A bear flag is a steep price drop (the pole), then a short pause that drifts slightly upward or sideways (the flag), then a possible break lower. The common way to set a target is to subtract the pole's height from the breakdown price. That number comes from a charting convention, not a forecast. The largest public catalogue of flags, kept by researcher Thomas Bulkowski, reports that 46% of flags with downward breakouts reached their price target. His target rule is more conservative than the full-pole rule.
What the pattern looks like
A CoinDesk explainer republished on Yahoo Finance and dated July 21, 2018, splits a flag into two parts. "The 'pole' represents a strong impulsive move (higher/lower) and is backed by a surge in trading volume." The flag is the pause that follows. It "looks like a falling or rising channel." The bear version is "spotted in a downtrend when the price is likely to sink further."
A guide from chart-tools vendor LuxAlgo, dated July 17, 2025, defines it more narrowly. The flag "resembles a small, roughly parallel channel angled against the prior decline."
Bulkowski's identification notes on thepatternsite.com add limits on time and shape:
- Length: "Flags are short, less than 3 weeks long." Longer pauses count as rectangles or channels.
- Pole: "The flagpole which leads to the flag should be unusually steep and last several days."
- Boundaries: Price moves between two parallel or nearly parallel trendlines.
- Volume: Volume trends downward during the flag in 77% of his patterns with downward breakouts.
When a break counts, and what cancels it
LuxAlgo separates "a temporary move beneath the lower boundary" from "a completed candle close below it." Only the close counts as a breakdown under its rules. Requiring an extra follow-through candle is optional, and the guide says to decide that before testing.
The guide's section on failed flags names three conditions that cancel the pattern:
- Close above the top: "A close through the upper flag boundary can invalidate the bearish structure."
- Retracement too deep: The bounce retraces more of the pole than your preset limit. LuxAlgo's own indicator defaults to a maximum flag depth of 50% of pole height and says this is not a universal rule.
- Pause too long: The pause outlasts the allowed duration. The indicator's default is a 4–15-bar flag, also presented as a setting rather than a requirement.
What volume can and cannot show
Both guides expect heavy volume on the pole and quieter trading in the flag. The 2018 explainer warns that "a low volume move usually ends up trapping investors on the wrong side of the market." LuxAlgo is more cautious: "A volume surge cannot certify continuation." Its indicator's volume filter is off by default.
Bear flag target from three chart levels
Enter a pole high, a pole low and a breakdown price from your own chart to get the projected bear flag target. The preset levels are hypothetical, not market data.
| Pole high (top of the steep fall) | Pole low (bottom of the steep fall) | Breakdown price (close below the flag) | Projected target (arithmetic, not a forecast) |
|---|---|---|---|
| 100 | 80 | 85 | 65 |
| 50 | 40 | 42 | 32 |
| 20 | 16 | 17 | 13 |
- The preset and example levels are hypothetical, not market data.
- All three levels are your own chart readings in the same price unit.
- Pole height is the pole high minus the pole low, as the 2018 explainer on Yahoo Finance defines it.
- The target subtracts that height from the breakdown price, the rule both guides give.
- A result at or below zero is not a usable price level.
The result is arithmetic on the levels you enter, not a probability or a promise that price will reach it.
Working out the target
The full-pole rule
The 2018 explainer and LuxAlgo use the same subtraction:
- Pole height = pole high − pole low
- Target = breakdown price − pole height
Hypothetical example (not market data): pole high 100, pole low 80, breakdown 85. The pole height is 20, so the target is 85 − 20 = 65.
Bulkowski's adjusted rule
Bulkowski scales the move down by how often targets were actually reached. You measure the pole height, multiply it by the share of patterns that met the target (46% for downward breakouts), and subtract the result from the top of the flag.
Same hypothetical pole, with the flag topping at 88: 20 × 0.46 = 9.2, so the target is 88 − 9.2 = 78.8. The full-pole rule gives 65 for the same chart, so the two methods put the target far apart.
What the record shows
Bulkowski's flags page shows statistics last updated on August 27, 2020. They come from a single bull-market table, and the page does not say which stocks or years the sample covers.
| Measure (flags, bull market) | Downward breakouts |
|---|---|
| Break-even failure rate | 45% |
| Average decline | 8% |
| Share meeting price target | 46% |
He notes that flag results "are based on the short-term price swing." For that reason flags are not ranked against other patterns, and the average move looks small.
LuxAlgo, for its part, says "there is no substantiated universal 60–70% continuation rate or 8% median move" for the rules in its guide. The 2018 explainer's line that "an asset usually mimics the pole" after a bear flag breakdown comes with no supporting data. Bulkowski's figures suggest a full repeat of the pole is the less likely outcome.
Stops and short selling
LuxAlgo says the pattern "does not make short selling safe or guarantee another decline." It describes two stop placements and what each one tests:
- Above the whole flag: tests whether the entire structure has failed.
- Above a retest high: tests only a narrower entry idea.
It rejects "a universally correct 1–2% buffer." Instead, it ties the buffer to the instrument's tick size, spread and volatility. It also warns that "a gap through a stop can create a larger loss than the initial cash-risk calculation." The U.S. SEC's investor education site, Investor.gov, defines a short sale as selling stock you do not own.
Limitations: the only tested figures come from one researcher's stock-market catalogue, last updated in 2020, which does not cover crypto. Neither guide reports tested results of its own.
