Data snapshot: 2026-06-30 22:22 UTC (stag/paper combined). Cohort = strategies created in June (creation inferred from
uptime). Pre-June strategies that ran into June are excluded from the cohort total; their lifetime P&L was restated by
the Jun-3 dust closure and is not reliable (see MRRP incident note).
The OKX MM v16 arms — the breakeven story. Huge volume, ~63-70% win rate, per-fill P&L near zero.
Strategy
Status
P&L
Trades
Volume
Win Rate
Margin (bps)
MMTV_v16_eth_okx_basiscancel
ACTIVE
+$0.11
1956
$49,188
66%
+0.02
MMTV_v16_btc_okx_basiscancel
ACTIVE
-$0.69
1643
$48,771
67%
-0.14
MMTV_v16_btc_okx_sigcancel2
ACTIVE
-$0.12
6051
$180,352
70%
-0.01
MMTV_v16_eth_okx_sigcancel2
ACTIVE
-$1.06
8072
$202,101
69%
-0.05
MMTV_v16_eth_okx_exitonly
ACTIVE
-$1.56
6962
$175,312
66%
-0.09
MMTV_v16_btc_okx_exitonly
ACTIVE
-$1.72
5317
$158,797
63%
-0.11
MeanReversion_v21_loosen_fast_eth
ACTIVE
+$0.17
1
$50
100%
+33.4
MR_v21 (tight_sr ×2, loosen_fast_btc)
ACTIVE
$0.00
0
—
—
—
Active cohort total: -$4.88 — all near-flat; the v16 arms are the "OKX MM at breakeven" evidence (exit-policy
baseline exitonly, +imb5 sigcancel2, +basis A/B basiscancel).
v16 arm churn — cost of driving OKX MM to breakeven + A/B levers
StatsArb
-$1.88
v20 ETH/BTC (retired — not cointegrated)
MeanReversion
+$0.58
v19→v21; live-positive but knife-edge
Read: like May, the cost is almost entirely MarketMaking iteration — but this time it bought a result (breakeven
+ the exit-policy finding + the follower/basis path), not just a verdict. The two KILLED arms (r2, r5 at -$5
each) were naive controls, killed as designed.
only positive deriv sub-item; Sharpe ~0.65, decaying, 2-3× cost
derivatives access
BUS cross-venue MM
+0.5-0.7 bps/fill with rebate
volume-capped ~$/day — option/relationship
All survivors sit in the Risk-transfer root of the edge taxonomy (paid to bear risk, not to predict). Nothing in
the Information (prediction) or Structure (scheduled/mechanical) roots survived. Carry ≈ 0 net and cross-sectional
−140%/yr are dead; neither is additive to VRP.
non-viable every axis (kline, funding/OI, order-book, cross-sectional L/S)
Order-book OFI microstructure
signal real (IC ~0.20) but sub-fee as a taker + rebate-positive fills can't be gated as a maker — a cost/economics wall, not latency (which is untested — decay-curve question)