Log inSign up

// research

Buy the dip? A backtest of -5% and -10% days on 19 coins

There is a bounce, and it is short. On all 19 coins the day after a -5% or -10% day was better than usual, but over the next 30 days the dip was followed by a worse month than usual on most of them — and buying every dip and selling a month later ended above the random median in only 16 of 38 cases.

3 min read

Six headline numbers from the scans, the backtests and the random-entry check
Two thresholds, 19 coins, six years — every number is from the scans and runs below

Setup

  • A dip: a daily close-to-close drop worse than -5%, or worse than -10%. On BTC that happened on 103 and 13 days since 2020, on the other 18 coins on 133–324 and 25–77 days.
  • Coins and data: the 19 coins of the golden cross study with spot history, Binance spot, daily candles, 2020-01-01 → 2026-10-01.
  • Statistics: 38 Research Lab scans (19 coins × 2 thresholds): the mean return 1, 7 and 30 days after the dip against the coin's usual return over the same horizon, with a multiple-testing correction (q-value), a rotation test and a split into two halves.
  • Trading: buy with 99% of the account at the next open after a dip, sell 30 days later, one position at a time; 0.1% fee and 0.05% slippage per side. Then the platform's null model: 100 runs with the same exits and size, but entries on a coin flip at the same rate as the rule.

The bounce is real — and over in days

For each of 19 coins: the mean 30-day return after a -5% day, after a -10% day and over an ordinary 30 days
Over a month, the dots after a dip mostly sit left of the usual 30 days

The day after a dip was better than an ordinary day on 19 of 19 coins for both thresholds, and that is where every finding that passed the correction sits: 8 of 114 tests, all on the one-day horizon — BTC after a -10% day, for example, +4.59% the next day against the usual +0.15%. A week later the dips were still ahead on 15 of 19 coins (-5%) and 17 of 19 (-10%). Over 30 days the order flipped: a better month than usual followed a -5% day on only 7 of 19 coins and a -10% day on 6 of 19.

Trading it

Total return of buying every -5% or -10% day and selling 30 days later, against holding, for each of 19 coins
Log scale; buy-the-dip runs against buying once and holding

Buying every -5% day made money on 10 of 19 coins, every -10% day on 8 of 19. Against simply holding, the -5% rule came out ahead on 6 of the 14 coins that rose and the -10% rule on 2.

Backtest

Buy the dip: after a -5% day, hold 30 days

BTCUSDT · 1d · Spot · 2020-01-01 → 2026-10-01

Return
-28.34%
Max drawdown
77.72%
Sharpe
0.14
Win rate
50.0%
Trades
40
Open the report →
BTC: buy after every -5% day, sell after 30 days — -28.34%

Backtest

Buy the dip: after a -10% day, hold 30 days

BTCUSDT · 1d · Spot · 2020-01-01 → 2026-10-01

Return
+124.82%
Max drawdown
36.73%
Sharpe
0.55
Win rate
66.7%
Trades
12
Open the report →
BTC: buy after every -10% day, sell after 30 days — +124.82%

Was the dip the reason?

For each of 19 coins: the percentile of the two rules among 100 random-entry runs
Below 50, random entry dates with the same exits did better

A rule can make money simply by being in a rising market. The null model takes that away: the same 30-day exits and the same size, with entries on a coin flip. Buying the dip ended below the random median in 22 of 38 rule × coin pairs — the -5% rule was above it on 11 of 19 coins, the -10% rule on 5 of 19 — and in the top 10% in 3, about the 3.8 that chance alone gives.

BTC and ETH: the rule's return against the spread of 100 random-entry runs
BTC after a -10% day is the one that beat most random dates — on 12 trades

Conclusion

  • After a big down day, crypto tended to bounce the next day — on every one of 19 coins.
  • The bounce did not last: over a month, the dip was followed by a worse-than-usual 30 days on most coins.
  • Buying every dip and holding for a month was close to chance — above the random median in 16 of 38 cases.

More research: funding as a signal and DCA or lump sum.

Test your own idea

Create a free account, start from a template or from your own rules, and read your first report.

Create a free accountBrowse the leaderboard

More articles