Optimal Bankroll Fractalization in Oxbet s Multi-Layered Market Structure
Oxbet s enjoin book operates across three distinct liquid tiers primary feather, synthetic, and dark each with distinct latency, slippage, and untoward natural selection profiles https://oxbett.jp.net/. The uninformed approach of unvarying roll storage allocation fails to exploit the non-linear risk premiums embedded in these layers. Instead, deploy a fractalized roll model where capital is partitioned off according to the Hurst index of each tier s unpredictability touch.Primary liquid state exhibits H 0.65, indicating mean-reverting deportment with finite drawdowns. Allocate 40 of bankroll here, shrew-sized via Kelly criterion well-adjusted for the tier s operational half-life( 120ms). Synthetic liquidity, with H 0.8, is trending and requires a 30 storage allocation, but only if the synthetic substance-to-primary spread out exceeds 1.5x the real median. Dark liquid state(H 0.95) demands the unexhausted 30, but only when the dark pool s fill probability, plagiarised from a hidden Markov model of counterparty demeanour, exceeds 70.
Latency Arbitrage with Predictive Order Flow Imbalance
Oxbet s matched engine processes orders in 50 s batches, creating microsecond-scale arbitrage windows between the primary quill and synthetic tiers. Traditional many-sided arbitrage models fail here because they don fast execution. Instead, simulate the enjoin flow imbalance(OFI) as a random work with a jump-diffusion part:dOFI(OFI)dt(OFI)dW JdNWhere J is the jump size distribution(empirically fitted to a Pareto with 2.3) and N is a Poisson work on with volume(t), calibrated to Oxbet s existent say book readjust relative frequency( 12Hz). The arbitrage signal is triggered when the predicted OFI at t 50 s exceeds the synthetic tier s bid-ask spread out by 3, well-adjusted for the tier s liquidness skew.
Adverse Selection Mitigation via Counterparty Fingerprinting
Oxbet s dark pool is impressionable to cyanogenetic tell flow from high-frequency commercialize makers who work stale quotes. Detect these counterparties by cluster their say submission patterns using a variational autoencoder trained on 30-dimensional sport vectors: entomb-order arrival time, size distribution entropy, and cancellation rotational latency. The encoder s latent space reveals two distinct clusters”informed”(low S, high cancellation rate) and”uninformed”(high entropy, low rate).Route orders to the dark pool only when the probability of veneer an knowing counterparty, estimated via a Bayesian supply statistical regression on the latent space, falls below 15. For hep clusters, preemptively let out the unfold by 0.5bps and reduce tell size to the 25th percentile of the dark pool s real fill statistical distribution.
Synthetic Liquidity Provision with Dynamic Skew Control
Synthetic liquidity providers on Oxbet must dynamically adjust their skew to keep off unfavourable selection from latency arbitrageurs. The optimal skew is derivable from a stochastic verify problem where the supplier s stock-take I(t) evolves as:dI(q_ask- q_bid)dt dWThe verify variable star is the skew s(t), distinct as the difference between the synthetic ask and primary quill bid. The supplier s object glass is to maximize expected P L over a tensed purview, subject to a terminus take stock punishment. The Hamilton-Jacobi-Bellman equation yields a unsympathetic-form root for s(t) as a run of I(t) and the primary tier s order book instability. Implement this via a search hold over indexed by I(t) and the unbalance s z-score, updated every 10ms.
Cross-Tier Hedging with Cointegrated Pairs
Oxbet s primary feather and synthetic tiers often diverge due to temporary worker liquid state imbalances, creating cointegrated pairs that regress at predictable half-lives. Identify these pairs using a Johansen test with a rolling 5-minute window, then hedge positions across tiers using a mean-reverting spread out simulate:dS(- S)dt dWWhere S is the spread between the synthetic and primary feather mid-prices, is the mean turnabout speed up(empirically 0.3s), and is the long-term equilibrium unfold. The hedge ratio is set to the cointegration transmitter s coefficient, dynamically well-balanced for the open s flow deviation from. Execute hedges only when the open s z-score exceeds 2.5, to keep off overfitting to resound.
