Skip to content
QuantCalc Tools
Directory

Bond Duration Convexity Math

50/50 vs Regression-Weighted 2s5s10s Butterfly Curve Trades: DV01 Math

Standard 50/50 DV01-weighted 2s5s10s butterfly trades suffer hidden front-end volatility bias; empirical regression and PCA weighting immunize 5-year belly curvature against 2s10s slope steepeners.

By Elena Rostova, FRM, CQFReview not recorded
50/50 vs Regression-Weighted 2s5s10s Butterfly Curve Trades: DV01 Math — QuantCalc Tools 2026 quantitative & supervisory data card
QuantCalc Tools (Modified Duration, Convexity, DV01 & Key-Rate Hedging): 50/50 equal-DV01 vs OLS slope-regression butterfly sizing, net convexity, and repo drag.Source: QuantCalc Tools Institutional Research DeskLicense: CC-BY-4.0

Relative-value sovereign bond traders express pure curvature views on the 5-year yield curve belly using a 2s5s10s butterfly—going long $100$ million of the 5-year note ("body") against short positions in both the 2-year and 10-year notes ("wings"), quoted in basis points as $Fly{2s5s10s} = 2 y5 - y2 - y{10}$. If you split the short wing DV01 equally (50% / 50%), front-end monetary policy shocks from Secured Overnight Financing Rate (SOFR) repricings leak directly into butterfly PnL because 2-year and 5-year yields exhibit $1.28\times$ higher empirical basis-point volatility than 10-year yields! Anchor partial tenor sensitivities in Key Rate Duration (KRD) & Partial DV01 Yield Curve Bucketing and review foundational duration mechanics in our Modified Duration, Convexity & DV01 master guide.

Exact DV01 System of Equations for 50/50 vs Beta-Weighted Butterflies#

Let $V2, V5, V{10}$ denote dirty market values and $DV012, DV015, DV01{10}$ denote dollar basis-point sensitivities of the 2Y, 5Y, and 10Y legs. Every butterfly enforces net parallel DV01 neutrality: $$w2 \cdot DV012 + DV015 + w{10} \cdot DV01{10} = 0 \implies (-w2\,DV012) + (-w{10}\,DV01{10}) = DV015$$ How we allocate that total short wing DV01 ($DV015$) between the 2Y short leg ($\alpha \cdot DV015$) and the 10Y short leg ($(1 - \alpha) \cdot DV01_5$) distinguishes three institutional desk conventions:

  1. 50/50 Equal-DV01 Weighting ($\alpha = 0.50$): Sets $DV012^{short} = 0.50 \cdot DV015$ and $DV01{10}^{short} = 0.50 \cdot DV015$, tracking $-\Delta(2 y5 - y2 - y_{10})/2$ with zero econometric estimation risk.
  2. Slope-Regression Weighting ($\alpha = \hat{\beta}_{slope}$): Regresses daily yield changes $(y5 - y{10})$ on $(y2 - y{10})$ via Ordinary Least Squares ($\Delta(y5 - y{10}) = c + \beta{reg}\,\Delta(y2 - y{10}) + \epsilont$). Setting the 2Y DV01 share equal to $\alpha = \beta{reg}$ (empirically $0.56\text{ to }0.62$ on US Treasuries) makes butterfly PnL strictly orthogonal ($\text{Cov}(PnL, \Delta(y{10} - y_2)) = 0$) to 2s10s steepeners and flatteners!

Worked $100M 5-Year Belly Butterfly Notional & Carry Calibration#

Formula or model from the articleNot independently verified
Worked $100,000,000 Long-Belly 2s5s10s Treasury Butterfly Sizing Audit
  • Instrument Inputs (Dirty Price = 100.00 Par):
  • 2Y Treasury Wing: YTM = 4.15%, D_mod = 1.890 yrs => $1,890 DV01 per $10M par
  • 5Y Belly Body: YTM = 4.28%, D_mod = 4.420 yrs => $44,200 DV01 on +$100,000,000 Long Body
  • 10Y Treasury Wing: YTM = 4.45%, D_mod = 7.820 yrs => $7,820 DV01 per $10M par
  • 90-Day Rolling OLS Slope Beta: Cov(d(y5-y10), d(y2-y10)) / Var(d(y2-y10)) = 0.5850
  • Case A — Classical 50/50 DV01-Neutral Butterfly (alpha = 0.5000):
  • Target Short 2Y DV01 = -0.5000 * $44,200 = -$22,100/bp => Short $116,931,217 of 2Y Notes
  • Target Short 10Y DV01 = -0.5000 * $44,200 = -$22,100/bp => Short $28,260,870 of 10Y Notes
  • Net Cash Residual = +$100.00M - $116.93M - $28.26M = -$45.19M (Invested in overnight repo!)
  • Case B — OLS Regression Slope-Neutral Butterfly (alpha = 0.5850):
  • Target Short 2Y DV01 = -0.5850 * $44,200 = -$25,857/bp => Short $136,809,524 of 2Y Notes
  • Target Short 10Y DV01 = -0.4150 * $44,200 = -$18,343/bp => Short $23,456,522 of 10Y Notes
  • Slope Shock Stress (2Y drops -30 bps, 5Y drops -17.55 bps, 10Y flat 0 bps => +30bp Bull Steepener):
  • 50/50 Fly PnL: +$44.2k(17.55) - $22.1k(30) - $22.1k*(0) = +$775.7k - $663.0k = +$112,710 LEAK!
  • Regression Fly PnL: +$44.2k(17.55) - $25.857k(30) - $18.343k*(0) = +$775.71k - $775.71k = $0.00 IMMUNIZED!

Notice the dramatic difference in par notional sizing: because the 2-year note has less than a quarter of the 10-year note's Modified Duration ($1.89$ vs $7.82$ years), achieving slope neutrality requires shorting $$136.81\text{M}$ of 2Y notes and only $$23.46\text{M}$ of 10Y notes against $+$100\text{M}$ of the 5Y belly!

Institutional 2s5s10s Treasury butterfly construction pipeline from OLS beta estimation to repo net carry execution
Institutional 2s5s10s Treasury butterfly construction pipeline from OLS beta estimation to repo net carry executionVerified Exhibit

Repo Specialness Traps and Net Convexity Asymmetry in Butterflies#

Never enter a cash Treasury butterfly without checking net convexity AND special collateral repo rates in our SOFR overnight compounding & carry calculator! Unlike a long-wings barbell, a long-belly / short-wings 2s5s10s butterfly is SHORT net convexity ($C{net} < 0$) because the 10-year short wing carries far higher quadratic curvature ($C{10} = 73.4$) per unit of DV01 than the 5-year body ($C5 = 22.8$): $$\text{Dollar Convexity}{\text{reg}} = \$100\text{M}(22.8) - \$136.81\text{M}(4.5) - \$23.46\text{M}(73.4) = -\$57.5\text{M}\cdot\text{yrs}^2 < 0$$ Furthermore, if the on-the-run 2-year or 10-year Treasury trades "on special" in repo at $150\text{ bps}$ below General Collateral SOFR (meaning short-sellers earn a penalized rebate rate when borrowing the bond), financing $$160.27\text{M}$ of short wings bleeds substantial daily carry!

Comparing Institutional Butterfly Weighting Schemes Across Curve Regimes#

Quantitative Distribution · 2Y Short DV01 Share ($\alpha$)
2Y Short DV01 Share ($\alpha$)10Y Short DV01 Share ($1-\alpha$)
01Classical 50/50 DV01-Neutral$50.0\%$ ($-\$2…02Cash- & DV01-Neutral Fly$36.0\%$ ($-\$1…03OLS Slope-Regression Fly$58.5\%$ ($-\$2…04Volatility-Parity Weighted …$44.2\%$ ($-\$1…05Two-Factor PCA (PC1+PC2=0) …$60.2\%$ ($-\$2…
Butterfly Weighting Methodology2Y Short DV01 Share ($\alpha$)10Y Short DV01 Share ($1-\alpha$)Net Cash Balance ($100M Body)Primary Desk Advantage & Vulnerability
Classical 50/50 DV01-Neutral$50.0\%$ ($-\$22.10\text{k}$)$50.0\%$ ($-\$22.10\text{k}$)$+\$45.19\text{M}$ excess cashTracks quoted $2y5 - y2 - y_{10}$; leaks 2s10s slope
Cash- & DV01-Neutral Fly$36.0\%$ ($-\$15.92\text{k}$)$64.0\%$ ($-\$28.28\text{k}$)$0.00 (Zero net cash!)No net repo balance; heavily exposed to 10Y term premium
OLS Slope-Regression Fly$58.5\%$ ($-\$25.86\text{k}$)$41.5\%$ ($-\$18.34\text{k}$)$+\$60.27\text{M}$ excess cashImmunizes 2s10s twists; requires rolling beta updates
Volatility-Parity Weighted Fly$44.2\%$ ($-\$19.54\text{k}$)$55.8\%$ ($-\$24.66\text{k}$)$+\$34.80\text{M}$ excess cashEqualizes standalone leg VaR; leaves residual slope beta
Two-Factor PCA (PC1+PC2=0) Fly$60.2\%$ ($-\$26.61\text{k}$)$39.8\%$ ($-\$17.59\text{k}$)$+\$63.25\text{M}$ excess cashPurest PC3 curvature factor; model-window sensitive
Residual 2s10s yield-curve steepener PnL leakage across Cash-Neutral, 50/50, Volatility-Parity, OLS Regression, and PCA butterflies
Residual 2s10s yield-curve steepener PnL leakage across Cash-Neutral, 50/50, Volatility-Parity, OLS Regression, and PCA butterfliesVerified Exhibit

Live Execution via SOFR OIS Swaps vs Cash Bonds and PCA Governance#

To dodge cash Treasury repo specialness altogether, macro funds regulated by the UK FCA frequently execute 2s5s10s butterflies in cleared SOFR OIS interest rate swaps—eliminating bond borrow squeezes while posting portfolio-margined initial margin under ISDA SIMM v2.6. Compare wing-versus-body trade mechanics with DV01-neutral barbell vs bullet convexity pickup, generalize regression weights to full eigenmodes in PCA level, slope, and curvature DV01 hedging, and browse all 20 guides on the Modified Duration & Key-Rate Hedging Topic Hub.

Frequently asked questions

Why does a 50/50 DV01-neutral 2s5s10s butterfly make or lose money when the 2s10s curve steepens?

During FOMC-driven steepeners or flatteners, the 5-year yield typically moves closer to the 2-year yield (~58.5% of the 2s10s spread change) rather than sitting at the 50% midpoint. A 50/50 split under-hedges the front end by ~8.5% of belly DV01.

Why do cash-neutral 2s5s10s butterflies distort wing DV01 weights so severely?

Forcing w_2 + w_{10} = -100M in market value alongside w_2*D_2 + w_{10}*D_{10} = -100M*D_5 fixes w_2 = -$58.3M and w_{10} = -$41.7M—placing 64% of short DV01 in the 10Y bond and turning the trade into a massive 5s10s steepener!

How does on-the-run Treasury repo specialness erode cash butterfly profitability?

Shorting on-the-run 2Y and 10Y benchmark notes requires borrowing them in the repo market. If either benchmark trades 100+ bps special below General Collateral SOFR, the lost repo interest on $160M of short wings can wipe out 15+ bps of belly carry!

Should institutional traders execute 2s5s10s butterflies in cash Treasuries, futures, or SOFR OIS swaps?

SOFR OIS swaps avoid repo specialness and balance-sheet leverage charges cleanly, whereas cash Treasuries capture bond-specific liquidity premia and Treasury futures (TU/FV/TY) offer CME SPAN 2 cross-margin offsets up to 80%!

Publication record

Published
14 November 2026
Updated
14 November 2026
Author
Elena Rostova, FRM, CQF
Review
Review not recorded
Automated fact check
Passed in the stored pipeline record
Source URLs
3 recorded

AI assistance recorded: anthropic/claude-opus-4.6. This field does not establish human review or source verification.

Academic & Institutional Attribution

Cite This Research & Embed Live Data Card

Open Attribution License (CC BY 4.0)

Standard academic & institutional research paper citation format.

Elena Rostova, FRM, CQF. (2026, November 14). 50/50 vs Regression-Weighted 2s5s10s Butterfly Curve Trades: DV01 Math. QuantCalc Tools. https://quantcalctools.com/guides/fifty-fifty-vs-regression-weighted-2s5s10s-butterfly-curve-trades
Explore Bond Duration Convexity Math Hub (20) →

QuantCalc Tools Data, Benchmarks & Contributor Records

50/50 vs Regression-Weighted 2s5s10s Butterfly DV01 Guide | QuantCalc Tools