Does skipping LB6 entries taken against another engine's open position improve Config B?
FRAME
- Subject: LB6 Config B, a BTC strategy of 6 engines. Each engine reads hourly options data and opens only long or only short (3 long, 3 short). 337 simulated trades, 2025-04-23 to 2026-08-22, net of 0.05% fee per side.
- Why it matters: Config B is the version planned for live; any rule added to it now is cheaper to test than to fix after launch.
- Prior work: lb6-opposite-weaker (SURVIVED): entries taken while another engine holds the opposite side earn less per trade, within engines and with any one engine removed.
- Current best: Config B as is: total +239.4%, max drawdown 9.7 pts, Calmar 24.8.
- Scoreboard: pak Yon judges by drawdown and Calmar: frozen-core3 was killed on drawdown budget, live runs a 15% drawdown kill switch, the 34-sleeve roster was picked by Calmar. [not confirmed for LB6]
- Question: those entries are weaker but still profitable (+37.8% in total across the 110 that a skip rule removes). Does skipping them make Config B both more profitable and safer?
- Scope: only an entry rule is added; signals and exits stay fixed. In-sample; decisive test = data after 2026-08-26. Data:
trading-infracommit53c6993^,docs/lb6_tweaked/reference_tradelists/*_oracle_B_trades.csv, joined to hourly options without look-ahead (poc/join_b.py). - Terms: "opposite entry" = at entry time, an accepted trade from another engine is still open on the other side.
BET
Skipping opposite entries makes Config B more profitable and safer.
- Adjective: more profitable, and safer.
- why: this bet was locked before the one-measure rule, with two measures; both are reported as locked. Under the current format it would be two bets.
- Measured by: total net % (higher = more profitable); max drawdown (lower = safer).
- why: total is the strategy's outcome; max drawdown is the number behind pak Yon's kill switch.
- Because: see Prior work.
- Falsified if: total < 239.4, or max drawdown > 8.7 (not at least 10% lower). Locked in
docs/bets/2026-11-17-lb6-f1.md, commitc272e6e, written before any test ran.
KILL
Sequential replay: trades in entry order; an entry is skipped if an accepted trade from another engine is still open on the opposite side.
| test | falsified if | result | status |
|---|---|---|---|
| total net % | < 239.4 | 201.6 | FAIL |
| max drawdown | > 8.7 | 15.1 | FAIL |
Also observed, not deciding:
| variant | trades | total | max_dd | max_dd_at | calmar | mean | win_rate |
|---|---|---|---|---|---|---|---|
| Config B | 337 | +239.4 | 9.7 | 2026-07 | 24.8 | +0.710 | 62.6% |
| skip opposite | 227 | +201.6 | 15.1 | 2026-03 | 13.4 | +0.888 | 64.3% |
110 entries are skipped, not the 166 opposite entries in the original tradelist: a skipped trade never opens, so later entries often no longer meet an opposite position. The skip variant leads by up to +7.1 pts (2025-10) and trails by up to -37.8 pts (2026-08).
UPDATE
- Status: FALSIFIED.
- What it means: the skipped entries were weak but profitable; removing them cut total by 37.8 pts and deepened the worst drawdown. Calmar fell from 24.8 to 13.4.
- Next bets:
lb6-half-size-opposite(Does trading opposite entries at half size make Config B more efficient?) - Rejected:
lb6-opposite-hedge(true by construction; all engines trade the same BTC, so a long and a short open together always offset (decided 2026-11-17)) - Open: whether position sizing counts as an allowed change (ask pak Yon); whether LB6's scoreboard is Calmar.