To Catch a Falling Knife

So, What's the Catch?

Essentially, when a trader "catches a falling knife", he is making an educated trade and guessing where a retracement ends. A retracement is a technical term used to describe a minor pullback (i.e. decrease) in the direction of a financial instrument. Key to note is that a retracement should be temporary, unlike a reversal, and it should not indicate a shift in the prevailing uptrend. (Note: we can speak of positive retracements too (i.e. increases in price) but for this article, we focus on negative ones.)

Why the Interest?

Well, if you're a trader, you should always be thinking whether a dip in the price of an asset is a long-term downtrend or a temporary market blip. Selling stock in a retracement only to see it rocket upwards days later is painful, but holding onto stock in a reversal can be devastating for traders.

General Retracement Rules

So to be sure what a retracement is, let's discuss in context and take a look at a real-life price chart of the Singapore stock market. Nikko AM STI ETF tracks the Straits Time Index, indicative of the Singapore market performance. Below, I generated a custom plotly chart of its prices from 5 Oct 2011 to 12 Aug 2013:

• Generally, it's an uptrend period. The rough support region is shown as a dotted line.
• Of interest are the retracement zones, which have been highlighted in red.
• There are four retracements, first on the leftmost and the last on the rightmost.

Three general rules follow:

• A retracement should not exceed a significant previous low. Using the chart above as an example, the second retracement doesn't challenge the lowest point of the first retracement, and so on for the subsequent retracements. Another popular way of saying this is that "retracements in an uptrend are characterized by higher lows and higher highs".
• Retracements have strong knacks for ending near round/nice numbers. Do take note that this is more of a heuristic than a strong rule. Again, looking at the chart, the retracement ended near 2.65, 2.75, 3.0 and 3.1 for the four zones respectively.
• Retracements are temporary. This is actually quite subjective. In a volatile market like the US stock market, temporary might mean 2-3 weeks. But in the relatively inert Singapore stock market where trading volume is lower, this might mean 2 months. If you look at the retracements above, generally, it is about 2 months. I recommend understanding the trading behaviour of your market well, before coming to any conclusions about any suitable timeframe of a retracement.

Other than these simple chart patterns, we should look at other indicators that point toward a retracement as well:

• Fundamentals Remain Unchanged: Watch the news, understand the company, and make sure that nothing fundamental about the company's profit and business has changed.
• Low Volume Trades: generally the retracement occurs due to profit-sell off by retail traders. So watch out for small block trades in contrast to large blocks of institutional selling.
• Buying Interest & Cashflow: watch the trade discussions, generally you'll see buying interest still, since people believe they're getting a temporary bargain.
• Indecision Candlesticks: Since nothing has fundamentally changed, expect buyers to be in a standoff with sellers - so watch out for dojis and spinning tops.
• Recent Rally: Was there a recent rally? If so, a retracement is almost certain - profit-taking is human, holding assets for the long-term is not.

That's about it for now, we can discuss more about what constitutes a trend reversal in future, and more complicated ways to calculate the scope of a retracement, such as Fibonacci Retracements. Python code for the chart above in full, appended below for reference:

## Loading Libraries
import plotly.express as px
import plotly.graph_objects as go
from plotly.subplots import make_subplots
import pandas as pd
import yfinance as yf

## Accessing stock price data from yfinance
sti_df=yf.download(
    tickers="G3B.SI", 
    period="max",     
    interval="1d",    
    progress=False,   
    auto_adjust=True  
)

## Change the "date" index into a "date" column
sti_df = sti_df.reset_index()
sti_df["Date"] = pd.to_datetime(sti_df["Date"])

## Selecting a certain period of the STI history
start_date = '2011-10-05'
end_date = '2013-08-12'
mask = (sti_df["Date"] >= start_date) & (sti_df["Date"] <= end_date)
sti_segment = sti_df.loc[mask]

sti_segment["Type"] = "General Uptrend"

## Selecting retracement periods
retrace_start1 = '2011-10-28'
retrace_start2 = '2012-04-02'
retrace_start3 = '2012-10-05'
retrace_start4 = '2013-05-22'
retrace_end1 = '2011-12-20'
retrace_end2 = '2012-06-05'
retrace_end3 = '2012-11-16'
retrace_end4 = '2013-06-25'
retrace_mask1 = (sti_segment["Date"] >= retrace_start1) & (sti_segment["Date"] <= retrace_end1)
retrace_mask2 = (sti_segment["Date"] >= retrace_start2) & (sti_segment["Date"] <= retrace_end2)
retrace_mask3 = (sti_segment["Date"] >= retrace_start3) & (sti_segment["Date"] <= retrace_end3)
retrace_mask4 = (sti_segment["Date"] >= retrace_start4) & (sti_segment["Date"] <= retrace_end4)

sti_segment["Type"][retrace_mask1] = "Retracement"
sti_segment["Type"][retrace_mask2] = "Retracement"
sti_segment["Type"][retrace_mask3] = "Retracement"
sti_segment["Type"][retrace_mask4] = "Retracement"

## Creating retracement and uptrend dataframes
sti_retracement = sti_segment.copy()
sti_retracement["Close"][sti_retracement["Type"] == "General Uptrend"] = np.nan
sti_retracement["Type"] = "Retracement"

sti_upwards = sti_segment.copy()
sti_upwards["Close"][sti_upwards["Type"] == "Retracement"] = np.nan
sti_upwards["Type"] = "General Uptrend"

## Creating final dataframe of relevant STI data
sti_merged = pd.concat([sti_retracement, sti_upwards])

## Generating an empty Plotly chart object
fig = go.Figure()

## Creating the Plotly chart object for STI prices
fig = px.line(        
      sti_merged,          
      x="Date",       
      y="Close",          
      title="Historical Prices for Nikko AM Singapore STI ETF",
      color="Type"
)

fig.update_traces(connectgaps=False)

fig.update_traces(
    line_width=2,
)

fig.update_layout(
    xaxis_title="Date", 
    xaxis_gridcolor="grey",
    yaxis_title="Closing Price",
    yaxis_gridcolor="grey",
    
    font_color="black",
    
    title_font_size=20,
    title_x=0.5,
    paper_bgcolor='white',
    plot_bgcolor='white'
)

## Create a support line
fig.add_trace(go.
              Scatter(
                x=["2011-10-05", "2013-08-12"], 
                y=[2.59, 3.13], 
                name='Support Line',
                line=dict(dash='dash')
              )
)

## Final formatting details
fig.data[0].line.color = "lightcoral"
fig.data[0].line.width = 4
fig.data[1].line.color = "black"
fig.data[2].line.color = "black"

Find more Finalytics stories on my blog. Have a suggestion? Contact me at [email protected].