Contents

## Capital Structure Arbitrage with a Non-Gaussian Pricing Model Market CDS Rates v

Di Justo, Patrick, Contributing Editor has reference to this Academic Journal, PHwiki organized this Journal Capital Structure Arbitrage with a Non-Gaussian Pricing Model Market CDS Rates vs Our Model When markets differ from model predictions, will they converge How do we profit from convergence Our Theoretical CDS Model: Theoretical CDS Rates via Options market: Stock Default = -95% q-Alpha model to obtain default probabilities numerically differentiate deep OTM puts from the option price surface Bootstrap CDS curve from implied default probabilities

This Particular University is Related to this Particular Journal

Strategy 1: Basic Threshold Strategy If (theoretic market) > then go long $10M notional CDS in addition to short a delta neutral call option hedge. If (theoretic market) < do the opposite Every day, check as long as daily convergence, in addition to take profits if abs(theoretic market) < Stop loss if the trade diverges by In case of stop-loss, then flag the name in addition to dont trade again as long as T time. Our data set: 100 companies over 2 years Strategy 1 (cumulative P/L) (.01,.02,90,.0025) trade trigger level = .01 $ stop loss level = .02 Kick-out period = 90 Convergence level = .0025 days (.02,.05,30,.005) most parameter combinations produced losses Theoretical vs. Market CDS rates Some converge Eastman Kodak Halliburton Market Theoretic CDS spread Days
Theoretical vs. Market CDS rates Some diverge Dow Chemical Sprint Nextel Market Theoretic CDS spread Days Theoretical vs. Market CDS rates Some discrepancies converge in addition to reopen Tyco General Motors Market Theoretic CDS spread Days Theoretical vs. Market CDS rates Some appear to be persistent American Electric Power International Paper Market Theoretic CDS spread Days
Caveats This is a convergence trading strategy Spread may widen further, producing losses Discrepancies may be from: - Model or parameter misspecification - Unperceived systematic risk factors - Inherent liquidity differences - Genuine mispricings NO guarantee that the difference will dissipate over a reasonable horizon Strategy 1 Many parameter combinations produce losses Many discrepancies do not converge We take on all openings & too many bad trades. Stop-loss is the dominating trade Maybe the biggest discrepancies are more likely to have genuine mispricings which converge Strategy 2: Rank in addition to Hold Rebalancing period length = T. At each T, trade the top 10% discrepancies. Take profits daily At the end of T close everything, go back to 1. We only trade egregious differences We capture partial convergence during each holding period
Strategy 2 (cumulative P/L) H = 30 Flat regions mean no trades 10^4 $ H = 60 Days 15 different combinations gave positive P/L Strategy 3: Active Holding Period Interval length = I, Holding period = H In strategy 2, we are idle during the holding periods but here we as long as m new portfolios at every I. At each I, close out the positions from t-H in addition to as long as m a new portfolio. Take profits daily. Strategy 3 (cumulative P/L) (interval, hold) = (15,45) 10^4 $ Days (10,120)
Strategy 3 (cumulative P/L) (50,150) 10^4 $ (40,160) Days For combs tried cum P/L was positive Results seem more Volatile in the interval Length than in H Strategy 4: Capture the Momentum In previous strategies we saw that wide differences may become wider. Use a different ranking criteria: convergence momentum. Similar to strategy 3, but compute in addition to rank the rates of spread convergence during a lookback/ as long as mation period as long as each company Strategy 4 (cumulative P/L) (15,30,60) interval = 15 10^4$ as long as mation = 30 hold = 60 Days (15, 60,90)
Areas as long as Further Analysis Margin effects. Maximum draw-downs effect Sharpe ratios analysis Transaction costs Out-of-sample testing Leverage cycle strategies Check constrained mean, long term time-averaged variance decay. Statistical arbitrage More of an instinct than science Appendix: Default Probabilities Monte Carlo is best, but too slow. Instead: We have as long as mulas as long as option prices under q-alpha dynamics. The option surface implies the S distribution: dP/dK = exp(-rT)Q{ST

## Di Justo, Patrick Contributing Editor

Di Justo, Patrick is from United States and they belong to Wired Magazine and they are from San Francisco, United States got related to this Particular Journal. and Di Justo, Patrick deal with the subjects like Computers

## Journal Ratings by Treasure Valley Community College

This Particular Journal got reviewed and rated by Treasure Valley Community College and short form of this particular Institution is OR and gave this Journal an Excellent Rating.