Should LB6 skip trades that go against another engine?
No. Those trades are weaker, but skipping them cuts LB6's total return from +239% to +202% and deepens its worst drawdown from 9.7 to 15.1 points.
What LB6 is
LB6 is a BTC trading system at PekkaLabs. It runs 6 small strategies, called engines, side by side, each trading only long (bet the price rises) or only short (bet the price falls). The data here is a simulation of the latest version, Config B: 337 trades from April 2025 to August 2026.
Where the question comes from
The report Are LB6 trades that go against another engine weaker? found that trades opened while another engine holds the opposite side earn +0.35% on average, against +1.06% for all other trades.
Weaker, but still profitable. So the question is: if LB6 gets a rule "do not open a position while another engine holds the opposite side", does it do better?
How it was tested
LB6's trades were replayed in time order. Whenever another engine held the opposite side, the new trade was skipped.
The pass criteria were written down before testing. The rule counts as an improvement only if both hold:
- Total return is higher than without the rule.
- Worst drawdown (deepest fall from a peak) drops by at least 10% from 9.7 points, to 8.7 points or less.
The result is worse
The rule skips 110 trades, not 166. A skipped trade never opens, so later trades often no longer meet an opposite position.
| no rule | skip rule | |
|---|---|---|
| trades | 337 | 227 |
| total return | +239.4% | +201.6% |
| worst drawdown | 9.7 pts (July 2026) | 15.1 pts (March 2026) |
| mean per trade | +0.71% | +0.89% |
| winning trades | 62.6% | 64.3% |
The mean per trade does rise. But the 110 skipped trades still add up to +37.8%, so total return falls.
Until December 2025 the skip rule is actually slightly ahead, by up to about 7 points in October 2025. From January 2026 it falls behind, and by the end of the data the gap is about 38 points.

The drawdown also gets deeper: 15.1 points with the skip rule, against 9.7 points without it.

Neither criterion holds, so the skip rule is not used.
Why, and what to try next
Opposite trades work as a hedge by construction: every engine trades the same BTC, so while one engine is long and another is short, a loss on one side is offset by the other. That needs no test. Remove those trades and LB6 is more often one-sided, so when the market turns, every position loses at once, which fits the deeper drawdown.
The open question is how much of that offset to keep. The middle path is to shrink opposite trades, for example to half size, instead of skipping them. That idea will be tested as its own report, with pass criteria written before the data is looked at.
Limits
- All numbers come from the same simulated data where the pattern was found. The real test is new data after August 2026.
- "Total return" is the sum of each trade's %, after a 0.05% fee per side. Real position sizes and leverage are not modelled.
- It is not yet confirmed whether changing position size is a change pak Yon allows.