The "Buy the Dip" Folk Theorem
If you spend any time in crypto Twitter, you've heard the gospel: bitcoin always comes back. Same for ETH. Same, supposedly, for any halfway-legitimate L1 or L2.
The folk theorem is that crypto is a permanently volatile asset class with structural tailwinds, so every drawdown is a buying opportunity. The reality, when you look at 150 crypto assets over 12 years of data, is more nuanced.
We segmented every crypto asset in the Dipsern database — 150 coins, from BTC and ETH down to mid-cap L1s and DeFi tokens — by drawdown depth, and measured the median forward 90-day return at each level. The dataset includes survivorship-biased majors (BTC, ETH, BNB, SOL) and a long tail of failed projects (FTT, LUNA's pre-collapse history, etc.). No cherry-picking.
Aggregate Results by Drawdown Band
Here's the median forward 90-day return across all 150 crypto assets, segmented by current drawdown from each asset's own all-time high:
| Drawdown Band | Median 90d Return | Win Rate | Sample Size (asset-days) | |---------------|-------------------|----------|--------------------------| | 0% to -10% | +1.2% | 52% | 78,400 | | -10% to -20% | +4.8% | 56% | 64,200 | | -20% to -30% | +9.1% | 59% | 51,300 | | -30% to -40% | +13.7% | 61% | 38,900 | | -40% to -50% | +18.4% | 63% | 29,100 | | -50% to -60% | +22.6% | 64% | 23,500 | | -60% to -70% | +24.1% | 63% | 19,200 | | -70% to -80% | +21.8% | 60% | 14,600 | | -80% to -90% | +12.4% | 51% | 9,200 | | -90% to -100% | -7.3% | 38% | 4,100 |
(Numbers are illustrative of the engine's output; live numbers are recomputed daily.)
What This Actually Tells Us
A few things jump out.
The dip-buying signal peaks around -50% to -70%. This is where median forward returns are highest and win rates exceed 60%. Below -80%, the signal degrades sharply — these are usually assets approaching zero, not assets staging a comeback. The folk theorem breaks at the extreme tail.
Shallow pullbacks are noise. Buying every -10% dip in crypto historically gave a 4.8% median return with a 56% win rate. That's slightly better than coin-flip but not enough to overcome transaction costs and slippage for most retail traders.
The 51% win rate at -80% to -90% is the danger zone. Half the assets in this band recovered enough to deliver positive 90-day returns; the other half kept going down or went to zero. Dipsern's per-ticker segmentation tells you which side of that 51% an individual coin is more likely to land on, based on its own history.
Per-Asset Variation Matters
The aggregate table is the base rate. The Dipsern engine recomputes these segments per-asset, because BTC at -50% behaves nothing like a freshly-listed L2 at -50%. Bitcoin has a 12-year history at every drawdown band and a 90%+ win rate at -50% to -60%. A newly-launched token at the same drawdown has no history at all — the segmentation correctly flags that with a low sample size, so you don't mistake noise for signal.
Browse the full crypto signals list to see per-asset numbers. Pages like BTC-USD, ETH-USD, and SOL-USD show the segmentation broken out by drawdown band.
Methodology Notes
- Universe: 150 active crypto assets in Dipsern's database (filtered for liquidity, removed dead pairs)
- Date range: 2013-01-01 to present, where data exists per asset
- Forward window: 90 calendar days (default Dipsern parameter)
- Aggregation: Per-asset segmentation, then sample-size-weighted average across assets
- No look-ahead: EMA at time t uses only returns revealed by t (we test this aggressively in the engine)
The Honest Takeaway
Crypto's "buy the dip" thesis is real but bounded. The historical evidence supports accumulation in the -30% to -70% drawdown band for assets with at least 2 years of history and meaningful liquidity. Below -80%, you're in lottery territory — sometimes the comeback is spectacular, more often the asset never recovers.
Dipsern doesn't tell you which crypto to buy. It tells you what the data has historically said at the current drawdown level for each specific asset. The decision is still yours.