Claim check: Buy when there is blood in the streets
“Buy when there is blood in the streets: a 20% crash is the time to buy.”
What the data says
Not enough data Not enough data to say: no market had more than 9 independent events after the signal, fewer than the 10 needed even to start judging, so any pattern is an anecdote.
Too few independent events or observations to judge either way.
Who says it
Attributed to Baron Rothschild; repeated by Warren Buffett fans and Reddit.
Origins are as commonly cited and were not independently verified.
How we tested it
Drawdown is the close divided by the highest close so far, minus one. Event: the first day it reaches -20% after having been above -10%. We record the 250-trading-day return from that close and compare it with the average 250-day return of the same market. Such events are rare, so a milder -10% version is shown for contrast.
Results
Drawdown reaches -20% (after being above -10%) 250-day (claim: higher than average)
| Market | Events | Mean after | All days | Difference | Median after | Up share | t-stat | Reading |
|---|---|---|---|---|---|---|---|---|
| S&P 500 | 2 | +38.8% | +14.9% | +23.9% | +38.8% | 100% | 1.1 (likely noise) | Too little data |
| Gold | 2 | +1.0% | +9.6% | -8.5% | +1.0% | 50% | -0.6 (likely noise) | Too little data |
| Silver | 2 | +40.1% | +13.2% | +26.9% | +40.1% | 50% | 0.5 (likely noise) | Too little data |
| Oil | 2 | +1.3% | +2.5% | -1.2% | +1.3% | 50% | -0.1 (likely noise) | Too little data |
| Natural gas | 1 | -39.3% | -16.3% | -23.0% | -39.3% | 0% | — (likely noise) | Too little data |
| Commodities | 2 | -1.1% | +4.5% | -5.6% | -1.1% | 50% | -1.4 (likely noise) | Too little data |
| Agriculture | 1 | +1.9% | +2.2% | -0.3% | +1.9% | 100% | — (likely noise) | Too little data |
| EUR/USD | 1 | -7.8% | -0.5% | -7.3% | -7.8% | 0% | — (likely noise) | Too little data |
| Bitcoin | 9 | +57.6% | +79.9% | -22.3% | +33.1% | 67% | -0.6 (likely noise) | Too little data |
| Ether | 4 | +6.3% | +58.3% | -52.0% | -56.4% | 25% | -0.7 (likely noise) | Too little data |
Difference from the average, by market (250-day)
By horizon: mean difference from all days
| Market | 60-day | 120-day | 250-day |
|---|---|---|---|
| S&P 500 | +10.4% (3) | +19.6% (2) | +23.9% (2) |
| Gold | -4.9% (3) | -5.3% (2) | -8.5% (2) |
| Silver | -3.0% (4) | -11.1% (3) | +26.9% (2) |
| Oil | -3.8% (3) | +1.2% (3) | -1.2% (2) |
| Natural gas | -13.3% (1) | -9.5% (1) | -23.0% (1) |
| Commodities | +8.3% (4) | +2.4% (3) | -5.6% (2) |
| Agriculture | +0.5% (1) | -9.2% (1) | -0.3% (1) |
| EUR/USD | -8.8% (1) | -5.0% (1) | -7.3% (1) |
| Bitcoin | +15.8% (15) | +1.8% (11) | -22.3% (9) |
| Ether | -13.3% (6) | +17.0% (5) | -52.0% (4) |
Drawdown reaches -10% (after being above -5%) · For comparison 250-day (claim: higher than average)
| Market | Events | Mean after | All days | Difference | Median after | Up share | t-stat | Reading |
|---|---|---|---|---|---|---|---|---|
| S&P 500 | 8 | +19.9% | +14.9% | +5.0% | +20.0% | 88% | 1.0 (likely noise) | Too little data |
| Gold | 5 | +10.8% | +9.6% | +1.2% | +6.0% | 80% | 0.2 (likely noise) | Too little data |
| Silver | 3 | +18.8% | +13.2% | +5.7% | +19.0% | 67% | 0.2 (likely noise) | Too little data |
| Oil | 2 | +3.8% | +2.5% | +1.3% | +3.8% | 100% | 0.6 (likely noise) | Too little data |
| Natural gas | 1 | -34.2% | -16.3% | -18.0% | -34.2% | 0% | — (likely noise) | Too little data |
| Commodities | 3 | +1.0% | +4.5% | -3.4% | -2.3% | 33% | -0.3 (likely noise) | Too little data |
| Agriculture | 2 | +12.1% | +2.2% | +9.9% | +12.1% | 50% | 0.3 (likely noise) | Too little data |
| EUR/USD | 2 | +4.5% | -0.5% | +5.0% | +4.5% | 50% | 1.0 (likely noise) | Too little data |
| Bitcoin | 9 | +53.4% | +79.9% | -26.5% | +37.2% | 67% | -0.8 (likely noise) | Too little data |
| Ether | 4 | +9.3% | +58.3% | -49.0% | -25.6% | 25% | -1.0 (likely noise) | Too little data |
0 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 | +23.9% (2) | — | +22.5% (2) |
| Gold | -8.5% (2) | -13.2% (1) | -5.0% (1) |
| Silver | +26.9% (2) | +39.1% (2) | — |
| Oil | -1.2% (2) | +8.8% (2) | — |
| Natural gas | -23.0% (1) | -20.7% (1) | — |
| Commodities | -5.6% (2) | +5.3% (1) | -17.4% (1) |
| Agriculture | -0.3% (1) | +5.9% (1) | — |
| EUR/USD | -7.3% (1) | -6.6% (1) | — |
| Bitcoin | -22.3% (9) | -11.4% (4) | -14.5% (5) |
| Ether | -52.0% (4) | -68.0% (3) | -70.1% (1) |
Context and cross-checks
Shown for context; not part of the verdict.
The site's own strategies
| Strategy | Market | Sharpe, 5 years | Benchmark, 5 years | Sharpe, full | Benchmark, full | CAGR, full | Benchmark, full | Max DD, full | Benchmark, full | Reading |
|---|---|---|---|---|---|---|---|---|---|---|
| RISK-01 Drawdown Control | SPY | 0.63 | 0.72 | 0.80 | 0.85 | +11.3% | +13.8% | -28% | -34% | No effect |
| RISK-01 Drawdown Control | GLD | 1.00 | 0.89 | 0.51 | 0.51 | +6.7% | +7.5% | -48% | -46% | No effect |
| RISK-01 Drawdown Control | USO | 0.27 | 0.68 | 0.02 | 0.05 | -2.0% | -4.8% | -66% | -95% | No effect |
| RISK-01 Drawdown Control | EUR/USD | -0.09 | -0.05 | -0.16 | -0.13 | -1.6% | -1.5% | -34% | -35% | No effect |
| RISK-01 Drawdown Control | DBC | 0.55 | 0.63 | 0.27 | 0.21 | +2.9% | +2.2% | -46% | -66% | Full history only |
| RISK-01 Drawdown Control | SLV | 0.71 | 0.57 | 0.37 | 0.35 | +6.0% | +6.1% | -67% | -76% | No effect |
| RISK-01 Drawdown Control | DBA | 0.67 | 0.68 | 0.11 | 0.13 | +0.6% | +0.8% | -59% | -62% | No effect |
| RISK-01 Drawdown Control | UNG | -0.32 | -0.27 | -0.18 | -0.35 | -6.0% | -25.1% | -66% | -99% | Full history only |
| RISK-01 Drawdown Control | BTC | 0.47 | 0.25 | 0.98 | 0.80 | +49.1% | +52.0% | -45% | -83% | Clear and stable |
| RISK-01 Drawdown Control | ETH | -0.04 | 0.09 | 0.73 | 0.56 | +34.2% | +22.9% | -51% | -94% | Full history only |
Sharpe ratio minus benchmark, full history · Benchmark = the buy-and-hold shown on the strategy pages.
The honest caveat
A 20 percent drop can continue to 50 percent (2008, 2022 crypto); the average hides that.
Descriptive statistics from historical data, before costs. Not a forecast and not investment advice.
Questions
Buy when there is blood in the streets: a 20% crash is the time to buy.
Not enough data to say: no market had more than 9 independent events after the signal, fewer than the 10 needed even to start judging, so any pattern is an anecdote.
How was this claim tested?
Drawdown is the close divided by the highest close so far, minus one. Event: the first day it reaches -20% after having been above -10%. We record the 250-trading-day return from that close and compare it with the average 250-day return of the same market. Such events are rare, so a milder -10% version is shown for contrast.
Related strategies on Quant Arena
Markets used
S&P 500 · Gold · Silver · Oil · Natural gas · Commodities · Agriculture · EUR/USD · Bitcoin · Ether