The two September bot rules, re-tested on 7 years of prices
Volume spike: 2,496 trades from Oct 26, 2019 to Oct 1, 2026, +1.31% a trade on average, −1.32% for the middle trade, 45% won, luck test 2.9. On paper since Sep 29: 15 trades closed (+1.69% on average), 7 open. Alt-bleed pair: 1,285 trades from Feb 25, 2020 to Oct 4, 2026, +0.57% a trade on average, +2.44% for the middle trade, 60% won, luck test 1.2. On paper since Sep 29: 0 trades closed, 4 open. Before fees. Past patterns, not financial advice.
Volume spike, top-40 perpetuals, Bitcoin above its 50-day average
as of · our daily closes, Aug 17, 2017 to Oct 7, 2026
Each trade is held 3 days, with at most 8 open at once and 2 sides a trade (each buy and each sell). Each decision reads only its own close and the closes before it; the paper record starts Sep 29, 2026.
the average trade each year, up or downthe middle tradeunder each year: its trades
Record
Trades
From – to
Average
Trimmed
Middle
Won
Luck test
Fee a side
Before Sep 29, 2026
2,496
Oct 26, 2019 – Oct 1, 2026
+1.31%
−1.05%
−1.32%
45%
2.9
0.65%
On paper since Sep 29, 2026
15 7 open
Sep 29, 2026 – Oct 7, 2026
+1.69%
−0.22%
−2.72%
33%
0.5
0.84%
Without the Bitcoin gate
3,378
Oct 25, 2019 – Oct 1, 2026
+0.51%
−1.66%
−2.04%
43%
1.3
0.25%
§ 02 · Alt-bleed pair
Alt-bleed pair: short the weakest 20% of the top-100 perpetuals in a downtrend, long Bitcoin
as of · our daily closes, Aug 17, 2017 to Oct 7, 2026
Each trade is held 10 days, with at most 8 open at once and 4 sides a trade (each buy and each sell). Each decision reads only its own close and the closes before it; the paper record starts Sep 29, 2026.
the average trade each year, up or downthe middle tradeunder each year: its trades
Record
Trades
From – to
Average
Trimmed
Middle
Won
Luck test
Fee a side
Before Sep 29, 2026
1,285
Feb 25, 2020 – Oct 4, 2026
+0.57%
+2.12%
+2.44%
60%
1.2
0.14%
On paper since Sep 29, 2026
0 4 open
—
—
—
—
—
—
—
The pair paid +0.38% a trade in Binance funding over 213 trades with rates on record: shown beside the record, not taken off it.
§ 03 · How it is tested
How it is tested
as of · methodology
A luck test (t) says how far a result sits from zero, in units of its own noise: about 2 is rarely luck for one test, but many tests at once need more, which is what the first cut is for.
These are the two rules our own trading bots ran in September, recomputed every night from our daily closes, bar for bar as the bots compute them. A coin counts as a perpetual on a day our exchange record lists its perpetual from before that day; perpetuals delisted before our record began are missing from the history.
Every result is before fees: "fee a side" is the fee that would eat the whole average trade. The trimmed average leaves out the best and worst tenth of trades; the middle trade shows what a typical trade did.
Our price history holds today's coins: coins that died before we began collecting are missing, which flatters the older years.
§ 04 · Revisions1 dated
Revisions
as of · every change to its headline numbers, dated
This study is re-run every night on the whole history. It was first written on Oct 6, 2026; its numbers last moved on Oct 6, 2026. Each night its headline numbers moved is listed here, newest first.
Alt-bleed pair: 1,285 trades, +0.57% a trade; on paper 0 closed, 4 open; Volume spike: 2,496 trades, +1.31% a trade; on paper 15 closed, 7 open