Backtesting my post-earnings momentum strategy revealed that one of my most basic trading rules may have been holding it back. Across 146 historical setups, replacing my +9% profit target/-5% stop loss with a -15% catastrophic stop and passive EOD exits more than doubled simulated expectancy, from +1.52% to +3.07% per setup. In this article, I’ll break down how I used my growing trading database to test different exit strategies, what the results taught me about risk management, and why backtesting is essential for finding—and fixing—weaknesses in a trading strategy.
Backtesting can tell you things about your trading strategy that intuition, experience and individual trades simply cannot.
A trading rule can sound perfectly reasonable on paper and still be poorly suited to the actual price behavior of the setups you’re trading.
That’s something I’ve learned firsthand while building and analyzing my own database of post-earnings momentum setups.
Throughout 2026, I’ve been manually tracking post-earnings price reactions and recording variables including first-hour price movement, technical breakouts, earnings results, guidance, maximum favorable excursion (MFE), maximum adverse excursion (MAE), and next-day end-of-day performance.
With that data, I’ve been able to run simulations that apply different exit rules to the exact same historical setups.
In this article, I’m going to show you one of the most important things those experiments have taught me:
The +9% profit target and -5% stop-loss strategy that looked perfectly sensible—and was actually profitable—appears to have been substantially less efficient for my post-earnings momentum strategy than simply giving trades more room to breathe.
The post-earnings setups used for this experiment are available in my Post-Earnings Momentum Database, where you can explore the underlying data used throughout this article.
Quick Answer: Does Backtesting Actually Work?
Yes. Backtesting works because it allows traders to test whether the assumptions behind their trading rules are actually supported by historical data. In my case, I simulated different exit strategies across 146 post-earnings momentum setups using a fixed $1,000 position size. My original +9% profit target/-5% stop-loss strategy generated a simulated $2,214.20 profit, but replacing the -5% stop with a -15% catastrophic stop and otherwise holding until the next-day close increased simulated profit to $4,476.20—a 102.2% improvement. Eliminating the stop entirely increased simulated profit slightly further to $4,614.90, or 108.4% more than my original strategy. That doesn’t prove traders shouldn’t use stop losses. It shows why backtesting matters: my data suggested that a rule I believed was protecting my strategy was actually interfering with its expectancy.
Key Statistics: Backtesting & Refinining Trading Strategies
- 146 setups analyzed: My current backtesting sample includes 146 completed post-earnings momentum setups.
- +$2,214.20 with my original strategy: A +9% profit target and -5% stop loss produced a simulated $2,214.20 profit using fixed $1,000 positions.
- +1.52% average return per setup: That was the average simulated return under the original +9%/-5% exit strategy.
- +$4,476.20 with a -15% catastrophic stop: Allowing trades to run until next-day EOD unless they reached -15% more than doubled simulated profits.
- +3.07% average return per setup: The -15% catastrophic-stop strategy generated roughly twice the expectancy of the original exit strategy.
- +102.2% improvement in simulated profit: Moving from +9%/-5% exits to a -15% catastrophic stop with passive EOD exits increased total simulated profit by 102.2%.
- +$4,614.90 with no stop loss: A passive next-day EOD strategy with no stop produced the highest simulated profit in the sample.
- Only $138.70 separated no-stop and -15% results: The catastrophic stop preserved nearly all of the historical expectancy of the no-stop strategy while still defining an emergency exit.
- 61.6% win rate with the -15% strategy: The simulation produced 90 winners, 52 losers and four breakeven trades.
- 47.3% win rate with +9%/-5% exits: My original strategy produced 69 winners, 76 losers and one breakeven trade.
- 28 stopped-out trades later finished profitable: Twenty-eight setups crossed the -5% stop threshold but ultimately closed the following session above my original entry.
- More expectancy also meant more drawdown: Maximum simulated drawdown increased from approximately $756.50 with +9%/-5% exits to $1,109.30 with the -15% catastrophic stop.
- Backtest overfitting is a documented risk: Bailey et al. (2017) found that repeatedly testing strategy variations can produce impressive historical results that reflect noise rather than a persistent edge.
- Multiple testing can exaggerate performance: Harvey and Liu (2015) argue that backtest results should account for how many different strategy variations have been tested.
- Performance metrics can be misleading: Lo (2002) found that ignoring serial correlation could overstate an annualized hedge-fund Sharpe ratio by as much as 65% in one empirical example.
- Real-world execution matters: Barber and Odean studied 66,465 brokerage households and found the most active traders earned 11.4% annually versus 17.9% for the market during their study period.
What Backtesting My Trading Post-Earnings Momentum Strategy Taught Me
The biggest benefit I’ve gotten from backtesting hasn’t been finding a magical combination of parameters that would have made the most money. It’s been finding weaknesses in rules I previously assumed were helping me.
My post-earnings momentum strategy attempts to identify stocks experiencing significant price discovery immediately following an earnings announcement.
I typically measure the initial reaction using the first hourly candle following the release and track what happens from the close of that candle through the following trading day.
That means these aren’t necessarily quiet stocks.
They’re moving because new fundamental information has entered the market, expectations are being repriced and buyers and sellers are fighting over what that new information means.
Volatility is part of the setup.
That’s important, because a stop loss shouldn’t necessarily be judged by whether it limits losses. The more important question is whether the stop sits outside the normal volatility of the strategy you’re trading.
My backtesting suggested mine didn’t.
My Original +9% Profit Target and -5% Stop-Loss Strategy Worked—But It Had a Weakness
My original exit strategy was simple. After entering a qualifying post-earnings momentum setup at the close of its first hourly earnings candle, I would target a +9% profit while risking -5%.
If neither threshold was reached, the position would eventually be closed at the end of my holding period.
On the surface, I still think this sounds like a perfectly reasonable trading strategy.
My potential profit target was considerably larger than my predetermined loss. Risk was clearly defined before entering the trade. And most importantly, when I simulated those rules across my historical dataset, the strategy was profitable.
Across 146 setups at $1,000 per position, the +9% PT and -5% SL strategy generated approximately $2,214.20.
That’s positive expectancy.
The -5% Stop Was Cutting Off Recovering Trades
Across 146 completed post-earnings momentum setups, my original exit rules frequently removed trades before the underlying momentum had finished playing out.
38.9% of the setups that hit my -5% stop before reaching the +9% target ultimately finished the next-day session profitable relative to my original entry.
But as my database grew, I started noticing something strange.
Some of my losing setups weren’t really failing, per se. They were retracing towards my EMA cloud, hitting my -5% stop loss, but then holding the broader trends and continuing in the same direction as the initial move AFTER I was stopped out.
Because my post-earnings momentum database tracks both MAE and eventual EOD performance rather than simply recording whether my stop was hit, those reversals remained visible in the data.
In fact, 28/146 setups that crossed the -5% stop threshold and stopped me out ultimately finished the next-day session profitable relative to my original entry in the direction of the first-hour earnings candle.
That doesn’t automatically mean the stop was wrong. But it gave me a hypothesis worth testing… What happens if I get out of the way and give these trades more room?
+9%/-5% Vs. -15% Catastrophic Stop and Passive EOD Exits
This is where maintaining detailed data became incredibly useful.
I didn’t need to change my trading rules and spend another year waiting to see what happened. I already had MFE, MAE and EOD performance recorded for my historical setups.
I realized I could simply run an experiment using my recorded setup and outcome data.
For these simulations, I assumed an exact same $1,000 position size and used the exact same historical setups. The only variable I changed was how the trades were allowed to exit.
Here are some of the results and the differences I found between a +9%/-5% and an exit strategy using only a -15% catastrophic stop loss and passive EOD exits.
| Metric | +9% PT / -5% SL | -15% SL / Passive EOD |
|---|---|---|
| Completed setups | 146 | 146 |
| Position size | $1,000 | $1,000 |
| Simulated profit | +$2,214.20 | +$4,476.20 |
| Average return/setup | +1.52% | +3.07% |
| Win rate | 47.3% | 61.6% |
| Winning outcomes | 69 | 90 |
| Losing outcomes | 76 | 52 |
| Profit factor | 1.61 | 2.22 |
| Maximum simulated drawdown | -$756.50 | -$1,109.30 |
The difference between the two strategy is substantial. The catastrophic-stop strategy generated approximately $2,262 more simulated profit, representing an improvement of approximately 102.2%.
It also increased the simulated win rate from 47.3% to 61.6%.
But there’s an extremely important number in that table that shouldn’t be ignored:
Maximum drawdown got worse.
The original exit strategy experienced a maximum simulated drawdown of approximately $756.50. Allowing trades considerably more room increased that figure to approximately $1,109.30.
That’s roughly 47% more drawdown, which can hurt both financially and psychologically.
More Room Increased Both Expectancy and Drawdown
Giving post-earnings momentum setups more room improved simulated returns, but it also increased the amount of capital exposed during losing stretches.
Switching from +9% PT / -5% SL to a -15% catastrophic stop more than doubled simulated profit.
The higher-expectancy strategy also experienced a significantly deeper simulated drawdown.
And that’s exactly why I don’t think the lesson is that stop losses are bad. The lesson is that risk and expectancy have to be evaluated together.
My -5% stop reduced the severity of individual losing trades and helped control drawdowns. But my data suggests that it did so too aggressively. It frequently removed positions during adverse movement that was apparently still within the normal volatility of these post-earnings setups.
The -15% strategy, on the other hand, accepted more volatility in exchange for allowing the underlying edge more room to play out.
In this particular historical sample, that trade-off substantially improved expectancy.
No Stop or a Catastrophic Stop Appears to Fit MY Strategy Better
Next, I took the experiment one step further. What if there were no stop at all?
Instead of closing a position because it reached an arbitrary adverse percentage, I also simulated holding every setup until the next-day EOD exit regardless of its MAE.
While trading without a stop loss is NOT something I recommend other traders do unless they know what they’re doing, this strategy (no stop loss) actually produced the highest historical return of the three simulations:
A total P/L of +$4,614.90.
Compared with $2,214.20 under my original +9%/-5% strategy, that’s an improvement of approximately 108.4%.
-15% Catastrophic Stop vs. No Stop Loss
Both simulations held trades until the next-day EOD unless the catastrophic stop was triggered. The difference in total simulated profit was surprisingly small.
Interestingly, however, the difference between no stop and a -15% catastrophic stop was tiny compared with the difference between either strategy and my original -5% stop.
The no-stop simulation produced just $138.70 more than the -15% catastrophic-stop strategy across 146 setups.
That’s important to me.
Based on the data I’ve collected so far, a -15% catastrophic stop appears to preserve most of the historical expectancy produced by passive exits while still giving me a predetermined point at which a trade has moved dramatically beyond what I’m willing to tolerate.
That’s very different from saying:
“Traders shouldn’t use stop losses.”
I would never draw that conclusion from this experiment.
Trading without risk management can result in enormous losses, particularly when leverage, concentrated positions, illiquid stocks, overnight gaps or unexpected news events are involved.
My conclusion is much narrower:
A -5% stop appears to have been too tight for MY post-earnings momentum strategy during the period I’ve studied.
That’s what backtesting helped me discover. And that’s why this experiment matters beyond the specific numbers.
I started with a trading rule. I collected data. The data exposed a possible weakness. I developed a hypothesis. I tested alternative rules against the same historical setups.
And now I can continue collecting new data to determine whether those results persist.
That’s strategy development. And it’s exactly what I’m using backtesting to help me do.
Additionally, I plan on using similar backtesting simulations to answer specific questions about my trading strategy, including:
- Is -5% normal adverse movement for these setups?
- Does taking profits at +9% unnecessarily cap strong momentum trades?
- How much additional drawdown am I accepting by allowing trades more room?
- Does a catastrophic stop preserve most of the expectancy of passive exits?
Those are questions my database can help answer.
The next 100 or 500 setups will tell me whether the answers remain the same.
What Backtesting My Post-Earnings Momentum Data Has Taught Me So Far
Backtesting my exit strategy is only one experiment I’ve conducted with my growing post-earnings momentum database. As I’ve collected more setups, I’ve been able to test assumptions about first-hour price action, volatility, technical structure, fundamentals, short interest and trade management.
These findings should be treated as observations and working hypotheses, not universal trading rules. My sample is still growing, and one of my goals is to see whether these patterns remain intact as hundreds of additional setups are added.
Key Insights & Hypotheses So Far
- Bigger first-hour moves don’t necessarily produce bigger continuations. My data suggests the size of the initial earnings reaction may tell me more about subsequent risk and volatility than potential reward.
- The 5%–10% first-hour range has shown an attractive risk/reward balance. These setups have historically produced strong continuation while experiencing considerably less adverse movement than many of the largest initial movers.
- Extremely large first-hour moves tend to become much harder to manage. Setups moving 20% or more initially have still produced major winners, but my data suggests they also experience substantially greater MAE and less predictable EOD outcomes.
- Small first-hour moves shouldn’t automatically be ignored. Some relatively modest earnings reactions have developed into surprisingly clean and significant post-earnings trends.
- Technical structure appears useful, but more confirmation isn’t always better. Hourly, four-hour and daily breakouts can help characterize a setup, but my early results haven’t shown that simply stacking every possible technical confirmation guarantees stronger continuation.
- Short interest may identify especially explosive opportunities. Some of the strongest continuation statistics in my sample have appeared among stocks with elevated short interest, although these subsets remain relatively small.
- EPS and revenue beats alone don’t appear to explain post-earnings momentum. The magnitude of an earnings beat has shown little obvious relationship with subsequent MFE, suggesting that expectations, guidance and market positioning may matter more than the headline beat itself.
- Fundamental and price-action alignment may be less important than I originally assumed. Some setups with seemingly conflicting earnings fundamentals have continued strongly in the direction of the initial market reaction.
- The market’s reaction may matter more than my interpretation of the earnings report. One emerging hypothesis is that sustained post-earnings price action can contain information that isn’t obvious from headline EPS and revenue numbers alone.
- Explore the Trading Statistics Hub to understand how different sectors behave across market cycles
- Study real setups inside the Trade Reviews section
- Learn the framework behind high-probability setups in the Post-Earnings Momentum Strategy
Conclusion – The Importance Of Backtesting Trading Strategies
Backtesting didn’t show me that stop losses don’t work.
It showed me that my stop loss may not have worked for my strategy. That’s a much more useful discovery.
A trading rule shouldn’t exist because it sounds disciplined, because another trader uses it or because the numbers look good on paper.
It should exist because you have evidence that it improves the way your strategy actually performs.
Build the strategy. Track the data. Test your assumptions. Refine. Repeat.
If you want to go deeper:
This is how you turn raw market data into repeatable trading edge.
More Trading Statistics…
Frequently Asked Questions
Does backtesting actually work?
Yes, but backtesting is best understood as an experimental tool rather than proof that a strategy will remain profitable. Historical data can help traders estimate expectancy, drawdowns, win rates and how different trading rules interact with their setups. However, backtests can be distorted by overfitting, selection bias, transaction costs and changing market conditions, which is why historical testing should ideally be followed by out-of-sample and forward testing.
What is backtesting in trading?
Backtesting involves applying predetermined trading rules to historical market data to estimate how a strategy would have performed in the past. Depending on the strategy, traders might test entries, exits, stop losses, profit targets, position sizing, technical indicators or other variables.
Why is backtesting important?
Backtesting allows traders to replace assumptions with measurable evidence. A rule that sounds logical doesn’t necessarily improve a strategy. By testing the same rule across a sufficiently large collection of historical setups, traders can determine whether it appears to improve expectancy, reduce risk or produce more consistent outcomes.
Can ChatGPT backtest a trading strategy?
Yes—with the right data. That’s actually how I’ve been running many of the experiments on my post-earnings momentum strategy. I collect and organize the setup data myself, including variables such as entry conditions, MFE, MAE and EOD performance, and then use ChatGPT to process the dataset and simulate different exit rules across the same historical trades. ChatGPT can calculate results such as total simulated P&L, expectancy, win rate, profit factor and drawdown, but the quality of the analysis ultimately depends on the quality and completeness of the underlying data.
Is 100 trades enough for backtesting?
One hundred trades can provide useful preliminary evidence, but there is no universal number of trades that makes a backtest statistically reliable. Sample size should be considered alongside the frequency of the setup, variability of returns, number of strategy parameters being tested and number of different market environments represented in the sample. My current experiment contains 146 completed setups, which I consider large enough to identify patterns worth investigating—not large enough to declare those patterns permanent. I plan to continue collecting data and testing whether these results survive as the sample grows.
What are the limitations of backtesting?
Backtesting tells you what would have happened under a particular set of assumptions using historical data. It cannot tell you with certainty what will happen in the future. Results can be affected by overfitting, survivorship bias, look-ahead bias, inaccurate historical data, changing market regimes, unrealistic fills, bid-ask spreads, slippage, commissions and liquidity. Backtests also can’t perfectly reproduce the psychological pressure involved in managing real money. For these reasons, strong historical performance should be treated as evidence worth investigating rather than a guarantee of future profitability.
What is backtest overfitting?
Backtest overfitting occurs when a strategy is optimized so extensively against historical data that it begins capturing random characteristics of that particular sample instead of a persistent trading edge. Testing enough combinations of indicators, stops, targets and filters will eventually produce impressive historical results simply by chance. This is why researchers including Bailey et al. and Harvey and Liu emphasize the dangers of multiple testing when evaluating investment strategies.
Should traders optimize their stop losses using backtesting?
Backtesting can help determine whether a particular stop-loss level fits the historical behavior of a strategy, but the goal shouldn’t simply be to find whichever stop produced the highest historical profit. Traders should examine how different stops affect expectancy, drawdowns, average losses, tail risk and the frequency of trades that stop out before recovering. A stop should ultimately reflect both the behavior of the setup and the amount of risk the trader is actually capable of accepting.
Does this backtest mean a -15% stop loss is better than a -5% stop loss?
No. It means a -15% catastrophic stop historically performed better for my specific post-earnings momentum strategy and this particular sample of 146 setups. Different strategies have completely different volatility characteristics. A -15% stop could be wildly inappropriate for another strategy, account size or position-sizing model.
Is trading without a stop loss a good strategy?
Not necessarily. My no-stop simulation produced the highest historical return in this particular experiment, but that doesn’t mean trading without a stop is universally safe or advisable. Stop losses are only one component of risk management. Position sizing, leverage, liquidity, diversification and maximum portfolio exposure can be equally important. My experiment suggests that a tight stop may have interfered with my particular strategy’s historical expectancy—not that risk management should be abandoned.
What’s the difference between backtesting and forward testing?
Backtesting applies a strategy to historical observations that already exist. Forward testing applies predetermined rules to new setups as they occur, without knowing the outcome in advance. Forward testing is especially useful after developing or refining a strategy through historical analysis because it provides genuinely unseen observations that can help determine whether the patterns identified during backtesting persist outside the original sample.
References
Bailey, D. H., Borwein, J. M., López de Prado, M., & Zhu, Q. J. (2017). The probability of backtest overfitting. The Journal of Computational Finance, 20(4), 39–69. https://doi.org/10.21314/JCF.2016.322
Barber, B. M., & Odean, T. (2000). Trading is hazardous to your wealth: The common stock investment performance of individual investors. The Journal of Finance, 55(2), 773–806. https://doi.org/10.1111/0022-1082.00226
Harvey, C. R., & Liu, Y. (2015). Backtesting. The Journal of Portfolio Management, 42(1), 13–28. https://doi.org/10.3905/jpm.2015.42.1.013
Laforest, J. (2026a). Post-earnings momentum database. Paper Trading Journal. https://papertradingjournal.com/post-earnings-momentum-database/
Laforest, J. (2026b, August 20). What is post-earnings momentum trading? Paper Trading Journal. https://papertradingjournal.com/2026/08/20/what-is-post-earnings-momentum-trading/
Lo, A. W. (2002). The statistics of Sharpe ratios. Financial Analysts Journal, 58(4), 36–52. https://doi.org/10.2469/faj.v58.n4.2453
Leave a Reply