Resize rule widened: about a third fewer copied orders

From the 08:00 UTC cycle on 14 October 2026 the book resizes a position it already holds only when the target size has moved further from the current size than before. The entry and exit signals, the four-coin limit, the 25% cap per coin, the stop on every position and the profit lock are unchanged. No order was sent by the change itself; it only raises the bar for future adds and trims.

Why

In the first 15 days of the demo account, 22 of 37 fills were adds or trims of positions already held; one ETH position produced eight fills. On a copy-trading platform every lead fill is copied to every follower as a separate order with its own fee, and a fill that scales to less than the exchange minimum is skipped on the follower's account, so small followers drift away from the lead. Fewer and larger resizes cost followers less and keep them closer to the book.

Evidence

The change was tested on the backtest, January 2021 to September 2026, at 0.07% and 0.15% per unit traded with Binance funding, against the rule as traded until now, with the acceptance limits fixed before the runs. Orders fall from 689 to 459 a year, fills per position from 4.9 to 3.3, and resizes below 5% of equity from 302 to 89 a year. Return and drawdown are unchanged within the backtest's noise: CAGR 44.1% against 41.7% at 0.07%, maximum drawdown 20.4% at 0.15% in both cases, and both halves of the sample, 2021 to 2023 and 2024 to 2026, equal or better. The difference in return is not evidence of improvement; settings next to the chosen one sit at the old figures. A simulated follower with 1,000 USDT has 0.9% of orders skipped instead of 2.9% and a tracking error of 1.8% a year instead of 2.1%. The one measurable cost: a coin can drift to about 30.5% of equity before a trim, against 28.8% under the old rule.

Two of three independent reviews could not refute the result; the third corrected how the peak coin weight was measured and confirmed the order reduction. Alternatives that cut more orders were rejected: a higher entry threshold is a knife edge on the score grid, a 50% band costs two points of drawdown, and daily-only resizing weakened 2024 to 2026.

Also changed

The drawdown circuit breaker halves every position when it fires. That halving now bypasses the resize filter, so a small position is halved too; under the wider rule it could have been held back.

Effect on the published figures

The tear sheet on this site is the backtest of the rule as traded until 14 October. The new resize rule changes those figures by less than their sampling noise, so they stand until the next scheduled refresh, and the live record from 14 October is produced by the rule described here. Expected live order pace: about 580 a year instead of 870.