Trend table
4h candles, Binance USD-M futures, top 30 by volume.The only family of rules that survived our tests, and it is not a guarantee.Trend-following rules were positive against the coin basket in three separate periods (spot 2017–2020, futures 2020–2023 and 2024–2026), while mean-reversion, funding and seasonality rules lost. The edge was weaker in the newest period, drawdowns of 25–55% occurred, and the coin list contains only survivors. It is a candidate for a forward test, not proof.
Forward journal
Every change of state is written at the moment it happens, with the price. This is the honest out-of-sample record.
The rules
- Momentum: the close is above the close 60 four-hour bars (10 days) ago.
- Donchian: enter when the close breaks the highest high of the previous 100 bars, leave when it falls below the lowest low of the previous 50.
- Moving averages: the 50-bar average is above the 200-bar average.
- Uptrend = at least two of the three rules are bullish. Closed candles only; the state is for the next bar.
Backtest of exactly this vote against an equal-weight basket (costs 0.05% per side and actual funding included): alpha +54%/year in 2017–2020 (t=2.4), +33% in 2020–2023 (t=1.8), +18% in 2024–2026 (t=1.3). Positive each time, shrinking, and not yet statistically strong.
Crypto trend table: momentum, Donchian and moving averages on 4h
This table shows the trend state of the top 30 Binance USD-M futures on 4h candles using three classic rules: momentum over ten days, a Donchian breakout with a 100-bar entry and a 50-bar exit, and the 50/200 moving average relationship. A market is in an uptrend when at least two of the three rules are bullish. The state is computed from closed candles only and applies to the next bar, exactly as in the backtest.
Every change of state is written to a forward journal with the time and the price, so the page builds its own out-of-sample record from the day it was switched on.
What the tests say
We searched 133 rules on 4h data: trend following, mean reversion, funding, seasonality and cross-sectional momentum, with costs of 0.05% per side and actual funding. No single rule passed our strict threshold, but the trend-following families were positive against an equal-weight coin basket in three separate periods: spot 2017–2020, futures 2020–2023 and futures 2024–2026. Mean-reversion, funding and seasonality rules lost in every period. For the exact 2-of-3 vote used here the alpha was +54% a year in 2017–2020, +33% in 2020–2023 and +18% in 2024–2026 with t-statistics of 2.4, 1.8 and 1.3.
What it does not tell you
The effect is shrinking, the sample is 15 instruments, the coin list contains only survivors, and drawdowns of 25–55% occurred even in the good periods. Trend following also lags: it buys after a rise has started and sells after a fall has started. The table is a candidate for a forward test and context for your own analysis, not a recommendation.
Questions and answers
Is this a trading signal?
No. It is a rule-based description of the current trend with an honest history of how such rules behaved.
Why 4h and not 15m?
The rules were tested on 4h bars, where costs are small compared with typical moves. Short timeframes lose the edge to fees.
How are Telegram alerts related?
When the vote of a market flips, the owner receives a message. The same flips appear in the forward journal on this page.
