Fair Order Matching on LOBSTER NASDAQ
0.85SpreadUnconstrained PPO
Evaluation results
| Method | Links | |||||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|
| Unconstrained PPOMethod Category=Learned Baseline2026.04 | 0.85 | 5,250 | 95 | 8,500 | 12 | 0.95 | 45.5 | 100 | — | — | — | |
| CPO-FOAMMethod Category=Ours2026.04 | 0.95 | 4,850 | 91 | 8,150 | 15 | 1.02 | 1.2 | 2.5 | 5 | 5 | 1.8 | |
| No Spectral NormMethod Category=Ablation Studies2026.04 | 0.98 | 4,680 | 89 | 7,950 | 16 | 1.05 | 22.5 | 6.5 | 25 | 15 | 8.5 | |
| PI-only ControllerMethod Category=Ablation Studies2026.04 | 0.98 | 4,720 | 90 | 7,850 | 16 | 1.04 | 1.8 | 4.5 | 18 | 12 | 4.5 | |
| P-only ControllerMethod Category=Ablation Studies2026.04 | 1.02 | 4,550 | 88 | 7,650 | 17 | 1.06 | 2.8 | 7.2 | 35 | 20 | 9.5 | |
| No PID (ξ = 0)Method Category=Ablation Studies2026.04 | 1.05 | 4,250 | 85 | 7,450 | 18 | 1.08 | 5.8 | 11.2 | 75 | 28 | 19.5 | |
| Lagrangian PPOMethod Category=Learned Baseline2026.04 | 1.15 | 3,840 | 76 | 6,250 | 25 | 1.12 | 12.4 | 28.5 | 145 | 85 | 45.2 | |
| No Trust RegionMethod Category=Ablation Studies2026.04 | 1.18 | 3,800 | 78 | 6,450 | 28 | 1.14 | 11.5 | 25.5 | 130 | 82 | 41 | |
| PID-LagrangianMethod Category=Learned Baseline2026.04 | 1.2 | 3,750 | 74 | 6,100 | 28 | 1.15 | 9.5 | 14.5 | 45 | 35 | 12.5 | |
| RCPOMethod Category=Learned Baseline2026.04 | 1.25 | 3,650 | 72 | 5,850 | 30 | 1.18 | 10.5 | 22.4 | 112 | 72 | 38.4 | |
| Vanilla CPOMethod Category=Learned Baseline2026.04 | 1.35 | 3,580 | 70 | 5,450 | 32 | 1.2 | 6.5 | 12.4 | 64 | 45 | 18.2 | |
| No KL ProjectionMethod Category=Ablation Studies2026.04 | 1.38 | 3,500 | 72 | 5,800 | 36 | 1.22 | 4.5 | 9.5 | 42 | 22 | 11.5 | |
| No Recovery StepMethod Category=Ablation Studies2026.04 | 1.4 | 3,250 | 68 | 5,100 | 42 | 1.28 | 1.5 | 5.5 | 15 | 10 | 3.8 | |
| IPOMethod Category=Learned Baseline2026.04 | 1.45 | 3,420 | 68 | 5,200 | 35 | 1.25 | 8.2 | 15.6 | 85 | 55 | 25.4 | |
| FIFOMethod Category=Rule-Based Baseline2026.04 | 2.45 | 1,250 | 42 | 4,200 | 85 | 1.45 | 18.5 | 100 | — | — | — | |
| Size-Time Interp.Method Category=Rule-Based Baseline2026.04 | 2.65 | 1,650 | 40 | 4,100 | 88 | 1.52 | 20.1 | 100 | — | — | — | |
| Pro-RataMethod Category=Rule-Based Baseline2026.04 | 3.1 | 1,840 | 38 | 3,850 | 92 | 1.6 | 25.4 | 100 | — | — | — |