Daily Close Price Rate-of-Change Strategy: 66 of 90 segments profitable, but deepest drawdown 56.7%

Bets on momentum continuing once the daily close turns up: it goes long as soon as the rate of change turns positive and exits when it turns negative, so choppy markets stop it out repeatedly.

In 90 split backtests of the Daily Close Price Rate-of-Change Strategy on BTCUSDT daily data, 66 segments were profitable and 0 blew up, suggesting a positive-return bias; but the median segment return was only +28.7%, while the average buy-and-hold return across the same segments was +112.6%, indicating that most of the time it did not beat doing nothing. The numbers to look at first are profitable segments 66/90 and deepest drawdown 56.7%—the former shows most split methods can make money, while the latter shows that under extreme splits the equity curve can be cut by more than half. From the return distribution, the best segment was Round 4 Segment 1, 2020-01-01 to 2021-02-24, with segment return +174.5% and deepest drawdown in that segment 20.2%; the worst segment was Round 9 Segment 2, 2021-03-11 to 2022-07-10, with segment return -49.2% and deepest drawdown in that segment 50.8%. The same rule set went from large gains to near halving across different windows, showing that results are highly sensitive to start point and market state. Average profit per trade $15,132 and average loss per trade $2,902 are monetary measures and are affected by position sizing assumptions, so they cannot be directly treated as strategy edge; they only show that after a trade is triggered, the profit/loss magnitude is not symmetric, but the drawdown data deserves more caution. What is this strategy betting on? It treats the rate of change of the daily close price as a trading signal, with a default threshold of 0, meaning a position adjustment may be triggered whenever price direction changes. The core assumption is that the close price rate of change contains tradable short-term directional information, and that this direction has persistence over the following period, at least enough to cover the cost of frequent rebalancing. If the market is in a trend segment, such as the sustained one-way move from early 2020 to early 2021, the signal can easily capture the main swing; but if the market enters range-bound or high-volatility reversal conditions, the rate of change will repeatedly cross zero, and the strategy will keep getting hit in the opposite direction. The worst segment from March 2021 to July 2022 is an example of such a failure environment. Its weaknesses therefore concentrate in three types of conditions: ranging markets, low-volatility noise, and parameter sensitivity. The threshold, lookback window, and position were not given in the original; here, following industry convention, threshold default 0, lookback_bars compared with the previous close, and position.value full-position adjustment were filled in. A threshold of 0 means almost all directional changes are treated as signals, so trading frequency and noise exposure will be very high; if the actual strategy has a stricter threshold or filter conditions, the results could be completely different. The 90 backtests are also not 90 independent samples, but rather the same history split into 6 non-overlapping windows, repeated 15 rounds with randomly jittered boundaries. The random seed has been stored, so it is reproducible, but sample independence is limited. The comparison with buy-and-hold is the one thing to remember most from this validation. The strategy's median segment return was +28.7%, the highest +174.5%, the lowest -49.2%, and the deepest drawdown 56.7%; over the same period, doing no trades and simply holding the asset gave an average segment return of +112.6%. This means the strategy may only catch up to or exceed holding in a few trend windows, while in most split methods the median return is clearly below the buy-and-hold benchmark, yet it bears a deepest drawdown of 56.7%. In other words, it is not a solution that can confidently replace holding; it is more like a timing attempt sensitive to specific market states. This content is a reproduction and validation of historical data, does not constitute investment advice, and this platform does not recommend, rate, or rank.