WEBVTT

00:00:00.102 --> 00:00:01.661
Can a probability be twenty?

00:00:02.552 --> 00:00:04.665
QuantLib one point forty-three returns

00:00:04.678 --> 00:00:06.701
that result for a deterministic option.

00:00:07.968 --> 00:00:09.798
Forward: one hundred twenty.

00:00:10.689 --> 00:00:12.158
Strike: one hundred.

00:00:13.047 --> 00:00:14.542
Probability should be one.

00:00:15.431 --> 00:00:17.841
Black returns zero; Bachelier returns

00:00:17.854 --> 00:00:18.395
twenty.

00:00:19.642 --> 00:00:21.575
I reproduced both errors in released

00:00:21.588 --> 00:00:23.341
code and current C plus plus.

00:00:24.231 --> 00:00:26.590
The fixes removed ninety mismatches across

00:00:26.603 --> 00:00:28.265
six hundred sixty inputs.

00:00:29.576 --> 00:00:31.753
The tested cap and floor sensitivities

00:00:31.766 --> 00:00:32.694
were also wrong.

00:00:33.585 --> 00:00:35.711
Corrections restored them; prices stayed

00:00:35.724 --> 00:00:36.536
unchanged.

00:00:37.425 --> 00:00:39.603
These synthetic tests do not establish

00:00:39.616 --> 00:00:40.827
customer losses.