Reference 09 November 2026 · 6 min

A plain-English guide to daily yield noise: how far the 5- and 10-year move on an ordinary day, and what a 1 basis point hit can and cannot show.

Written Monday November 9, accompanying the November 9 analysis, which graded a weak 3-year auction as one hit on a 5-year close 1.1 basis points above Friday's. Measured on the Yahoo series the plans grade: over the last 250 sessions the 5-year's median close-to-close move was 3.2 basis points and the 10-year's 2.8, about one day in five moved 1.1 or less, and up and down days split 130 to 119, close to a coin flip. A typical 30-minute ^FVX bar moved 0.6 basis points against Monday's 0.1 in the auction window. Why a direction-only grade cannot separate the auction from everything else that happened, why the site still grades small margins without a minimum, and why 10 hits in 19 directional lines is not evidence of skill.

By Igor Komolov

Written Monday November 9, 2026, accompanying the Monday November 9 analysis note, which graded a pre-registered plan as one hit after a weak 3-year auction and a 5-year close 1.1 basis points above Friday's. A yield moves every day whether or not anything happens. Before a close-to-close change can be read as the effect of an auction or a data release, it has to be compared with how much the same yield moves on an ordinary day. This guide measures that for the 5- and 10-year on the same series the plans grade, sets out what a direction-only hit can and cannot show, and explains why the site keeps grading small margins anyway and what it does not claim from them.

What a basis point is, and what the plans grade

A basis point is one hundredth of a percentage point: a 5-year yield going from 4.648 to 4.659 percent is a rise of 1.1 basis points. The plans on this site grade one number per line: the Yahoo daily close of the CBOE yield index (^FVX for the 5-year, ^TNX for the 10-year) against the prior session's close. Direction only. A close above the line is above it by 0.1 basis points or by 20, and the scorecard records both the same way. That is deliberate, because it is the only rule that can be fixed before the day without choosing a size, and it is also why a margin has to be read separately.

How big an ordinary day is

The table is computed from Yahoo's daily closes for ^FVX and ^TNX (v8/finance/chart, one year of daily bars, pulled about 5:15 PM ET Nov 9), as the change from each close to the next, rounded to a tenth of a basis point. The 250-session window runs from November 11, 2025 to November 9, 2026; the 60-session window from August 17; the 20-session window from October 13.

Series, window	Median absolute daily change	Middle half of days	One day in ten above	Days at 1.1 bp or less	Up / down / flat
5Y (^FVX), 250 sessions	3.2 bp	1.6 to 5.1 bp	7.5 bp	47 (19%)	130 / 119 / 1
5Y (^FVX), 60 sessions	3.0 bp	1.5 to 5.4 bp	8.4 bp	11 (18%)	34 / 25 / 1
5Y (^FVX), 20 sessions	3.4 bp	1.6 to 7.0 bp	9.3 bp	3 (15%)	10 / 9 / 1
10Y (^TNX), 250 sessions	2.8 bp	1.4 to 4.8 bp	6.4 bp	46 (18%)	133 / 113 / 4
10Y (^TNX), 60 sessions	3.5 bp	2.2 to 5.2 bp	6.0 bp	10 (17%)	36 / 23 / 1
10Y (^TNX), 20 sessions	3.5 bp	2.2 to 5.8 bp	10.1 bp	4 (20%)	9 / 11 / 0

Three things come out of it. The typical day moves the 5-year about 3 basis points in one direction or the other, and the largest single day in the year moved it 20.4. About one day in five moves 1.1 basis points or less, so Monday's change was smaller than roughly four days in five. And the split of up days and down days is close to even (130 up and 119 down for the 5-year over the year), so a line that only calls a direction has something near a coin flip as its no-skill base rate on any one day. Over the last ten sessions the 5-year's changes were +1.6, +3.1, -7.0, -8.0, -2.8, +5.0, -5.4, -8.1, +4.0 and +1.1 basis points, ending Monday; the 10-year's were +0.7, +5.2, -5.6, -6.0, -3.5, +5.8, -5.8, -5.7, +2.2 and +0.5.

The same is true inside the day. Over the 50 sessions from August 31 to November 9 (700 complete 30-minute bars of ^FVX), the median move from a bar's open to its close was 0.6 basis points, with one bar in ten moving more than 2.1. Monday's auction window, the 12:50 bar open to the 1:20 bar close, moved 0.1 basis points across two bars: less than a typical single half hour. Event windows explains why the site records that window and does not grade it.

What a small hit can and cannot show

A hit says the close landed on the side the plan named. It does not say how the close got there. On Monday the 5-year opened below Friday's close, crossed above it in the 9:20 AM bar and stayed above from 9:50 AM, more than three hours before the auction result existed, while the S&P 500 rose 2.35 percent and Brent fell more than 6 percent. A 1.1 basis point margin on that day is inside the range the series produces without any catalyst, so it cannot separate three stories: the auction lifted the 5-year; something else lifted it and the auction did nothing; or the auction pulled it down slightly and something else lifted it more. All three end at 4.659.

That is why the analysis notes report margins next to grades. The Oct 27 note called a 0.7 basis point hit on the 10-year "real by the plan's terms and small enough to be noise" (the Oct 27 confidence note). Friday's 10-year hit was 2.2, below that series' median day. A grade near the median move is a weak piece of evidence; a grade several times it, with the window showing the move arriving at the release, is a stronger one. Neither is a proof.

Why grade small margins at all

Because the alternatives are worse. A plan could require a minimum margin, say 3 basis points, for a hit. But a miss would then need its own margin too, and the space between them would turn into a second no-call zone that nobody fixed in advance; or the margin could be set after the day, which would let the writer choose it. Grading direction at the close, with the line written down beforehand, removes the choice. The cost is that the record contains thin hits and thin misses, and the honest response is to show their size, not to drop them.

The evidence, such as it is, comes from many plans, not one. The scorecard currently holds 16 plans with 10 hits, 9 misses, 21 no-calls and 2 void lines, and one plan pending. With a near-even base rate, 10 hits in 19 directional lines is about what a coin would produce; this site does not claim skill from it, and it would take far more graded lines before a record like this could say anything. Several plans have noted a session-range line under consideration, which would grade a high and a low as well as a direction; it has not been adopted, and if it is, it will be stated in a plan before it is used and not applied backwards.

A checklist for reading a close-to-close move

Put the move next to the series' median day: about 3 basis points for the 5- and 10-year over the past year.
Check when the move happened. If the yield was already on the graded side before the event, the event did not put it there.
Check the event window against a typical half hour, about 0.6 basis points for the 5-year recently.
List what else happened between the two closes. A plan graded close to close carries all of it.
Treat one hit or one miss as one observation from a near coin-flip base rate.