Nine rose, nine fell, and the bill curve moved from the wrong end

Five complete verified sessions ran from 5 October to 9 October, so the Friday-to-Friday comparison is available without a caveat for a sixth week running. From the 2 October close to the 9 October close the All-Share (DSEI) fell from 4,631.77 to 4,597.48, about −0.74%. The domestic Tanzania Share Index (TSI) fell from 10,255.60 to 10,142.06, about −1.11%.

Banks, Finance & Investment fell from 23,188.19 to 22,842.38, about −1.49%; a second consecutive weekly fall, and −3.68% across the two weeks from 23,714.23 on 25 September. That is the whole of the +3.44% this ledger reported three weeks ago, and then some. Commercial Services rose from 3,507.72 to 3,531.40, about +0.68%, and was the only sector index higher; Industrial & Allied fell from 5,075.25 to 5,034.51, about −0.80%, having been the only one higher a week ago. The spread between the best and worst sector index was about 217 basis points, against roughly 265, 592 and 1,354 in the three weeks before it. Fourth consecutive narrowing.

The sessions ran −0.40%, +0.22%, −0.27%, −0.46%, +0.17%; three of five lower, none of them dramatic. Breadth across those five sessions was 33 advances against 45 declines, against 27 and 48 the week before; slightly better, and still negative.

Then the arithmetic that gives this chapter its name. Across the 28 counters with a verified close on both Fridays, 9 rose, 9 fell and 10 were unchanged. The count was exactly even, and the All-Share fell 0.74% anyway. An index is weighted and a count is not: the two counters that supply most of this market's turnover, CRDB (2,950 → 2,900, −1.69%) and NMB (2,040 → 2,010, −1.47%), were both among the nine that fell. On 8 October those two alone carried 83.8% of the day's equity turnover value. When a board is even and the index is not, the honest reading is that size did it, not sentiment.

The nine that rose did so by more than the nine that fell: TOL +6.36% (1,730 → 1,840), MCB +4.05%, MBP +4.02%, NMG +3.17%, MKCB +2.17%. The nine that fell were led by MUCOBA −14.44% (450 → 385), then PAL −4.92%, AFRIPRISE −4.70% and TCCL −3.80%. The Ghost records the 14.44% and offers no reason for it: the DSE listed-company filing register this book reads was last generated on 23 September and has not been regenerated since, so there is no disclosure evidence to cite either way. A price without a filing is a price, not a story.

And it cost a third less to produce. Equity turnover across the five sessions was TZS 18.019bn, against the previous week's TZS 27.874bn; about −35.4%, and the fourth weekly fall in five weeks. Foreign sales exceeded foreign purchases in all five sessions, TZS 3.131bn net across the week, against TZS 1.358bn net over a prior week that contained two net-buying sessions. No motive is inferred from that; it is the DSE's own participation arithmetic, and it says the week's fall came on less money rather than more.

The model book went from TSh 612.41m on 2 October to TSh 611.22m on 9 October; −TSh 1.19m, about −0.19%. That is the second-largest weekly fall this ledger records, after last week's, and the first time two weekly falls have landed back to back since inception. Contributed capital did not move, so the whole of it is market P&L.

Holding by holding, against the verified closes: NMB 2,040 → 2,010, −1.47%, −TSh 1.676m. CRDB 2,950 → 2,900, −1.69%, −TSh 0.849m. Vodacom 1,260 → 1,270, +0.79%, +TSh 0.833m. Tanzania Cigarette 13,350 → 13,380, +0.22%, +TSh 0.134m. Twiga Cement unchanged at 5,670, nil. Two of five rose, one did not move, two fell; and the two that fell are the two largest bank allocations, which is the same sentence as the index's.

The equity sleeve therefore lost TSh 1.559m. The fixed-income and cash sleeves accrued TSh 0.370m. Those two numbers add to −TSh 1.189m, which is the NAV move exactly; there is no third thing. The ballast covered 23.7% of the equity loss this week against 6.2% last week, and it is worth being precise about why: the accrual was the same TSh 0.370m in both weeks. The ballast did not work harder. The hole was smaller. Income is a constant that arrives whatever equities do; it is not a hedge that scales with the loss, and a week where it covers a quarter of the damage is a week where the damage was small, not a week where the ballast was clever.

The money-weighted return went from +17.89% to +17.64%, about 25 basis points lower. Sleeve weights at the 9 October mark: equities 55.44% against a band of 30 to 60%, fixed income 31.52% against 25 to 50%, cash 9.44% against 5 to 20%, gold 3.60% against 0 to 8%. Nothing is breached, and fixed income is the thinnest thing in the book relative to its band for a fourth month.

The largest model-book equity allocation is NMB at 18.38% of the book, down from 18.61% a week ago and 19.26% the week before that. The single-name cap published on 1 October is 20%, and the distance to it has widened in each of the two weeks since; both times because the price fell, and neither time because a decision was taken.

This chapter will not claim that as evidence the cap works. The cap binds on new savings only, it never forces a sale, and no new savings arrive before 1 November. A rule that has had two quiet weeks has been untested for two weeks. The first occasion on which it can bind on anything at all is the November allocation, and the only honest status until then is: published, not yet exercised.

Bank of Tanzania auction 1208 settled on 7 October, inside this window, and it is the most interesting print the ballast has had in months. The weighted average yields: 35-day 3.36%, 91-day 3.4908%, 182-day 5.63%, 364-day 6.2839%.

Read the two ends separately, because they said opposite things. The 364-day rung fell for a sixth consecutive auction; by one basis point, from 6.29%, against an average step of about 17 basis points across the prior five. The direction that Chapter 21 leaned on survived. The movement it leaned on has all but stopped. At the other end the 35-day rung rose 84 basis points in a single auction, from 2.52%; the second-largest single-auction rise in the 83 auctions this evidence store holds back to January 2023, the largest since November 2023, and the highest level the 35-day has printed since 6 May 2026 (3.39%). The 182-day rose 34. The spread from the 35-day to the 364-day collapsed from 3.77 percentage points to 2.92; about 85 basis points of flattening, delivered entirely from the short end.

That looks, at a glance, like the short end offering to pay. Put inflation next to it and the glance fails. September annual inflation on the Mainland basket was 4.3%, unchanged, in the National Bureau of Statistics release dated 8 October, with transport prices up 14.0% inside that unchanged average. Against 4.3%, the 35-day pays about 0.94 percentage points less than inflation and the 91-day about 0.81 less; the 182-day pays about 1.33 more and the 364-day about 1.98 more. The rung that moved 84 basis points is still a rung that loses to inflation. That is the entire short-end case, in one line, and it is why an eye-catching move is not automatically an opportunity.

No bond auction settled in the window, so the curve store's readings stand as they were: 15 years at 10.79% as at 30 September, and 10 years at 10.69% as at 2 September, still untested by a fresh auction. The policy rate remains 6.25%, now displayed by the Bank of Tanzania against Q4 2026.

And here is the part that decides what the book does with all of it. The Ghost's fixed-income sleeves accrue at the indicative yields they were entered at; T-bonds at 11%, bills at 10%, the collective income fund at 6%, cash at 5%. Six auctions of falling one-year yields have cost this book nothing on what it already owns, and they change everything about what the next shilling buys. A curve moves the hurdle for new money, not the marks on old money; which is precisely why the duration question belongs at an allocation date and not in a week with no money to allocate.

October's TSh 7.0m of savings is still not recorded in the canonical tracker. The public book as at 9 October carries contributed capital of TSh 521.0m; July, August and September only. This is the second consecutive weekly note to disclose it, and Chapter 21 carries its own dated 3 October correction saying the same thing: that chapter's allocation is a published proposal, not a recorded contribution, and TSh 528.0m is the capital the proposal would produce rather than capital the book holds.

What changes when it is entered, stated in advance so nobody has to take it on trust later: contributed capital becomes TSh 528.0m; the market P&L is unchanged; and the money-weighted return moves by roughly two thousandths of a percentage point, because a deposit is capital and not a gain. A book whose reported return barely notices new money is a book measuring return correctly.

The Ghost would rather carry an unreconciled line in public for two weeks than quietly stamp it with an earlier date. Performance is never back-dated here. When the contribution is recorded it will be recorded on the date it is recorded, and the sleeve weights above will change on that date and not before.

Hold. Nothing trimmed, nothing sold, no sleeve opened or closed. Three reasons, in order.

One: there is no new money this week. The next savings arrive on 1 November. Every change available in this chapter would therefore have to be funded by selling something the book already owns, and nothing in the week's evidence asks for a sale. The cheapest way to act badly is to treat an allocation question as a trading question because the calendar is inconvenient.

Two: one auction is a reading, not a trend. This book's own July lesson is that evidence moves an allocation and headlines do not, and the 84-basis-point jump at the 35-day rung is, so far, a single print. It may be the start of a repricing at the front of the curve; it may be one auction's bid-cover arithmetic. A second auction will say which, and the next one lands on the fortnightly cadence after 1208, before the November allocation.

Three: the sleeve that is arguably mispositioned is already the designated recipient. Fixed income at 31.52% sits in the lower half of a 25 to 50% band and has done for a fourth month. The remedy for a thin ballast is to feed it with new savings on 1 November, which is already the standing plan; not to fund it by selling an equity into a week that took TZS 3.131bn of net foreign sales on 35.4% less turnover. Thin markets are where forced sellers find out what their holdings are really worth.

One thing the Ghost explicitly declined to do, published as a non-decision so the reasoning is on the record: it declined to read the 84-basis-point rise at the short end as a reason to send November's money back to the bill ladder. The rung that moved still pays less than inflation. If the short end wants this book's savings, it can start by paying more than 4.3%.

Four dated things. The next Bank of Tanzania bill auction, on the fortnightly cadence after 1208, and specifically whether the 35-day holds 3.36% or gives it straight back, and whether the 364-day finally rises after six falls. A regeneration of the listed-company filing register, last generated 23 September, which would say whether anything was disclosed behind MUCOBA's 14.44% fall, and behind the seventeen days before it. The next bond auction at ten years, still the first real test of 2 September's 10.69%. And the 1 November allocation, which is both the first occasion the single-name cap can bind and the first occasion the short end's new shape can be acted on.

Two smaller marks. The Bank of Tanzania's official USD/TZS indicative mean moved from 2,641.4186 on 2 October to 2,631.3587 on 9 October, about −0.38%; the shilling's second consecutive firmer week, and the larger of the two, with four straight daily readings lower from 6 October. And the Mining Commission's world-gold reference was about flat at $4,192 against $4,191, which ends five consecutive weekly falls; the domestic buying-centre price was $3,689, about 12.0% below world, that gap being royalty, clearing and margin between the global market and the Tanzanian mine gate. The model book's gold sleeve is 3.60%, inside its 0 to 8% band, carried at contributed value and not marked, so none of that movement is in the NAV either way.

The discipline for the week is three lines. When the advance-decline count is even and the index still falls, read the weights and not the count. When a yield curve moves, ask which end moved, and then ask whether that end pays more than inflation before calling it an opportunity. And when a rule drifts further from binding, notice that the price did that, not the rule; a quiet constraint has not been proven, only left alone.

; The Ghost. Hypothetical and educational only; a model book, not real money, and never advice.