September 1, 2026 – September 30, 2026

September 2026 rebalancing bot results: +9.21% vs BTC +8.71%

Versus the same 25/25/50 start with no rebalancing: +2.62 pts

September was a good month for my rebalancing bot, though not a spotless one. The portfolio returned 9.21% from September 1 to September 30. Bitcoin returned 8.71% over the same stretch and QQQ returned 3.95%. So the bot finished slightly ahead of bitcoin and well ahead of the Nasdaq fund. The setup is simple: 25% BTC spot, 25% SOXL perpetual at 1x long, and 50% USDT on Bybit. With half the portfolio in USDT, I did not expect to keep up with bitcoin in a rising month, so I am a little surprised by this result.

The longer view is more sober. At month end the bot's cumulative return is 0.98%, against 32.81% for bitcoin and 0.29% for QQQ. One good month does not change the fact that holding bitcoin alone has been far ahead of this strategy over the whole period. The bot is only a little above QQQ on that cumulative measure. Anyone looking for rebalancing bot results should keep both numbers in mind: the monthly win and the long gap behind bitcoin.

The best day was September 21, when the bot gained 5.12%, while bitcoin gained 6.09% and QQQ gained 1.8%. The worst day was September 28, when the bot lost 3.35%, bitcoin lost 1.63% and QQQ lost 0.72%. I have no notes on what drove either day, so I will not guess at causes. What the numbers do show is that on the worst day the bot fell by more than twice as much as bitcoin. Holding half the portfolio in USDT clearly did not cushion that day. The month had 17 up days and 13 down days.

The rebalancing is the part I am most interested in. The rule is a hourly check, with a rebalance whenever a weight drifts 2 percentage points or more from target. This month it made 9 trades: 6 buys and 3 sells, with 2 in BTC and 7 in SOXL. SOXL went from 113.65 to 146.81 and bitcoin went from 77,862.1 to 83,823.3, so SOXL did most of the moving and most of the trading. Measured against a buy-and-hold portfolio that started at the same 25/25/50 weights, the rebalanced version came out 2.62 percentage points ahead. That is the effect people usually call Shannon's demon: selling a bit of what ran up and buying what lagged, so volatility turns into a small gain. At month end the weights were 25.1% BTC, 25.46% SOXL and 49.44% USDT, which is very close to where I started.

The drawdown needs an honest mention. The bot's worst peak-to-trough drop within the month was 6.05%. Bitcoin's was 5.23% and QQQ's was 2.21%. So the bot fell further than bitcoin at its worst, even though bitcoin is only 25% of the portfolio and half the money sits in USDT. The SOXL position is the likely reason, since it is the more volatile asset in the mix. I can see that in the worst-day numbers, but I cannot prove it from this data alone. Two days this month also included deposits, so my own cash flows are part of the picture and I try not to read too much into small differences.

The lesson for me is that a month where rebalancing beat buy-and-hold by 2.62 points is not proof that the method works. It helped because SOXL moved a lot and the bot traded around it. In a month with less movement, or a steady trend in one direction, the same rules could do nothing or lag. I also should not treat the USDT half as protection when the SOXL leg can still hand me a 6% drawdown. I will keep running the bot as it is, keep logging the results, and judge it over many months rather than this one.

Daily entries this month

How rebalancing works: Shannon's Demon →

Methodology

Total equity is logged by the bot every hour at :59. Deposits are removed at the hour they arrive and returns are chained as a time-weighted return (TWR). Daily return compares the last snapshot of the day with the previous day's last snapshot. Benchmarks use Bybit BTCUSDT spot and QQQUSDT perpetual hourly closes.

This site is a personal record, not investment advice. Past performance does not guarantee future results.