โ€บNavigation

Axis-2 ยท Cascade-Free Harvest โ€” Iteration 1 12 NEW

Family swept: order-flow toxicity & informed-trading (PIN family ยท spread-decomposition ยท trade-sign crowding) ยท campaign regime-invariant-orthogonality ยท append-only

12new candidates
0dupes
28running total (spot-legal/L2)
+3deriv (out-of-scope)

What this iteration deepened. The kickoff seeded VPIN / OFI / CVD / Hawkes for order flow. This firing exhausts the informed-trading & toxicity lineage around them: the full PIN family (PIN โ†’ MPIN โ†’ AdjPIN โ†’ iVPIN), two structural spread-decomposition models that isolate an adverse-selection (toxicity) component from trades only (MRR, Glostenโ€“Harris), the trade-signing primitives that feed every imbalance measure (BVC, Lee-Ready, tick rule), and the trade-sign crowding line (Lilloโ€“Farmer long-memory, Bouchaud propagator, and the 2023 Satoโ€“Kanazawa order-splitting inference).

Spot-legality. Binance spot aggTrades carry the exact aggressor side (is_buyer_maker), so trade signs are known โ€” Lee-Ready (H07) is the only needs-L2 item and is redundant here, catalogued only because BVC/tick benchmarks cite it. Everything else is spot-legal per principle #8. Derivatives-only toxic-flow ML papers (funding/quote/proprietary features) were reviewed and left out-of-scope.

New candidates (iter 1)

#NameFamilyMeasuresCascade / liquidity mappingSpot?ParamStable-IDFOSS
H01PIN (EKOP)informed-tradingPoisson buy/sell mixture MLE โ†’ fraction of information-motivated tradesadverse-selection/toxicity; high PIN โ†’ wider spread, higher exit costspotPPEasley, Kiefer, O'Hara, Paperman 1996, J. Finance โ€” DOI 10.1111/j.1540-6261.1996.tb04074.xVER monty-se/PINstimation (R, GPLโ‰ฅ3); mgao6767/frds (Py, MIT)
H02MPINinformed-tradingmultilayer PIN โ€” detects N info-event layerstoxicity flag robust on multi-news days where PIN biasesspotPPErsan 2016 โ€” SSRN 2874420 / DOI 10.2139/ssrn.2874420 ยท parent H01VER PINstimation (R, GPLโ‰ฅ3)
H03AdjPINinformed-trading / spread-decompadds symmetric order-flow shocks (PSOS) โ†’ splits info-asymmetry from illiquidityseparates toxic adverse-selection from benign symmetric liquidity shocksspotPPDuarte & Young 2009, JFE โ€” DOI 10.1016/j.jfineco.2007.10.008 ยท parent H01VER PINstimation (R, GPLโ‰ฅ3)
H04iVPINorder-flow toxicityMLE-based VPIN, stable on small volume buckets / rare informed tradesflow-toxicity โ†’ fragility / flash-crash early warningspotPPKe & Lin 2017, Critical Finance Review โ€” DOI 10.1561/104.00000046 ยท parent S11 VPINVER PINstimation (R) ivpin()
H05MRRspread-decompositionstructural GMM: ฮ”price โ†’ adverse-selection ฮธ + transitory ฯ† + order-flow persistence ฯฮธ = toxicity cost; ฯ = order-flow crowding/persistencespotPPMadhavan, Richardson, Roomans 1997, RFS โ€” DOI 10.1093/rfs/10.4.1035UNV short GMM; no canonical repo
H06Bulk Volume Classificationtrade-sign / toxicity feedbuy-fraction per bar via t-CDF of standardized price change (no per-trade signing)feeds VPIN-style imbalance; info-linked trade flowspotPPEasley, Lรณpez de Prado, O'Hara 2016, JFE โ€” DOI 10.1016/j.jfineco.2016.01.018VER jktis/Trade-Classification-Algorithms (Py)
H07Leeโ€“Readytrade-sign classificationquote-midpoint + tick-rule trade signingsigning primitive; redundant on Binance (aggressor exact)L2MNLee & Ready 1991, J. Finance โ€” DOI 10.1111/j.1540-6261.1991.tb02683.xVER jktis/Trade-Classification-Algorithms (Py); PINstimation (R)
H08Tick rule / tick testtrade-sign classificationsign trade by price move vs prior trade (uptick=buy)trade-only signing primitive; underpins BVC benchmarksspotPFdefined/evaluated in Easley-Lรณpez de Prado-O'Hara 2016 JFE โ€” DOI 10.1016/j.jfineco.2016.01.018VER jktis/Trade-Classification-Algorithms (Py)
H09Order-flow long memorytrade-sign crowdingsigned-flow ACF decays ฯ„โˆ’ฮณ (Hโ‰ˆ0.7): persistent signed order flowmetaorder / order-splitting crowding; directional-crowding & capacity signalspotDIMLillo & Farmer 2004, SNDE 8(3) โ€” DOI 10.2202/1558-3708.1226UNV nolds (Py, MIT โ€” generic Hurst/DFA); no flow-specific repo
H10Propagator / response fntrade-sign crowding / impactR(โ„“)=โŸจsignยทฮ”pโŸฉ response + propagator G(โ„“): permanent vs transient signed-flow impacthow signed flow moves price (impact/fragility); offsets long-memory to keep price diffusivespotPFBouchaud, Gefen, Potters, Wyart 2004, Quant. Finance 4(2) โ€” arXiv:cond-mat/0307332; DOI 10.1080/14697680400000022UNV
H11LMF order-splitting inferencetrade-sign crowding (SOTA 2023)infers metaorder-length dist P(L)โˆLโˆ’(ฮฑ+1) & #splitting traders from sign-ACF exponentquantifies metaorder crowding / splitting-trader count โ†’ capacity + crowding readspotDIMSato & Kanazawa 2023, PRL 131 197401 โ€” arXiv:2301.13505; DOI 10.1103/PhysRevLett.131.197401 ยท parent H09UNV macro leg spot-legal; full validation needs trader IDs
H12Glostenโ€“Harrisspread-decompositionregress ฮ”p on signed volume โ†’ adverse-selection (volume-scaled) vs transitory spreadadverse-selection component = toxicity; transactions-only (no quotes)spotPPGlosten & Harris 1988, JFE 21(1) โ€” DOI 10.1016/0304-405X(88)90034-7UNV short OLS; frds has related spread measures

FOSS anchors surfaced this iteration

Scope & provenance notes
Next angle โ†’ liquidity-cascade / fire-sale / feedback-loop risk metrics (systemic fragility, connectedness, illiquidity spirals). ยท Data SSoT: axis-2-cascade-free/CANDIDATE-CATALOG.md ยท back to loop hub