Claim check: Do not catch a falling knife
“Don't catch a falling knife: never buy something in a downtrend.”
What the data says
Partly Partly: after “20-day fall of 10% or more, below the 200-day average”, the average 21-day return was -0.7% against +0.7% for all days (medians over 6 markets); it was lower than usual in 5 of 6 markets and statistically clear, with at least 20 independent events and the same sign in both halves, in 0. The direction mostly points the claimed way, but not strongly enough to be sure.
It holds in some markets or periods, leans the right way without proof, or helps on risk but not on return.
Who says it
Trader's proverb contradicting 'buy the dip'.
Origins are as commonly cited and were not independently verified.
How we tested it
Event: the close is down 10% or more over 20 trading days and is below its 200-day average. We record the 21-trading-day return from that close (events at least 21 days apart). The claim says buying here is a bad idea, so it holds only if the return is clearly lower than the market's average 21-day return. The same fall above the 200-day average is shown for contrast.
Results
20-day fall of 10% or more, below the 200-day average 21-day (claim: lower than average)
| Market | Events | Mean after | All days | Difference | Median after | Up share | t-stat | Reading |
|---|---|---|---|---|---|---|---|---|
| S&P 500 | 7 | +3.1% | +1.2% | +1.9% | +3.3% | 71% | 0.6 (likely noise) | Too little data |
| Gold | 5 | +3.2% | +0.7% | +2.5% | +5.0% | 80% | 1.2 (likely noise) | Too little data |
| Silver | 32 | +1.9% | +0.9% | +0.9% | +2.2% | 66% | 0.8 (likely noise) | No effect |
| Oil | 42 | -0.4% | +0.3% | -0.7% | +2.9% | 55% | -0.2 (likely noise) | Right way, likely noise |
| Natural gas | 78 | -1.6% | -1.5% | 0.0% | -0.6% | 47% | 0.0 (likely noise) | Right way, likely noise |
| Commodities | 10 | -1.6% | +0.4% | -2.0% | -2.7% | 40% | -1.0 (likely noise) | Right way, likely noise |
| Agriculture | 3 | -0.9% | +0.2% | -1.0% | +0.9% | 67% | -0.6 (likely noise) | Too little data |
| EUR/USD | 0 | — | — | — | — | — | — | Too little data |
| Bitcoin | 45 | +1.4% | +4.0% | -2.6% | -0.2% | 49% | -1.0 (likely noise) | Right way, likely noise |
| Ether | 46 | -1.0% | +3.6% | -4.6% | -1.2% | 48% | -1.5 (likely noise) | Right way, likely noise |
Difference from the average, by market (21-day)
By horizon: mean difference from all days
| Market | 5-day | 21-day | 63-day |
|---|---|---|---|
| S&P 500 | +0.6% (17) | +1.9% (7) | +7.0% (7) |
| Gold | +2.3% (10) | +2.5% (5) | -0.6% (4) |
| Silver | +0.6% (70) | +0.9% (32) | -0.9% (21) |
| Oil | -1.1% (109) | -0.7% (42) | +0.5% (24) |
| Natural gas | 0.0% (217) | 0.0% (78) | +1.5% (41) |
| Commodities | -0.7% (20) | -2.0% (10) | -1.7% (8) |
| Agriculture | +0.3% (7) | -1.0% (3) | -3.4% (3) |
| EUR/USD | — | — | — |
| Bitcoin | -0.4% (128) | -2.6% (45) | -14.1% (24) |
| Ether | -1.2% (144) | -4.6% (46) | -1.5% (24) |
Average -day return after the signal vs all days (Silver)
Same fall, above the 200-day average · For comparison 21-day (claim: lower than average)
| Market | Events | Mean after | All days | Difference | Median after | Up share | t-stat | Reading |
|---|---|---|---|---|---|---|---|---|
| S&P 500 | 0 | — | — | — | — | — | — | Too little data |
| Gold | 2 | +4.5% | +0.7% | +3.8% | +4.5% | 100% | 6.5 (likely noise) | Too little data |
| Silver | 15 | +6.0% | +0.9% | +5.1% | +5.5% | 87% | 2.2 (likely noise) | No effect |
| Oil | 10 | +1.1% | +0.3% | +0.8% | +4.0% | 60% | 0.2 (likely noise) | No effect |
| Natural gas | 17 | -2.0% | -1.5% | -0.5% | -4.2% | 24% | -0.1 (likely noise) | Right way, likely noise |
| Commodities | 3 | +4.2% | +0.4% | +3.8% | +3.9% | 100% | 2.0 (likely noise) | Too little data |
| Agriculture | 1 | +7.7% | +0.2% | +7.5% | +7.7% | 100% | — (likely noise) | Too little data |
| EUR/USD | 0 | — | — | — | — | — | — | Too little data |
| Bitcoin | 32 | +5.1% | +4.0% | +1.0% | +4.0% | 59% | 0.3 (likely noise) | No effect |
| Ether | 24 | +0.4% | +3.6% | -3.2% | -3.0% | 38% | -0.7 (likely noise) | Right way, likely noise |
6 of 10 markets had enough data.
Stability across the two halves of history
A saying that only works in one half of the history is not much of a rule. Each row shows the effect in the full history, the first half and the second half; green means the claimed effect was present.
| Market | Full history | First half | Second half |
|---|---|---|---|
| S&P 500 | +1.9% (7) | -0.4% (2) | +2.7% (5) |
| Gold | +2.5% (5) | +4.7% (4) | -4.8% (1) |
| Silver | +0.9% (32) | +1.4% (19) | +0.6% (13) |
| Oil | -0.7% (42) | +0.6% (23) | -2.0% (19) |
| Natural gas | 0.0% (78) | +0.5% (33) | -0.6% (45) |
| Commodities | -2.0% (10) | -1.7% (6) | -2.1% (4) |
| Agriculture | -1.0% (3) | +1.2% (1) | -2.4% (2) |
| Bitcoin | -2.6% (45) | -5.3% (19) | -0.3% (26) |
| Ether | -4.6% (46) | -5.2% (24) | -4.1% (22) |
The honest caveat
It overlaps with 'buy the dip'; real traders also use position size.
Descriptive statistics from historical data, before costs. Not a forecast and not investment advice.
Questions
Don't catch a falling knife: never buy something in a downtrend.
Partly: after “20-day fall of 10% or more, below the 200-day average”, the average 21-day return was -0.7% against +0.7% for all days (medians over 6 markets); it was lower than usual in 5 of 6 markets and statistically clear, with at least 20 independent events and the same sign in both halves, in 0. The direction mostly points the claimed way, but not strongly enough to be sure.
How was this claim tested?
Event: the close is down 10% or more over 20 trading days and is below its 200-day average. We record the 21-trading-day return from that close (events at least 21 days apart). The claim says buying here is a bad idea, so it holds only if the return is clearly lower than the market's average 21-day return. The same fall above the 200-day average is shown for contrast.
Markets used
S&P 500 · Gold · Silver · Oil · Natural gas · Commodities · Agriculture · EUR/USD · Bitcoin · Ether