Rebalance styles — what happens to your weights
Every Quant backtest holds the top N stocks equal-weighted and re-ranks them on a fixed schedule (every 1, 3, 6, or 12 months). The Rebalance style decides what to do with the weights when a rebalance lands. There are three — and on the backtest you can switch between them or select more than one to overlay and compare on the chart.
To see the difference, picture a top-3 portfolio. After a strong run for one holding, the weights have drifted from an even split to:
A 50% · B 30% · C 20%
1. Equal-weight (default)
Reset everything back to 1/N at every rebalance — even when the same names are held. In the example, A is trimmed from 50% back to 33% and the proceeds top up B and C:
A 50 / B 30 / C 20 → A 33 · B 33 · C 33
This is the textbook "equal-weight, rebalanced" rule. It has the highest turnover and systematically trims winners and tops up laggards — capturing a rebalancing / mean-reversion effect that can help or hurt depending on the factor.
2. Reset on change
Do that full equal-weight reset only when the membership actually changes. If the same top-3 come up again, do nothing — let the weights keep drifting (A stays near 50%). When the roster does change — say C drops out and D enters — reset all three to equal:
(C→D) A 50 / B 30 / C 20 → A 33 · B 33 · D 33
Medium turnover: you skip the "no-op" rebalances but still re-level whenever the line-up shifts.
3. Delta only
Never touch the names that stay. Only sell the dropouts and buy the new entrants — funded by the proceeds plus any new monthly cash, split equally among the entrants. With C→D you sell C's slice and buy D; A and B are left exactly as they are:
(C→D) A 50 / B 30 / C 20 → A ~50 · B ~30 · D ~20
This is the lowest-turnover style and the most "let your winners run": a long-held winner like A keeps compounding at its grown weight instead of being trimmed. The trade-off is concentration — over time one name can come to dominate the basket.
When are they the same?
If the top-N membership is unchanged at a rebalance, Reset on change and Delta only both do nothing (weights just drift), while Equal-weight still re-levels them. The three styles only diverge once the roster turns over.
A worked comparison (real numbers)
Backtest of Momentum, top 5, rebalanced quarterly, from Jun 2016 to Jun 2026, starting with $10,000 plus $1,000/month (about $131,000 invested in total):
| Rebalance style | Ending value |
|---|---|
| Equal-weight | ~$5.11M |
| Reset on change | ~$5.14M |
| Delta only | ~$3.95M |
| S&P 500 (same cash flow) | ~$0.28M |
Two things stand out. Equal-weight and Reset-on-change land almost on top of each other — Momentum's roster turns over so often that "reset only on change" ends up resetting nearly every quarter anyway. And Delta-only trails here (~$3.95M vs ~$5.1M): for this fast-rotating factor, regularly trimming the hottest names back to equal actually helped, so never trimming gave some of that back. For a slower-moving factor like Top-N Market Cap, the gap between the styles is usually much smaller.
These figures are a single, survivorship-biased snapshot of an unusually hot decade for a few semiconductor names. Treat the relationships — the turnover ordering, and when the styles diverge — as the durable lesson, not the exact dollar amounts.
Turnover & taxes. Delta-only trades the least, Equal-weight the most. The backtest ignores taxes, fees, and slippage — in a real taxable account, lower turnover means fewer taxable events, which would tilt the real-world comparison toward the lower-turnover styles relative to what you see here.