The worldly author of finance dwarves — like, Should I Buy That Stock? — happily introduces this Finance Shorties series. Its intention is to cover challenging and useful issues in trading and investing efficiently for immediate usage by both hardworking digging dwarves and humans.
Today: What a stop-loss is in theory, how it behaves in practice, and what the implications of setting conservative and risky stop-loss distances are.
Disclaimer: the following content is for informational purposes only and does not constitute financial advice. Do not take any decisions based solely on my writing.
Stop-loss: What it is used for and how it is commonly seen to behave
After an investor (rather long-term oriented, does thorough fundamental analysis) or a trader (rather short-term, has a great trading system) buys a security, he or she could simply keep it in their portfolio and not take any action for the time he or she wants to keep it. How long?
Our investor might hold on to the security until the position has played out, that is, until it might have reached its intrinsic value and is not considered “cheap” anymore. The trader might have reached the position’s profit target as defined in advance.
But what if the position declines, say, 50 % in value in a relatively short time? For the investor, it’s maybe time to hold on to it — only if he or she is convinced of the analysis done before.
For the trader, however, it’s a huge issue. After a 50 % decline in position, it takes a 100 % increase to make it up again. Moreover, the trading capital stays invested for a long term if he or she doesn’t sell— and is, thus, not available to open other positions.
What to do? The trader sets up a stop-loss order for the position. It will automatically sell the position if the security’s price falls below a pre-defined stop-loss price. This (for a long position) works as follows:
Say, the trader bought a stock at 100 $ via his or her broker.
- First, he or she defines a stop-loss price — e.g., 2 % below the entry price.
- This price is 100 $ — 2 %*100 $= 100 $ — 2 $ = 98 $.
- The trader enters a stop-loss order with this price (98 $) and some pre-defined duration at the broker. Let’s say that duration is unlimited.
- Now, if the stock price some time in the future drops below 98 $, the stop-loss order turns into a market order. And that detail is crucial! The security is then sold at the best available price in the next possible time frame. The trader is, thus, not guaranteed the stop-loss price.
- However, usually the price at which the stop-loss order is executed (i.e., at which the security is actually sold for) is reasonably close to the intended price. This is dependent on various factors like liquidity in the market but commonly holds true for widely traded securities (large stocks, large ETFs, etc.).
How stop-loss (approximately) behaves in practice
To demonstrate this practical stop-loss behaviour, here comes a simple analysis. It is based on a synthesized (randomly generated) open-high-low-close price time series which imitates the statistical behaviour of a major stock index closely.
The price data shown is synthetic (random generated) data mimicking the statistical behaviour of a large stock index — it is, thus, not a prediction. First close price is set to 100 for convenience. For more on the methodology, please see below under the article.
Rules for the analysis are:
- If the trader is not in the market, he will instantly buy one share.
2. After buying that share, a fixed-percentage stop-loss order is entered. The fixed percentage varies for our analysis / parameter study between 0.1 and 1.9 % distance to current open (buy) price in steps of 0.1 %.
3. If stop-loss is triggered by a daily low, a market sell order is (automatically) set up.
4. The share is then sold at the next possible (assumption: next-day) open price. In reality, it would sell off intraday, but this time series does not simulate intraday data.
5. So, the trader is not in the market anymore. He will then continue to instantly buy one share at market open price ( → first rule).
Obviously, that is a dumb trading system and also simplified in its execution (no fees considered, no intraday data). However, it allows to derive some useful knowledge about stop-loss behaviour. See examples of 0.5 % and 1 % stop loss in the time series below.
At the beginning (A), the trader enters a trade (rule #1, not shown explicitly), independent of both stop-loss settings.
The stop-loss orders are entered (rule #2), with 0.5 % distance (red) and 1.0 % (muddy gray) distance to current open price. Thus, the red curve is closer to the actual price data than the muddy gray one.
Right on the next day, the low price falls below 0.5 % (red arrow) and, thus, the stop-loss is triggered (rule #3-#4). Thus, for the 0.5 % stop-loss, the position is closed with next market open price, and the trader after that moves into the market again (rule #5 → #1).
For the 1 % stop-loss, nothing happens, as the low price has not yet fallen more than 1 % from the entry price level. On the third workday (two day gap is due to weekend) the low price actually drops below the stop-loss price. You know what happens next: market order, sell at next possible point in time (here: next open) at best possible price, close trade, open again.
This whole game repeats itself at B, where there is a lot of sideways movement and, thus, a lot of stop-loss triggering especially for the narrow 0.5 % setting.
→ Stop-loss should be set so that it considers current market conditions.
From C on, something special happens: the stop-loss price is never triggered again until the end of the time window (“August 23” in our synthesized data).
Let’s see how under these assumptions the stop-loss price was met on average.
For very narrow stop-losses (here: less than 0.7 % from entry price) the desired closing price is not guaranteed in this example. Actual closing prices are much less favorable than intended. Above that, the realized selling prices vary: We see that these might even be slightly better than the stop-loss our trader set, if the price has moved up again before the forced sell. This difference between intended and actual price is called slippage and is, obviously, dependent on the example / security chosen!
→ Traders should be aware of slippage in their systems and always calculate conservatively.
Implications of setting conservative (wide) and risky (narrow) stop-losses
Stop-losses always have a trade-off.
The trader might think: Couldn’t I just choose a very narrow stop-loss? If it is triggered, I won’t lose much; if not, I win a lot.
But this is a double-edged sword. As stop-losses get narrower, their being triggered increases vastly:
Yes, our trader (or trading dwarves, who knows, up to now they have only been investing) will not lose much each time, even though slippage occurs and probably worsens his or her results. But losing will get much more frequent, as the narrow stop-loss is triggered often.
And for the second part of his/her assumption — “if not, I win a lot”: this does not have to be the case, it just could. But anyways: Let’s appreciate the traders and everyone’s optimism and audacity!
So, what to do now?
- Stop-loss orders in general are a great way to mitigate losses when traders are wrong.
- Very narrow stop-loss distances increase the number of losing trades (not good).
- Very wide distances increase the loss per trade (also not good).
- Thus, traders should use something inbetween and may want to pick a certain price level from the past or “usual” variations in security price over a certain past time window.
To be more quantitative:
- Depending on the overall strategy, a suitable stop-loss value can be derived at least for historical data via backtesting, for example using backtrader.
- It usually makes sense to choose a stop-loss dependent on the range of current variation of a security, for example using standard deviation (like Bollinger Bands) or Average True Range.
Coding human’s acknowledgements: Studies were carried out within the PyCharm IDE (https://www.jetbrains.com/pycharm/), among other libraries. Thank you to the developers for providing such great user-friendly and reliable tools.
If you are interested in the code that was used for this, please give me a heads up and I will provide you with my GitHub repo.
DISCLAIMER: This article presents my own learnings based on studies generated on synthesized (random, i.e., artificial) data which is statistically similar to real-world time series, and personal experience. The content is, thus, purely educational. Past performance is not reliable indicator of future results. The article should not be considered Financial or Legal Advice. Consult a financial professional before making any major financial decisions.
