Skip to content
Menu
Finabase Learn

Strategies · Trend and momentum

Buying pullbacks in an uptrend: a dip-buying rule, tested

A rule for buying pullbacks in an uptrend: buy a five-day low above the 100-day average, sell on the bounce. Tested with costs on SPY and 30 stocks.

On this page
  1. The idea
  2. The rules
  3. What the backtest shows
  4. Where it fails: the losing periods
  5. Compared with its neighbours
  6. How Finabase does it
  7. Frequently asked questions
  8. Summary
  9. Check your understanding

"Buy the dip" is the most repeated advice in a rising market. The version traders take seriously adds a condition: buy the dip only when the larger trend is up. The trend supplies the direction, and the pullback supplies a better price than chasing a new high. It is a combination of the two families in this section, a trend filter with a short-term entry against the latest move.

This lesson turns that idea into a rule with no judgment in it and tests it on two years of daily prices. Two years is too short to say whether a strategy works, and the page does not claim to. What the test shows clearly is the rule's shape: it won four trades in five, each win was small, costs took a third of the profit, and it was in the market only a quarter of the time while the market rose. On 30 large stocks the average result was zero.

The idea

An uptrend does not rise in a straight line. It advances, gives some back, and advances again. The Trends and trendlines: how to identify a trend and draw it correctly lesson describes this as a series of higher highs and higher lows. A pullback is one of the "gives some back" stretches.

The strategy rests on two claims.

  1. In an uptrend, short falls tend to reverse. Buyers who missed the advance use the dip, so the fall is shallow and brief.
  2. Outside an uptrend, they may not. A dip in a falling market is often the next leg down. The trend filter is there to skip those.

The first claim is short-term mean reversion: after a few down days, the next few days have been slightly better than average in US stock indexes for most of the last three decades. The second claim is trend following. The rule needs both to be true to beat simply holding, because it gives up all the days when it is not in the market.

That last point is the one dip buyers tend to miss. A rule that is right four times in five can still fall far behind an investor who did nothing, if the market keeps rising while the rule waits for the next dip.

The rules

On daily bars, checked after each close:

  1. Compute the 100-day and the 10-day simple moving average of the close.
  2. Buy at today's close if you hold cash, today's close is above the 100-day average, and today's close is the lowest of the last five closes.
  3. Sell at today's close if you hold the fund and today's close is above the 10-day average. That is the bounce.
  4. Also sell if today's close is below the 100-day average. That is the trend failing, and it is the rule's only stop.

The usual trend filter in published versions is the 200-day average. The daily price history available for these lessons covers two years, and a 200-day average would use up most of the first year before the test could start, so this rule uses 100 days. Results with a 50-day filter are in the settings table.

SPY daily, 2025-04 to 2026-11: every buy (▲) and sell (▼) of the pullback rule $500 $600 $700 SPY close 100-day average May Jul Sep Nov Jan 2026 Mar May Jul Sep Nov
SPY daily, April 2025 to November 2026, with its 100-day average. Up triangles are buys at a five-day low and down triangles are sells.

A worked example

On October 10, 2025 SPY closed at $653.02, its lowest close in five days. The 100-day average was $630.99, so the trend filter passed and the rule bought.

Six trading days later, on October 20, SPY closed at $671.30, above its 10-day average of $665.32. The rule sold for a gain of 2.6% after costs. That was the best trade of the test.

The worst came a month later. On November 14 the rule bought at $671.93 with the 100-day average at $650.81. The dip continued. On November 20 SPY closed at $652.53, just below the 100-day average of $652.60, and the trend exit sold for a loss of 3.1%. One loss of that size cancelled about four average wins.

What the backtest shows

The test, stated in full:

  • Data. SPY daily closes from November 18, 2024 to November 17, 2026. The first 100 days build the trend average, so the test runs from April 14, 2025: 402 trading days, about 1.6 years.
  • Trading. The signal and the trade both use the day's close. In practice that means an order placed in the last minutes of the session.
  • Costs. 0.10% of the amount on every buy and every sell.
  • Left out. Dividends, interest on cash and taxes.
  • What the period is. It begins a few days after the low of the April 2025 fall and covers a strong rise. There is no bear market in it. It shows the rule's behaviour in an uptrend and cannot show whether the rule has an edge.
SPY daily, 2025-04 to 2026-11: pullback rule against buy and hold, 0.10% per trade, no dividends $1 $1.2 $1.4 Growth of $1 Buy and hold Pullback rule 0% −10% Fall from the peak May Jul Sep Nov Jan 2026 Mar May Jul Sep Nov
SPY daily, April 2025 to November 2026: the pullback rule against buy and hold, 0.10% per trade, dividends not included. Top: growth of $1. Bottom: the fall from the highest value so far.
Measure Pullback rule Buy and hold
Total return 15.3% 35.9%
Return per year (CAGR) 9.4% 21.2%
Deepest fall from a peak −4.2% −9.1%
Longest time below a peak 127 days 53 days
Time invested 25% 100%
Round-trip trades 30 1
Trades with a gain 80% n/a

Three things stand out.

The win rate is high and the wins are small. Twenty-four gains averaged 0.8% and six losses averaged 0.9%. Most trades lasted one to six days.

Costs are a large share. A round trip costs 0.20%, a quarter of the average win. With no costs the rule gains 22.5% and wins 90% of its trades; three trades that showed a small profit before costs became losses after them. A rule with many small wins is far more sensitive to the cost assumption than a rule that trades once a year.

For the time it was exposed, it did well. It earned 15.3% while invested a quarter of the time, and its worst fall was less than half of buy and hold's. Per day in the market, that is a better result than holding. Per calendar year it is much worse, and money can only be spent per calendar year.

The two halves of the period

Period Pullback rule Buy and hold
April 2025 – February 2026 7.0% 29.0%
February 2026 – November 2026 7.8% 5.3%

The table gives total returns, not yearly rates, because each half is under a year. In the first half the market rose almost without pause and the rule captured a quarter of the gain. In the second half the market went sideways and then recovered, with a fall of 9.1% from late January to the end of March 2026, and the rule came out ahead. It made no trades from the end of February to mid-May 2026, because SPY was below its 100-day average for much of that time. That is the trend filter working as designed.

Where it fails: the losing periods

A strong, steady rise. From April 2025 to February 2026 buy and hold gained 29.0% and the rule 7.0%. An uptrend with few dips gives the rule few entries, and each entry is sold after a bounce of a percent or so.

The dip that keeps going. The exit on a break of the 100-day average is far away when the price is well above the average, so a failed trade loses more than a successful one wins: −3.1% in November 2025 and −1.9% in February 2026 against a typical gain under 1%.

Individual stocks. The same rule on each of the 30 stocks in the Dow Jones Industrial Average as of late 2024, over the same days:

Measure Pullback rule Buy and hold
Average total return 0.0% 29.7%
Average deepest fall −14.3% −26.8%
Stocks where the rule did better 4 on return 24 on the deepest fall
Trades, and share with a gain 614, 63% n/a

Across 614 trades the average gain was 1.7%, the average loss 2.9%, and the average trade −0.03% after costs. Results ranged from −38.9% on IBM to +65.5% on Goldman Sachs. A single stock can gap down on an earnings report while the rule is holding it, and an index fund cannot. These are today's index members, so both columns benefit from hindsight; the comparison between them is what matters.

The trend filter did not help here. Buying every five-day low with no trend test, and selling above the 10-day average, gained 24.4% on SPY against 15.3% with the filter, and averaged 8.9% on the 30 stocks against 0.0%. In a period with no bear market, a filter can only remove trades. Its purpose is to avoid buying dips in a long decline, and this data contains none, so the test cannot show the filter's value. It can only show its cost.

Other limits:

  • Taxes and account rules. Every gain is short-term and taxed as ordinary income in a US taxable account (IRS Topic 409). Trades opened and closed on the same day do not occur in this rule, but a trader running similar rules intraday in a margin account should know FINRA's intraday margin rules. They replaced the pattern day trader rule on June 4, 2026, and a broker may still apply the old rule until October 20, 2027.
  • Filling at the close. The test assumes a fill at the exact closing price. A real order near the close gets a price close to it, not equal to it.

Neighbouring settings

On SPY over the same days. Buy and hold gained 35.9%.

Trend average Dip: lowest close of Exit above Total return Trades Trades with a gain
100 days 3 days 10-day 6.2% 56 61%
100 days 5 days 10-day 15.3% 30 80%
100 days 7 days 10-day 15.2% 23 83%
100 days 5 days 5-day 7.4% 31 74%
100 days 5 days 20-day 19.4% 37 78%
50 days 5 days 10-day 4.3% 25 60%

Every version trailed holding. The spread, from 4.3% to 19.4%, is wide for settings this close together, and with about 30 trades each the differences are well within what chance produces. On QQQ the main rule gained 1.7% against 43.6% for holding; on IWM, 16.1% against 44.2%.

Compared with its neighbours

Rule Bars Typical holding time Share of time invested Trades a year
Pullback in an uptrend Daily Days 25% About 19
Moving average trend following Monthly Months to years 76% Under one
Buy and hold None Always 100% None

The pullback rule is the opposite of the monthly trend rule in almost every respect: many trades, small gains, high win rate, large cost share. The monthly rule's weakness is the sharp fall that reverses. This rule's weakness is the steady rise it is not on board for.

How Finabase does it

No Finabase tool tells you when to buy a dip. These show the rule's ingredients.

  • Chart. The link above opens SPY with the 10-day and 100-day simple moving averages. A five-day low above the longer line and a close back above the shorter line are both visible by eye.
  • Backtest. The Backtest tool trades the signals of the indicators and candlestick patterns it offers on one symbol. It does not have this exact rule. Its results also differ in kind from this page's: each signal buys a fixed $10,000, and no trading cost is charged, which flatters any rule with many small wins.
  • Strategy. The Strategy tool replays if/then allocation rules over two years of daily prices, the same span as this test, and with the same limit: it shows behaviour in one kind of market.

Frequently asked questions

Does buying the dip work?

In this test, buying five-day lows in an uptrend won 80% of its trades on SPY and still earned less than half of what holding earned, because it was in the market only a quarter of the time. On 30 large stocks it averaged no gain at all. Two years of data cannot settle the question either way.

Why use a trend filter if it lowered the return?

Its job is to avoid buying dips during a long decline, when a "dip" is the start of the next leg down. The period tested has no long decline, so the filter could only cost trades. A fair test of the filter needs data that includes a bear market, which the daily history here does not.

How is this different from RSI mean reversion?

The entry idea is similar: buy short-term weakness and sell the rebound. A pullback rule adds the requirement that the larger trend is up and usually has a trend-based stop. Rules based on the RSI measure the weakness with an oscillator in place of a count of days.

Why are the wins so small?

The rule sells at the first close above the 10-day average, which usually comes within a few days of the entry. Holding longer, to the 20-day average, raised the total return in this test from 15.3% to 19.4%, but that is one sample, and it also means sitting through more of each failed trade.

Summary

Buying pullbacks in an uptrend means buying a short fall while the longer trend is up and selling the rebound. A rule built on a five-day low above the 100-day average won 24 of 30 trades on SPY between April 2025 and November 2026 and gained 15.3% after costs, against 35.9% for holding. It was invested a quarter of the time, costs took a third of its profit, and its losses were larger than its wins. On 30 Dow stocks it averaged zero. The test covers under two years of a rising market, so it describes how such a rule behaves, with many small wins and a large cost of waiting, and leaves open whether it has an edge.

Check your understanding

4 questions on this lesson. Each answer comes with the reason, and nothing is scored.

  1. A fund closes at $98, its lowest close in five days. Its 100-day average is $101. What does the pullback rule do?

    1. Buys, because the price has dipped
    2. Nothing, because the close is below the 100-day average
    3. Sells short
    Show the answer

    B. Nothing, because the close is below the 100-day average The trend filter comes first. A five-day low below the 100-day average is a dip without an uptrend, which is the case the rule is designed to skip.

  2. The rule won 80% of its trades on SPY and gained 15.3%, while buy and hold gained 35.9%. How can both be true?

    1. The wins were small and the rule was invested only a quarter of the time
    2. Buy and hold used leverage
    3. The win rate was measured wrongly
    Show the answer

    A. The wins were small and the rule was invested only a quarter of the time Win rate says how often, not how much. Wins averaged 0.8%, and the market’s rise mostly happened on days when the rule was waiting in cash.

  3. With no trading costs the rule gains 22.5%; with 0.10% per trade it gains 15.3%. Why does this rule lose so much to costs?

    1. It trades 30 round trips for average wins under 1%, so a 0.20% round trip is a large slice of each win
    2. Costs are higher for daily bars
    3. It trades at the open
    Show the answer

    A. It trades 30 round trips for average wins under 1%, so a 0.20% round trip is a large slice of each win Cost per trade is fixed while the profit per trade is small. The same 0.10% is trivial for a rule that trades once in two years and decisive for one that trades 19 times a year.

  4. Removing the trend filter raised the return in this test. What is the sound conclusion?

    1. The trend filter is useless
    2. The test period has no long decline, so it can show the filter’s cost and not its benefit
    3. The filter should use 200 days
    Show the answer

    B. The test period has no long decline, so it can show the filter’s cost and not its benefit A filter that keeps the rule out of falling markets can only be judged on data that contains one. From April 2025 to November 2026 there was none, so every skipped trade was a missed gain.

Sources and further reading

  1. Larry Connors and Cesar Alvarez, Short Term Trading Strategies That Work (TradingMarkets, 2008) — buying short-term weakness above the 200-day average
  2. IRS — Topic no. 409, Capital gains and losses
  3. FINRA — Understanding the New Intraday Margin Requirements (the 2026 replacement for the pattern day trader rule)
  4. Finabase chart data — daily bars for SPY, QQQ, IWM and 30 Dow stocks, November 2024 to November 2026; rules and results computed by Finabase