below 1, confirming that mechanically shorting on a MACD death cross alone did not generate an overall profit in this historical sample.
The largest loss (single trade) was largest (most negative) in pattern A, next in B and smallest in C. Pattern A, which holds until the MACD golden cross, has no stop loss; its largest loss was also the largest, and with no stop loss, losses grew substantially in some cases when a rise continued. Cases where the largest loss fell far below -100% were also observed, a figure that shows a risk specific to short selling (no theoretical upper limit to a price rise).
The median MAE (the median of the largest adverse move experienced after entry) was most negative in pattern C, next in B and least negative in A.
The average holding period was longest in pattern A (12.1 trading days), next in B (10.0 trading days) and shortest in C (8.2 trading days); pattern A holds until MACD recovers, so its holding period was the longest.
Results by trend
Pattern A (hold until the MACD signal turns)
| Trend | N | Share of trades closed in profit | Average P/L | PF | Largest loss |
|---|---|---|---|---|---|
| Uptrend | 22,564 | 28.6% | -1.14% | 0.59 | -54.0% |
| Downtrend | 8,237 | 30.5% | -0.81% | 0.73 | -299.0% |
| Range | 6,489 | 29.8% | -0.94% | 0.66 | -55.0% |
Note: Past sample (N=22,564, N=8,237, N=6,489); not indicative of future results.
In pattern A, the share of trades closed in profit, average P/L and profit factor were all highest in downtrends, next in ranges and lowest in uptrends. In uptrends, where short trades go against the rising trend, average P/L was the lowest (most negative) and the average holding period was 13.1 trading days.
Pattern C (stop loss on a 5% rise + exit after 10 trading days)
| Trend | N | Share of trades closed in profit | Average P/L | PF | Largest loss |
|---|---|---|---|---|---|
| Uptrend | 20,407 | 39.9% | -0.57% | 0.74 | -29.8% |
| Downtrend | 7,555 | 38.1% | -0.50% | 0.81 | -46.4% |
| Range | 5,920 | 39.3% | -0.51% | 0.78 | -43.4% |
Note: Past sample (N=20,407, N=7,555, N=5,920); not indicative of future results.
In pattern C, the order by trend for the share of trades closed in profit was reversed: highest in uptrends, next in ranges and lowest in downtrends. Average P/L and profit factor, however, were highest in downtrends, next in ranges and lowest in uptrends. One possible factor is that, although trades closed in profit more often in uptrends, the average gain on profitable trades was larger in downtrends (+5.57% vs +4.15% in uptrends).
Note also that pattern B's average P/L was highest in ranges, next in uptrends and lowest in downtrends, a different direction from patterns A and C. The 3 patterns cannot be summarized as "average P/L was better in downtrends for all of them."
Results by volatility (pattern A)
| Volatility | N | Share of trades closed in profit | Average P/L | PF | Median MAE |
|---|---|---|---|---|---|
| Low volatility | 7,780 | 29.2% | -0.59% | 0.69 | -2.18% |
| Normal | 22,872 | 29.5% | -0.91% | 0.66 | -3.08% |
| High volatility | 7,357 | 28.5% | -1.89% | 0.56 | -4.82% |
Note: Past sample (N=7,780, N=22,872, N=7,357); not indicative of future results.
In pattern A, both average P/L and profit factor were highest in low volatility, next in normal and lowest in high volatility. In high volatility, the average loss on trades closed at a loss was also larger (more negative) at -6.10%, and the median MAE was also the most negative (-4.82%). The -299.0% largest loss occurred in the high-volatility group, and short positions held without a stop loss in high-volatility conditions carried particularly large risk in this data.
Results around earnings (pattern A)
| Distance from earnings | N | Share of trades closed in profit | Average P/L | PF |
|---|---|---|---|---|
| Away from earnings | 13,362 | 28.7% | -0.97% | 0.63 |
| Near earnings | 3,055 | 33.4% | -0.97% | 0.65 |
Note: Past sample (N=13,362, N=3,055); not indicative of future results. Earnings date data is available only from 2008 onward.
The share of trades closed in profit and profit factor were higher near earnings than away from earnings. Average P/L was the same near earnings (-0.97%) and away from earnings (-0.97%). The average loss on trades closed at a loss was larger (more negative) near earnings (-4.22%) than away from earnings (-3.67%). The largest loss, on the other hand, was larger (more negative) away from earnings.
Observing periods when losses grew on shorts in uptrends
A point worth highlighting in this test is periods when a short was entered on a death cross during an uptrend and held without a stop loss. In pattern A's uptrends, the average holding period was the longest at 13.1 trading days, and average P/L was also the lowest (most negative).
In uptrends (the close is above SMA200 and SMA200 is rising), there are cases where a death cross occurs but the price only makes a short-term pullback and then continues to rise. In pattern A, which holds without a stop loss, cases where losses grew as unrealized losses were carried until MACD turned were observed. This is not a case of "the stop loss coming too late"; it is a property of the rule itself, which holds without a stop loss.
As mentioned in Why Backtesting Builds Confidence|Judge Drawdown by Rules, checking periods when losses grew in historical data helps you think about how to handle them in actual trading.
Limitations and caveats
- Survivorship bias: Because the test is based on current S&P 500 constituents, stocks that were delisted or removed from the index in the past are not included.
- Slippage and commissions: Not included in the test. In actual trading, fills at the open may not match the assumption.
- Gaps: In pattern C, if the stock gaps up sharply at the open, the stop loss can result in a loss beyond 5% (because the fill is at that open price).
- Fixed conditions: The MACD setting (12, 26, 9) and exit conditions are fixed. Changing the settings or stocks changes the results.
- Cost of short selling: Actual short selling incurs costs such as stock borrow fees, which are not included in the test.
Please also read Why Backtests Fail in Live Trading.
Summary
This article organized the results of short selling from the next trading day after a MACD(12, 26, 9) death cross, tested with 3 exit patterns.
- In all patterns, average P/L was negative and profit factor was below 1
- The share of trades closed in profit was highest in pattern B (fixed 10 trading days), while pattern C (5% stop loss + 10 trading days) had the least negative average P/L and the highest profit factor
- Pattern A recorded the largest single-trade loss; in this data, average P/L was lowest in uptrends and the -299.0% loss occurred in the high-volatility group
- By trend, average P/L was highest (closest to zero) in downtrends for patterns A and C, while for pattern B it was highest in ranges, so the direction differed by pattern
You can check how results change when conditions change by narrowing down to your own stocks and periods with ENTRIQ's Chart Replay. You can accumulate and review your test results with Strategy Tags.
Frequently asked questions (FAQ)
Q. Is a MACD death cross a basis for short selling?
A. In this test, average P/L was negative in all patterns. It confirms that mechanically shorting on a death cross alone was not a condition that accumulated profits in historical data, but what additional conditions might change that is outside the scope of this test.
Q. Why does pattern A, which has no stop loss, have a largest loss beyond -100%?
A. In a short trade, the loss grows as the stock rises above the entry price. Because there is no theoretical upper limit to a price rise, if the stock rises sharply before you buy to cover, the loss can exceed -100%. This is not a calculation error.
Q. Do the results change if limited to downtrends?
A. In pattern A, the share of trades closed in profit, average P/L and profit factor were all highest in downtrends. In pattern C, average P/L and profit factor were highest in downtrends, but the share closed in profit was the lowest. Even in downtrends, profit factor stayed below 1 (A 0.73, C 0.81), so restricting the test to downtrends did not produce an overall profit in this historical sample.
Q. Where can I see the test of buying on a MACD golden cross?
A. The test of buying on a MACD golden cross is covered in a separate article. For crosses between moving averages of the price itself, see Golden Cross & Death Cross|Backtest Your MA Signals.
This article is provided for informational purposes to examine and organize historical data and does not recommend any specific trade. The figures shown are past samples and do not indicate future results. All investment decisions are your own responsibility.
Test the same condition used in this article on symbols of your choice.
Backtest this conditionReproduce this validation yourself
With ENTRIQ's chart replay, you can trade through past charts using the same rules.
Try the demo — no sign-upNo sign-up or card needed. Runs right in your browser. Start 14-day free trial →
Validate your method on past charts
The same validation in this article can be reproduced with your own method. On past charts, do the classics really work?
- ✓ Replay past charts
- ✓ AI feedback on your trades
- ✓ Validate methods with statistics
No sign-up or card needed. Runs right in your browser.