Re: [Quantlib-users] PiecewiseZeroInflationCurve issues at change of month

[Luigi B. <lui...@gm...>]

Hello Magnus,
    one of our contributors just examined the problem and opened an issue
at <https://github.com/lballabio/QuantLib/issues/2454> — let's continue the
discussion there.

Regards,
    Luigi


On Tue, Feb 24, 2026 at 2:28 PM Magnus Gretton <mag...@ho...>
wrote:

> Hi all,
>
> I am using QuantLib across python and excel (the latter of which has had
> all inflation related classes and any required processes updated to 1.39 as
> a hybrid build) and currently experiencing some issues on building
> inflation curves.
> My issue is that when the months change, for the first few days of each
> month, calculations using the inflation curve fail due to the "root is not
> bracketed"-error.
> For instance, on February 3rd my setup failed although it was fully
> functional at the end of January.
> I've tried to broaden the bands for the guesses without luck in
> inflationtraits.hpp:  *constexpr double maxInflation = 0.5; *
> The error is the same in both my tests in python and excel.
>
> Ive attached my python code, for reference!
>
> *The errors in excel/Python:*
> Excel: "ZeroInflationTSZeroRate - 1st iteration: failed at 3rd alive
> instrument, pillar May 1st, 2026, maturity May 1st, 2026, reference date
> December 1st, 2025: root not bracketed: f[-0.5,0.5] ->
> [-2.184665e-01,-2.960285e-01]"
> Python: "RuntimeError: 1st iteration: failed at 3rd alive instrument,
> pillar May 1st, 2026, maturity May 1st, 2026, reference date December 1st,
> 2025: root not bracketed: f[-0.5,0.5] -> [-2.184665e-01,-2.960285e-01]"
> Note, the same error is produced when valuing swaps using the curve.
>
> *Observations:*
>
>    - During debugging, values seem reasonable across the board. No
>    unexpected anomalies in forecasted index fixings and corresponding rate
>    calculations. (Excel)
>    - When I change the settlement date to the 13th of Feb the
>    calculations work just fine. (Excel)
>    - Removing the first 7 ratehelpers (and reducing noisy short term
>    data) makes calculations work again. (Excel)
>    - Flat interpolation works fine, but using linear as I intend causes
>    issues. (Python / excel)
>    - Reducing curve accuracy has no effect (fails at 1st iteration).
>    (excel)
>
>
> Please find details on my current setup below:
>
> *Inflation setup:*
> *Index:* EUHICPXT
> *Base date: *Dec25 @ 128.89
> *Frequency:* Monthly
> *Curve:* PiecewiseZeroInflationCurve
> *Calendar:* Target
> *DayCount:* Simple
> *Convention:* Modified Following
> *Settlement days:* 0
> *Settlement date:* 03. Feb 2026
> *Lag:* 3m
> *Seasonality:* Omitted for testing purposes / or set to 1 for all months
> *Interpolation: *Linear (Flat does not produce an error interestingly
> enough)
> *Curve Accuracy:* 1.00E-09
> *ZCISratehelpers Tenor & Example Prices:*
>
> Tenor
> Rate
> 3m
> -2.81%
> 4m
> -0.73%
> 5m
> 1.17%
> 6m
> 2.27%
> 7m
> 2.19%
> 8m
> 2.22%
> 9m
> 1.75%
> 10m
> 1.86%
> 11m
> 1.74%
> 1y
> 1.82%
> 15m
> 1.46%
> 18m
> 2.19%
> 21m
> 2.08%
> 2y
> 1.78%
> 3y
> 1.84%
> 4y
> 1.88%
> 5y
> 1.90%
> 6y
> 1.92%
> 7y
> 1.95%
> 8y
> 1.97%
> 9y
> 1.99%
> 10y
> 2.01%
> 12y
> 2.06%
> 15y
> 2.11%
> 20y
> 2.19%
> 25y
> 2.24%
> 30y
> 2.29%
>
> Any help or insights would be greatly appreciated!
> Have a great week, everyone.
>
> Kind regards,
> Magnus
> _______________________________________________
> QuantLib-users mailing list
> Qua...@li...
> https://lists.sourceforge.net/lists/listinfo/quantlib-users
>