VLCC Convexity Hedging: DHT / FRO
Does deep-OTM tail-hedging work on a large VLCC position — and how much can you overpay?
July 20, 2026 — Applied convexity study (backtested)
Why this exists: extends the tail-hedging / convexity backtest from the S&P to VLCC equities (DHT, FRO) — the assets a large tanker holder actually owns. Three questions: (1) does buying deep-OTM puts protect a long VLCC position? (2) how reliable is that protection across scenarios (fast crash vs slow cyclical grind, strike depth, tenor)? (3) how does the answer depend on the vol-risk-premium (VRP = how overpriced the puts are) — i.e., how do you compute the “win-rate” of the hedge as a function of VRP? Code:
run_backtest_vlcc.py; data:data/.
TL;DR verdict
- VLCC is a different animal. DHT annualized vol 48%, FRO 61% (vs S&P 17%), with −97% / −98% drawdowns. Tanker downturns are slow multi-year grinds — the exact regime that defeats short-dated puts (report §7.5).
- The core new metric — CAGR break-even VRP. For DHT, tail-hedging with a 1-year 30%-OTM rolled put raises CAGR as long as you pay a VRP below ≈ 67% (i.e., IV up to ~1.67× realized). For the S&P that break-even VRP is ≈ 0%. The fatter the tail, the more you can overpay for insurance.
- But it is asset- and regime-specific — not universally reliable. FRO’s break-even VRP is ≈ 0% (its 61% vol makes premiums brutal, and the 2011–12 restructuring + choppy grind defeat the rolled put). And DHT’s entire positive contribution came from essentially ONE year (2011, −83%) — extreme lumpiness.
- Win-rate is the wrong headline. The hedge “wins” only 16% (DHT) / 4% (S&P) of one-year periods — you are “wrong” most of the time by design. Use expectancy, the CAGR break-even VRP, and the insurance-value band, conditioned on entry vol (buying puts after vol spikes is a near-guaranteed loss).
- Robustness + live pricing → it’s really a cycle bet (§3.6, §7). DHT’s 67% break-even VRP is entirely the 2008–12 crash: exclude it (2013+, 2019+ windows, +14–25%/yr) and the break-even collapses to 0%. Live options (Jul 2026) price DHT’s 1-yr 30%-OTM put at a paid VRP of ≈33% (FRO ≈26–31%). So hedging DHT is worth it only if you believe a 2008-scale downturn is ahead (33% < 67%) — i.e., hedge near a cyclical top (CRule 1 + CRule 5), or don’t hedge at all.
Education/analysis, not investment advice.
⚠️ Protocol Notice & Caveats
Applies the repo’s Two-Step Research Protocol and connects to CRule 5 (buy protection when vol is low / everyone is greedy) and CRule 8 (pre-committed convex exits). §1 Step-1 draft; §2 Step-2 review; §3 results; §4 the win-rate-vs-VRP framework; §5 VLCC-specific VRP thinking; §6 practical takeaways.
Data: DHT (2005-10 → 2024-12) and FRO (2005-01 → 2024-12) daily adjusted close (splits + dividends), yfinance. S&P from the sibling study for comparison. Puts priced by Black-Scholes, IV = trailing 63-day realized vol × (1 + VRP); PASSIVE rolled put (the design that dominated §7.5).
Read these caveats before trusting any number:
- Window bias: 2005 start is near a cyclical peak; the window includes the worst tanker depression in history (2009–2018). Buy-and-hold DHT/FRO lost money over this window (that is why it is a good hedge stress-test — not a claim the stocks are bad). A holder who bought in 2020–22 has a totally different experience.
- FRO 2011–12 restructuring is a near-total wipeout / credit-event embedded in the continuous ticker — treat FRO pre-2013 with care; DHT is the cleaner case.
- VLCC options are illiquid (thin, wide bid-ask) → the real VRP you pay is high and uncertain; frictions eat into the break-even cushion.
- Options modeled on the adjusted (total-return) series (real puts are on price — minor distortion). Gross of tax / transaction cost.
Section 1 — Step 1: Concise Research Draft
Core conclusion (first): For a concentrated long VLCC position, deep-OTM long-dated put hedging can raise the geometric return and is worth a much higher vol-risk-premium than for an index — because the tail (−97%) is so catastrophic that avoiding it justifies a large EV-negative premium. But the protection is neither cheap nor reliable across assets/regimes: it works for DHT (break-even VRP ≈ 67%), fails for FRO (≈ 0%), is dominated by a single crash year, and is destroyed if you initiate after volatility has already spiked. The right decision variable is not win-rate but the CAGR break-even VRP vs the VRP you actually pay.
3 supporting points (claim → evidence needed):
- Claim: VLCC’s fat tail raises the tolerable VRP far above an index’s. → Evidence: DHT CAGR break-even VRP ≈ 67% vs S&P ≈ 0% (breakeven_vrp); DHT mean 1-yr put payoff 4.9% vs S&P 0.4%.
- Claim: Long-dated beats short-dated on these slow grinds. → Evidence: DHT 252-day puts beat 63-day at every strike/VRP (hedge_grid).
- Claim: Cheap protection materially cuts the VLCC drawdown. → Evidence: DHT maxDD −97% → −78% (k20% 1y VRP0) while lifting CAGR (−6.2% → +0.7%).
2 opposing / counter points (claim → evidence needed):
- Claim: It is not reliable — asset/regime/luck dependent. → Evidence: FRO break-even VRP ≈ 0% (hedge is a net drag); DHT’s positive value came almost entirely from 2011 (+49% hedge P&L in a −83% year; most other years −5% to −15%) (reliability).
- Claim: Entry timing (vol regime) can flip the sign. → Evidence: buying 1-yr puts in high-IV years (2009, 2015) lost 37–40% even as the stock fell — the “re-buy at peak IV” trap, amplified by 48–61% vol.
Explicitly unknown (not fabricated): the real VRP a VLCC holder pays (options illiquid — likely high, unmeasured here); whether the next tanker downturn is a fast crash or a slow grind; results net of the wide bid-ask on DHT/FRO options; how much of FRO’s failure is the 2011–12 restructuring artifact.
Section 2 — Step 2: Strict Peer Review (draft NOT rewritten)
1. Facts that need verification
- DHT/FRO adjusted series continuity through splits & the FRO 2011–12 restructuring (yfinance adjustment quality); verify vs company filings.
- Break-even VRP (67% DHT) is window-dependent (2005 near-peak start); verify on sub-windows (e.g., 2013+, 2019+).
- BS with IV = realized×(1+VRP) is a proxy; VLCC option skew is steep and IV is often > this — verify against real DHT/FRO option chains if obtainable.
2. Logical leaps / equivocation
- “Break-even VRP 67%” ⇄ “hedging DHT is safe.” The 67% is a historical average dominated by one year; it is not a guarantee for the next cycle.
- “VLCC tail is fat, so hedge” conflates DHT and FRO — they diverge sharply; do not generalize “VLCC.”
- CAGR-positive ≠ EV-positive — the hedge is EV-negative above ~0% VRP; its value is purely the geometric/ergodicity benefit. Don’t sell it as “free.”
3. Missing counterexamples / competing explanations
- Just trim the position (AQR de-risk): for an illiquid-option asset, holding less DHT/more cash may beat paying wide option spreads — untested here.
- Dividends as a natural hedge: DHT/FRO pay 20–40% yields at cycle peaks; the income cushion (already in the adjusted series) competes with paying put premium.
- Freight derivatives (FFA) hedge the rate (the actual earnings driver) more directly and liquidly than equity puts — a competing convex hedge not tested.
4. Most important primary sources to add
- DHT/FRO option chains (real IV/skew) to measure the actual VRP paid.
- Clarksons/Baltic TD3C to align the equity hedge with the underlying freight cycle.
- Company filings on the FRO 2011–12 restructuring and DHT/FRO split history.
5. Sentences that are at most speculation, not fact
- “the more you can overpay for insurance” as a forward statement (it is a backward-looking average).
- “works for DHT, fails for FRO” as a durable property (both are one 20-yr path).
- any implied prediction about the next downturn’s speed.
Section 3 — Results
Full tables in data/. All returns from adjusted (total-return) series.
3.1 Risk profile — VLCC vs S&P → results_vlcc_profile.csv
| Asset | Ann. vol | maxDD | Worst day | Best day | Unhedged CAGR (window) |
|---|---|---|---|---|---|
| DHT | 48% | −97% | −26% | +27% | −6.2%/yr (2005–24) |
| FRO | 61% | −98% | −41% | +38% | −4.7%/yr (2005–24) |
| S&P 500 | 17% | −57% | −21% | +12% | +8.4%/yr (1974–24) |
VLCC vol is ~3× the index and drawdowns near-total. Over this full-cycle window (starting near a peak) both names lost money — the reason a tail hedge could add so much, and the reason to read every number as window-conditional.
3.2 Reliability grid — strike × tenor × VRP (DHT) → results_vlcc_hedge_grid.csv
| DHT hedge | CAGR | maxDD | FRO same | CAGR | |
|---|---|---|---|---|---|
| UNHEDGED | −6.2% | −97% | UNHEDGED | −4.7% | |
| k20% 1yr VRP0 | +0.7% | −78% | k20% 1yr VRP0 | −6.2% | |
| k30% 1yr VRP0 | +0.7% | −83% | k30% 1yr VRP0 | −6.0% | |
| k30% 1yr VRP50% | −4.3% | −90% | k30% 1yr VRP50% | −14.2% | |
| k30% 3mo VRP0 | −2.1% | −91% | k30% 3mo VRP0 | −1.6% |
Two lessons: (i) 1-year beats 3-month for DHT (slow grind); (ii) the same hedge that helps DHT hurts FRO — reliability is asset-specific, not a property of “VLCC.”
3.3 Win-rate vs VRP (canonical: 1-yr put, DHT/FRO 30%-OTM, S&P 20%-OTM) → results_vlcc_winrate_vrp.csv
| Asset | VRP | Win-rate | Expectancy | Payoff ratio | CAGR delta |
|---|---|---|---|---|---|
| DHT | 0% | 16% | +0.1% | 5.4 | +6.83pp |
| DHT | 50% | 12% | −5.4% | 2.5 | +1.83pp |
| DHT | 100% | 9% | −11.2% | 1.8 | −3.63pp |
| FRO | 0% | 18% | −2.5% | 2.8 | −1.30pp |
| FRO | 50% | 13% | −9.9% | 1.5 | −9.50pp |
| S&P | 0% | 4% | −0.4% | 9.9 | −0.42pp |
| S&P | 50% | 4% | −1.9% | 3.5 | −2.40pp |
Win-rate is 4–18% everywhere — you lose premium most years. Yet DHT’s CAGR *rises +6.8pp at VRP0: a low-win-rate, high-payoff-ratio (5.4) convex bet can be strongly geometrically positive. This is exactly why win-rate alone is the wrong metric.*
3.4 The headline framework — break-even VRP → results_vlcc_breakeven_vrp.csv
| Asset | Mean 1-yr put payoff | Expectancy break-even VRP | CAGR break-even VRP |
|---|---|---|---|
| DHT (30%-OTM) | 4.9% | ~0% | ≈ 67% |
| FRO (30%-OTM) | 5.4% | ~0% | ≈ 0% |
| S&P (20%-OTM) | 0.4% | ~0% | ≈ 0% |
- Expectancy break-even VRP = where mean payoff = mean premium. ~0% for all → at realized vol the puts are roughly fairly priced to their average payoff; any overpricing makes the raw bet EV-negative.
- CAGR break-even VRP = where hedged CAGR = unhedged. DHT ≈ 67%, FRO ≈ 0%, S&P ≈ 0%.
- The gap [0%, 67%] for DHT is the “insurance-value band” — the range of VRP where the hedge is EV-negative but still raises CAGR (pure ergodicity/variance-drain value from dodging the −97% tail). For FRO and the S&P that band is ≈ zero.
3.5 Reliability / lumpiness — annual hedge P&L (30%-OTM 1-yr, VRP 50%) → results_vlcc_reliability.csv
| DHT year | Underlying | Hedge P&L | Note | |
|---|---|---|---|---|
| 2011 | −83% | +49.4% | the one year that carried the whole hedge | |
| 2009 | −31% | −40.8% | bought puts at post-2008 peak IV → paid ~41%, stock only −31% | |
| 2008 | −48% | +1.0% | barely paid (slow grind, not a fast crash) | |
| most other years | mixed | −5% to −15% | steady premium bleed |
Two brutal, honest lessons: (a) the hedge’s value is lumped into ~one event (2011) — extreme fragility; (b) initiating in a high-IV year (2009, 2015) *lost 37–41% even as the stock fell — the peak-IV trap, amplified by 48–61% vol. When you buy the hedge matters more than that you buy it.*
3.6 Robustness — is the 67% break-even VRP stable? (sub-windows) → results_vlcc_breakeven_windows.csv
| Asset | Window | Years | Unhedged CAGR | CAGR break-even VRP |
|---|---|---|---|---|
| DHT | 2005+ | 19.2 | −6.2% | 67% |
| DHT | 2013+ | 12.0 | +14.3% | 0% |
| DHT | 2019+ | 6.0 | +24.8% | 0% |
| FRO | 2005+ | 20.0 | −4.7% | 0% |
| FRO | 2013+ | 12.0 | +3.8% | 0% |
| FRO | 2019+ | 6.0 | +25.8% | 0% |
The 67% is NOT a stable property — it is entirely the 2008–2012 catastrophe. Exclude that crash (2013+, 2019+) and DHT compounded +14% to +25%/yr, so the rolled put just bled premium → break-even VRP collapses to 0%. The break-even VRP is a function of whether a catastrophic crash falls inside the window — not a durable feature of the stock. Translation: the ~67% cushion is real only if a 2008-scale downturn is actually ahead. This is the ultimate form of the “convex bets are lumpy and regime-dependent” theme — and it turns the hedge decision into a cycle-position call (§7).
Section 4 — How to Compute the “Win-Rate” vs VRP (the framework)
Raw win-rate (4–18%) is useless as a standalone for a convex hedge — you are meant to lose small premiums most of the time and win big rarely. Here is the decision procedure that actually works, as a function of VRP:
Step 1 — Estimate the VRP you actually pay.
paid_VRP = (market option IV / trailing realized vol) − 1. For thin VLCC options this is high and uncertain; use a conservative (high) estimate.
Step 2 — Compute the two break-even VRPs (from history or a model).
expectancy(VRP) = mean(payoff) − mean(premium(VRP))→ expectancy break-even VRP (raw +EV boundary).ΔCAGR(VRP) = CAGR_hedged − CAGR_unhedged(from a backtest) → CAGR break-even VRP (the one that matters for a compounder).
Step 3 — Decide by the CAGR break-even, not the win-rate.
Hedge iff
paid_VRP < CAGR-break-even-VRP(with a margin for frictions). DHT ≈ 67% → lots of room; FRO ≈ 0% and S&P ≈ 0% → only hedge if you can buy at/below realized vol (rare).
Step 4 — Report win-rate correctly (three numbers, not one):
- Unconditional win-rate (low — ignore as a decision input).
-
Regime-conditional win-rate: P(hedge pays annual decline > X%) — high; this is the number that describes the insurance. - Magnitude-weighted expectancy = win-rate × avg-win − loss-rate × avg-loss, plus the payoff ratio (DHT 5.4 at VRP0). A 16%-win / 5.4-payoff-ratio bet is ≈ EV-neutral and geometrically positive — which the CAGR delta confirms (+6.8pp).
Step 5 — Condition on entry vol. Compute all of the above only for initiations when vol is low (cycle top). §3.5 shows initiating at high IV (mid-crash) turns the hedge into a guaranteed loss. The single biggest lever is buying the hedge cheap, at the top of the cycle (CRule 5).
Section 5 — VLCC-Specific VRP Thinking
- Fatter tail → higher tolerable VRP. The whole reason DHT’s CAGR-break-even VRP (67%) dwarfs the S&P’s (~0%) is that DHT can fall −97% while the S&P tops out near −57%. Dodging a −97% drawdown is worth an enormous EV-negative premium (a −90% loss needs +900% to recover; the variance/skew drain is colossal). For catastrophic-tail cyclicals, tail insurance is structurally more valuable than for an index.
- …but the grind and the premium fight back. VLCC downturns are multi-year grinds, so you must re-buy and re-pay premium through the whole decline (the FRO failure). And 48–61% vol makes even 30%-OTM puts expensive. Net: the hedge only wins if the eventual drawdown is deep enough to overcome years of premium — true for DHT’s −97%, not reliably for FRO.
- Entry-vol is destiny. Because IV explodes once the cycle turns, the hedge must be established at the top, when vol is low and complacency high — precisely CRule 5’s “buy protection when everyone is greedy.” Buying after the first −30% is a near-guaranteed loss (§3.5).
- Dividends are a partial natural hedge. DHT/FRO pay 20–40% yields near peaks; that income cushions the position and competes with paying put premium — one reason a dividend-harvesting holder may rationally under-hedge.
- Options are thin — consider substitutes. Real DHT/FRO option spreads are wide, so the paid VRP is high and may exceed even DHT’s 67% cushion. Practical convex alternatives: trim the position at the cycle top (de-risk), hold a cash buffer, or hedge the freight rate via FFAs (which track the actual earnings driver more liquidly than equity puts).
Section 6 — Practical Takeaways for a Large VLCC Holder
- Tail-hedging a concentrated VLCC book can be worth a lot — DHT’s history says you could overpay up to ~67% VRP and still raise CAGR — but only with long-dated (≥1yr) deep-OTM puts, bought at the cycle top when vol is low.
- Do not trust it as reliable insurance. It is asset-specific (DHT yes, FRO no), lumpy (one event carried DHT), and lethal if initiated mid-crash at peak IV. Size it as a small convex sleeve, not a portfolio pillar.
- Compute the decision correctly: compare your paid VRP to the CAGR break-even VRP, not the win-rate. If VLCC option spreads push your paid VRP above the break-even, de-risk (trim) or use FFAs instead.
- This is CRule 5 + CRule 8 in options form: buy cheap protection when the cycle is euphoric and vol is low; pre-commit the exit; and never chase insurance after the crash has started.
Bottom line: the convexity logic that was marginal for the S&P is materially stronger for a catastrophic-tail cyclical like DHT — the fat tail buys you a wide VRP cushion (~67%). But VLCC’s slow grinds, high premiums, thin options, and one-event lumpiness make it fragile and timing-dependent: worth doing small, long-dated, and bought at the top — or replaced by simple de-risking when the options are too dear.
Section 7 — Live Calibration: Your Paid VRP vs the Break-Even (it’s a cycle bet)
What you actually pay today (live DHT/FRO option chains, ~Jul 2026) → results_vlcc_paid_vrp.csv:
| Ticker | Expiry (T) | Spot | ~30%-OTM put IV | Realized vol (63d) | Paid VRP | Open int. |
|---|---|---|---|---|---|---|
| DHT | 2027-07 (1.0y) | $17.89 | 55% | 42% | ≈ +33% | 589 |
| FRO | 2028-01 (1.5y) | $36.93 | 59% | 47% | ≈ +26% | 23 (thin) |
| FRO | 2027-02 (0.6y) | $36.93 | 62% | 47% | ≈ +31% | 349 |
Paid VRP = option IV / trailing-realized vol − 1. Snapshot; VLCC options are illiquid (note FRO’s OI of 23) so treat as indicative.
The decision, made concrete:
- DHT paid VRP ≈ 33%. That is below the full-cycle break-even (67%) → hedging is worth it only if a 2008-scale downturn is plausibly ahead. It is far above the recent-regime break-even (0%) → in a continuing super-cycle the same put just bleeds.
- FRO paid VRP ≈ 26–31%, but its break-even VRP is ≈ 0% in every window → do not hedge FRO with these puts — de-risk (trim) or hedge the rate via FFAs.
So the hedge decision collapses to a CYCLE-POSITION call. The ~67% cushion only pays if the catastrophe that created it recurs; §3.6 shows that outside the 2008–12 crash the break-even is 0%. Therefore:
Hedge the VLCC book only near a cyclical TOP — when (a) rates/valuations are stretched, (b) vol is low so puts are cheap (low paid VRP), and (c) a large downturn is plausible. That is exactly CRule 1 (cycle positioning) + CRule 5 (buy protection into greed). Mid-cycle or early-cycle, the hedge is a near-guaranteed drag — de-risk by trimming instead.
This is where the whole tail-hedge study converges with this repo’s core competency: convexity hedging on a cyclical is a bet on cycle position, priced through the VRP. The options market hands you a live paid-VRP (~33% for DHT); your job is to compare it to a break-even VRP that is itself a function of where in the cycle you think you are.
Section 8 — Should You Hedge NOW? Mapping the Current Cycle Read to the Decision
This repo’s live cycle work (VLCC Cycle Position, Jun 2026 — p.37/38; Average × Duration, p.35) puts VLCC at mid-cycle, cheap-to-fair: DHT $17.44 / FRO $35.12 priced on a sustained ~$100k TCE (not the collapsed $420k Hormuz spike), PE 5.2–5.6× @ $100k, supply-backed through 2027 (near-zero orderbook to late-2028), sell-algo “do not sell.”
Cross that with §3.6 / §7: the hedge’s ~67% cushion only pays in a 2008-scale crash, the recent-regime break-even VRP is 0%, and you’d pay a live ~33% VRP on DHT puts today. Mid-cycle + 0% break-even + 33% paid = the hedge would just bleed. The cycle read and the convexity math agree.
Decision matrix — cycle phase (CRule 1) → hedge action:
| Cycle phase | Vol / put price | Hedge action |
|---|---|---|
| Trough / early upturn | vol high, puts dear | Don’t hedge — own the recovery |
| Mid-cycle ← we are here (Jun-2026) | vol moderate, break-even VRP ≈ 0% | Don’t hedge — it bleeds. Ride it; dividends are the cushion; “do not sell” |
| Late-cycle / approaching peak | vol still low, puts cheap | START buying long-dated deep-OTM puts — the CRule 5 window; the 67% cushion goes live |
| Peak / downturn confirmed | vol spiked, puts dear (peak-IV trap) | Too late for cheap insurance — de-risk by TRIMMING |
The trigger to start hedging (watch for the late-cycle flip while vol is still low): rates rolling over from a sustained high (not a spike), the orderbook filling / newbuild surge (CRule 5 sell signal), PE compressing toward its trough on peak earnings, sell-side >70% buys and “super-cycle” headlines. Because supply is backed through 2027, that window is most likely 2027–2028, not now — so pre-position the puts when those signals appear and vol is still cheap, not after the rate has already rolled over (by then IV has spiked → the peak-IV trap of §3.5).
Current verdict (Jul 2026): do NOT tail-hedge the VLCC book yet. You are mid-cycle, the stocks are cheap-to-fair, the break-even VRP here is ~0%, and paying ~33% VRP would bleed. Collect the dividends as your natural cushion, keep the powder dry, and buy the long-dated deep-OTM puts when the cycle signals flip to late-cycle while vol is still low (likely 2027). If you must cut risk before then, trim rather than hedge.
Reproduce it yourself
cd tail_hedge
python run_backtest_vlcc.py # pulls DHT/FRO (yfinance) -> data/results_vlcc_*.csv
python run_backtest_vlcc_windows.py # sub-window break-even VRP + live option paid-VRP
Sources
- yfinance / Yahoo Finance — DHT, FRO daily adjusted close.
- Sibling study: Tail-Hedging & Convexity — 50-Year Backtest (S&P baseline, §7.5 passive-vs-ladder).
- N. N. Taleb (Antifragile), M. Spitznagel (Safe Haven), AQR / Ilmanen (VRP); repo
RULES.mdCRule 5 / CRule 8.
Two-Step Research Protocol applied (§1 draft + §2 review). Bilingual mirror: 中文版 →. Data: data/. Education/analysis only — not investment advice.