CB Arbitrage Desk Refinitiv · LSEG Eikon updated 2026-05-17 08:07
DEMO MODE · Static snapshot. The live version (running locally) pulls fresh Refinitiv data every weekday at 08:04. View code on GitHub.

Today's plays

Ranked by cheap-to-model · click any row to drill into the bond · new here? read the primer ↓
1
Rohm NR Tree
The thesis: the bond trades at 128.38, but our model says fair value is 157.04 — a +22.3% gap. Buy ¥100M face of the bond and short 25,638 shares of 6963.T to neutralize stock direction. Historical hedged win rate for the 15-25% cheap bucket over 60 days: 50% (49% on a pure-bond-return basis, before hedging costs).
Maturity 2029-04-24 · Realized vol (HV60) 64% · Credit spread 429bp
+22.3% cheap
BUY
2
Taiyo Yuden NR Tree
The thesis: the bond trades at 148.38, but our model says fair value is 179.95 — a +21.3% gap. Buy ¥100M face of the bond and short 19,190 shares of 6976.T to neutralize stock direction. Historical hedged win rate for the 15-25% cheap bucket over 60 days: 50% (49% on a pure-bond-return basis, before hedging costs).
Maturity 2030-10-18 · Realized vol (HV60) 73% · Credit spread 429bp
+21.3% cheap
BUY
Snapshotdemo_screen
Bonds priced15
Rated0
MC priced5
Buy signals3
Median spread429bp

Full universe

click column to sort · ⊕ to add to watchlist · row colour by cheap%
How to read this table Every bond the model could price today. Green-tinted rows are cheap to model (potential buys); red-tinted are rich. Hover any column header to see what it means. The Rtg column shows the issuer's credit rating (greener = safer). Engine = whether the bond used the binomial tree (default) or the Monte Carlo pricer (used for bonds with reset clauses). Conf = the system's confidence in the signal — Low means we suspect a data anomaly like a stale conversion price.
Issuer Equity Rtg Spd Engine Maturity Mkt Model Cheap% Prem% HV60 IV Δ Vega Conf
Nxera Pharma 4565.T NR 429 Tree 2028-12-14 81.75 107.39 +31.4 24 0.65 0.031 0.39 Low
Rohm 6963.T NR 429 Tree 2029-04-24 128.38 157.04 +22.3 -6 0.64 0.026 0.59 High
Taiyo Yuden 6976.T NR 429 Tree 2030-10-18 148.38 179.95 +21.3 -4 0.73 0.019 0.63 High
Daifuku 6383.T NR 429 MC 0% 2030-09-13 199.62 207.14 +3.8 -4 0.50 0.026 0.57 High
Infroneer 5076.T NR 429 Tree 2029-03-30 133.38 137.35 +3.0 -1 0.39 0.044 0.47 High
CyberAgent 4751.T NR 429 Tree 2029-11-16 107.25 103.84 -3.2 23 0.36 0.42 0.037 0.57 High
FRANCE BED HLDGS 7840.T NR 429 MC 32% 2029-03-14 105.00 98.24 -6.4 14 0.22 0.36 0.040 0.52 High
Nippon Steel 5401.T NR 429 Tree 2031-02-14 102.00 95.08 -6.8 34 0.30 0.43 0.073 0.44 High
Nippon Steel 5401.T NR 429 Tree 2029-02-14 101.55 93.48 -7.9 32 0.30 0.47 0.049 0.43 High
Kansai Paint 4613.T NR 429 Tree 2029-03-08 103.00 94.59 -8.2 24 0.27 0.44 0.015 0.48 High
Daiwa House 1925.T NR 429 Tree 2029-03-30 102.25 93.04 -9.0 21 0.20 0.37 0.011 0.51 High
AZ-Com Maruwa 9090.T NR 429 MC 99% 2030-09-24 98.75 89.37 -9.5 65 0.27 0.54 0.020 0.26 High
Daiwa House 1925.T NR 429 Tree 2030-03-29 102.50 91.62 -10.6 19 0.20 0.39 0.011 0.57 High
Tokyu 9005.T NR 429 MC 40% 2030-09-30 102.25 90.88 -11.1 25 0.18 0.43 0.025 0.59 High
Kasumigaseki Cap 3498.T NR 429 MC 96% 2029-11-05 102.12 86.93 -14.9 200 0.47 1.00 0.000 0.14 Low

About this dashboard

Autonomous AI workflow that pulls live convertible-bond data from Refinitiv every weekday morning, prices each bond using a Tsiveriotis-Fernandes binomial tree (with Monte Carlo for path-dependent reset clauses), validates against real issuer ratings + JGB yields + credit spreads, and surfaces the day's high-confidence cheap-to-model opportunities. See How it works for the 10-step agent loop, the Backtest for the honest performance numbers (5.65% CAGR, Sharpe 0.49 over 3 years after realistic costs), and the Glossary for every term defined.

What is convertible bond arbitrage?

The trade

Buy a convertible bond (a bond that can be exchanged for the issuer's stock at a fixed conversion price), and short delta × shares of the underlying equity. The bond's price has both a debt component and an embedded equity option.

Why mispricings happen

Convertible bonds in Japan are dealer-quoted and illiquid. Dealer indicative quotes lag the underlying equity, so when the stock moves the implied option value gets stale — creating fair-vs-market gaps that the model surfaces.

How the model decides "cheap"

For each bond we compute model fair value = bond floor + embedded option value, using realised vol of the underlying, the issuer's credit spread, and the JGB yield curve. Cheap% = (model − market) / market.

Action labels

BUY = ≥5% cheap, high confidence, no data flags. WATCH = cheap but reset clause or stale-quote risk. AVOID = rich vs model. Always Δ-hedge with the underlying short.