Relative Strength Rotation Strategy: 69 of 90 segments profitable, deepest drawdown 72.7%

Bets on triple confirmation (MA, RSI, MACD) before entering and rides trends; signals lag, so repeated crossovers bleed capital in range markets.

On this BTCUSDT 1d data, the Relative Strength Rotation Strategy had 69 profitable segments out of 90 splits, so the surface win rate is not low; but the median segment return was +28.7%, while the average buy-and-hold return across the same segments was +101.7%, meaning most splits did not beat 'doing nothing'. Its highest segment return of +592.8% is eye-catching, and the lowest segment return of -61.5% and the deepest drawdown of 72.7% across 90 segments are equally striking: this is not a strategy that trades small drawdowns for steady returns, but one that uses larger volatility to capture the right-tail returns of a few strong trend segments. It bets that relative strength persists: the recently stronger direction keeps being strong after rotation, capital switches to the strong side, and profits come from trend continuation. The problem lies exactly there—once the strength relationship reverses quickly, rotation signals fail repeatedly. The worst segment, round 11 segment 2 (2021-02-21 to 2022-07-21), had a segment return of -61.5% and a deepest drawdown of 72.7% for that segment, which was also the deepest among all 90 segments, showing that in a persistently weakening environment after a trend reversal it may expose positions to the wrong direction; choppy markets and low-volatility environments are also unfriendly, because strength rankings are unstable and frequent switching amplifies false signals and trading friction. The most decision-useful numbers for readers are not the highest +592.8%, but the combination of 69/90 profitable segments and median +28.7%: most splits made money, but not much; at the same time, the lowest -61.5% and the deepest drawdown 72.7% tell us that a few splits can lose heavily. The best segment, round 4 segment 1 (2020-01-01 to 2021-03-11), had a segment return of +592.8% and a deepest drawdown of 29.3% for that segment; it looks strong, but this is more like a right-tail sample and cannot represent ordinary splits. Average profit per trade was $34,869 and average loss per trade was $4,470; the former is clearly higher than the latter, but this is in dollar terms and is affected by position size and capital scale, so it cannot be directly treated as a percentage edge; 0 liquidations across 90 segments also does not mean low risk—the 72.7% drawdown already shows that the holding process may be very uncomfortable. Methodologically, one should remain clear-headed: these 90 backtests are not 90 independent samples, but 90 split variants obtained by cutting the same historical data into 6 non-overlapping windows, repeating 15 rounds, and randomly jittering the boundaries; the random seed has been stored in the database, and the same seed can reproduce the same split. This can test split sensitivity, but it cannot replace independent validation across markets and cycles. The sample covers only one interval of this historical data. Regarding parameters, interval, ma_period, rsi_period, rsi_threshold, and macd_params were not given in the strategy description; this article uses the daily timeframe and common parameters by industry convention, and copies the rest as-is; therefore the results can only be regarded as a historical performance test under this set of assumptions, not as a strict replication of the original strategy implementation—what was not replicated is precisely the specific parameters and implementation details. The numbers most worth watching are: 69/90 profitable segments show that most splits did not lose money; median segment return +28.7% versus buy-and-hold average +101.7% shows that most splits underperformed holding; deepest drawdown 72.7% shows that tail risk is heavy. It may be effective when strong trends continue, and ineffective in trend reversals, choppy markets, and low volatility, and it is sensitive to parameters and split method. This content is a reproduction and validation based on historical data and does not constitute investment advice; this platform does not recommend, rate, or rank.