Forecasts — MacroBessent Research

S&P 5007,650.50+0.17%Nasdaq26,522.54+0.39%Dow51,682.64-0.18%Nikkei 22565,018.95+1.38%Hang Seng24,750.78+0.60%EuroStoxx 506,236.20-1.37%Dollar (DXY)100.22+0.00%EUR/USD1.15+0.18%USD/JPY156.85+0.46%GBP/USD1.34+0.26%USD/CNY6.70-0.15%US 10Y yield5.00+1.03%US 5Y yield4.86+1.15%Gold4,424.90+0.57%Silver66.56+1.66%Copper6.62+0.43%Platinum1,803.50+0.68%Palladium1,306.20+1.24%WTI100.30-1.58%Brent103.87-0.91%Henry Hub2.91+0.38%Gasoline3.53+0.58%Heating oil5.06-1.10%Wheat714.25-1.75%Corn527.50-0.57%Soybeans1,303.50-1.23%Coffee294.55+1.38%Sugar17.36-0.34%Cocoa5,327.00-7.71%Cotton77.38-1.31%Lithium ETF70.50-0.56%Uranium ETF41.65-2.41%Rare Earths ETF69.06-0.06%Copper Miners ETF87.28+0.77%Bitcoin81,244.98+0.42%Ethereum2,637.29+0.99%Solana111.48-0.99%S&P 5007,650.50+0.17%Nasdaq26,522.54+0.39%Dow51,682.64-0.18%Nikkei 22565,018.95+1.38%Hang Seng24,750.78+0.60%EuroStoxx 506,236.20-1.37%Dollar (DXY)100.22+0.00%EUR/USD1.15+0.18%USD/JPY156.85+0.46%GBP/USD1.34+0.26%USD/CNY6.70-0.15%US 10Y yield5.00+1.03%US 5Y yield4.86+1.15%Gold4,424.90+0.57%Silver66.56+1.66%Copper6.62+0.43%Platinum1,803.50+0.68%Palladium1,306.20+1.24%WTI100.30-1.58%Brent103.87-0.91%Henry Hub2.91+0.38%Gasoline3.53+0.58%Heating oil5.06-1.10%Wheat714.25-1.75%Corn527.50-0.57%Soybeans1,303.50-1.23%Coffee294.55+1.38%Sugar17.36-0.34%Cocoa5,327.00-7.71%Cotton77.38-1.31%Lithium ETF70.50-0.56%Uranium ETF41.65-2.41%Rare Earths ETF69.06-0.06%Copper Miners ETF87.28+0.77%Bitcoin81,244.98+0.42%Ethereum2,637.29+0.99%Solana111.48-0.99%

Win rate Das of 2026-09-19

ReadHistoryData

call	hit rate
our ensemble	4/5
market price	1/1
base rate	2/5

What this rate covers. Every claim this desk has ever scored: 5 in total — 5 resolved macro thresholds and 0 scored 28-day signal claims. 11 macro thresholds have been registered in all, so 6 are still open; 35 signal claims are standing and none has reached its horizon yet. The resolved Kalshi book below is the market's record and is deliberately not pooled into this number. The pre-registered floor for publishing an ensemble probability is 50 resolved bets.

Scored as calls: the published probability's side of 50% against the resolved outcome. The market row is scored on the same rule over the 1 resolved bets that carried a market price, so the two rates describe the same bets. The base rate is the reference-class anchor scored the same way on every row. Brier below is the skill metric; this is the plain-language version of the same record.

Reliability: resolved bets binned by the probability we published, against how often they resolved YES. On the dashed line a 30% call comes true 30% of the time. Dot size and label are n per bin. · research.db forecasts, 5 resolved

Running mean Brier after each resolution (lower is better): ours against the Laplace-smoothed base rate on every bet, and against the market price on the 1 bet that carried one. Bets resolved at the same time are shown in ledger order (ties ordered by id). · research.db forecasts, 5 resolved

bin	mean p	realised	n
0.2-0.4	0.29	0.00	3
0.6-0.8	0.70	0.50	2

Mean Brier over 5 resolved bets: ours 0.1668, Laplace base rate 0.3074, market 0.2450 (n=1). With n this small the chart is a scoreboard, not a calibration claim.

Hit rate after each resolution. The lifetime line covers every resolved bet; the trailing-20 line is not drawn, because it needs 20 resolutions and the ledger has 5 — 15 more. · research.db forecasts, 5 resolved

16 stored builds since 2026-09-04 — every numeric leaf of this panel is written to signal_history at each build, so a repeated value can mean the upstream had not published, not that nothing moved.

open › 0 › base rate

2026-09-04 0.60 → 2026-09-19 0.44

open › 1 › base rate

2026-09-04 0.39 → 2026-09-19 0.64

open › 2 › base rate

2026-09-04 0.48 → 2026-09-19 0.14

open › 3 › base rate

2026-09-04 0.51 → 2026-09-19 0.75

open › 4 › base rate

2026-09-04 0.27 → 2026-09-19 0.67

recently resolved › 0 › base rate

2026-09-12 0.60 → 2026-09-19 0.60

recently resolved › 1 › base rate

2026-09-12 0.39 → 2026-09-19 0.39

recently resolved › 2 › base rate

2026-09-12 0.48 → 2026-09-19 0.48

hit rate history · 5 rows

n	resolved at	cumulative	trailing
1	2026-09-12	0.00	not reported
2	2026-09-12	0.50	not reported
3	2026-09-12	0.67	not reported
4	2026-09-12	0.75	not reported
5	2026-09-12	0.80	not reported

scalars

field	value
n resolved	5
our hits	4
our rate	0.80
market rows	1
market hits	1
market rate	1.00
base rate hits	2
base rate hit rate	0.40

Open bets (6) Das of 2026-09-19

ReadHistoryData

bet	p	base rate	outside/inside/refuter	mkt	deadline
Will WTI front-month crude fall to at most 97.62 USD/bbl by 2026-09-20?	held	44%	—	77%	2026-09-20
Will Gold (LBMA PM fix) rise to at least 4,408.18 USD/oz by 2026-09-20?	held	64%	—	50%	2026-09-20
Will the VIX index fall to at most 13.84 points by 2026-09-20?	held	14%	—	no reading	2026-09-20
Will the 10-year Treasury yield rise to at least 4.985 % by 2026-09-20?	held	75%	—	no reading	2026-09-20
Will Bitcoin (BTC-USD) fall to at most 76,197 USD by 2026-09-20?	held	67%	—	13%	2026-09-20
Will WTI-Brent stay between -13.61 and -1.94 $/bbl through 2026-09-22? model wtibrent-nw-1 puts next-week WTI-Brent in [-13.61, -1.94] on features dated 2026-09-11; hit rate 15/24; grade T only after resolve	held	80%	—	no reading	2026-09-22

every bet carries a hard machine-checkable threshold; at deadline it resolves YES/NO from the official series — no judgment calls. pre-registered floor not met: 5/50 resolved bets; the ensemble's probabilities are held until the ledger matures.

Ensemble unavailable on 5 of 6 open bets. The three ensemble passes did not run for those bets, so the probability recorded is the base rate and nothing else. This is a note about the build, not about the bets. Separately, the published call is held: pre-registered floor not met: 5/50 resolved bets; the ensemble's probabilities are held until the ledger matures.

16 stored builds since 2026-09-04 — every numeric leaf of this panel is written to signal_history at each build, so a repeated value can mean the upstream had not published, not that nothing moved.

open › 0 › base rate

2026-09-04 0.60 → 2026-09-19 0.44

open › 1 › base rate

2026-09-04 0.39 → 2026-09-19 0.64

open › 2 › base rate

2026-09-04 0.48 → 2026-09-19 0.14

open › 3 › base rate

2026-09-04 0.51 → 2026-09-19 0.75

open › 4 › base rate

2026-09-04 0.27 → 2026-09-19 0.67

recently resolved › 0 › base rate

2026-09-12 0.60 → 2026-09-19 0.60

recently resolved › 1 › base rate

2026-09-12 0.39 → 2026-09-19 0.39

recently resolved › 2 › base rate

2026-09-12 0.48 → 2026-09-19 0.48

gate

key	value
allowed	no
reason	pre-registered floor not met: 5/50 resolved bets; the ensemble's probabilities a

rows · 6 rows

question	deadline	base rate	market price	probability	p outside	p inside	p refuter	gate
Will WTI front-month crude fall to at most 97.62 USD/bbl by 2026-09-20?	2026-09-20	0.44	0.77	not reported	not reported	not reported	not reported	held
Will Gold (LBMA PM fix) rise to at least 4,408.18 USD/oz by 2026-09-20?	2026-09-20	0.64	0.49	not reported	not reported	not reported	not reported	held
Will the VIX index fall to at most 13.84 points by 2026-09-20?	2026-09-20	0.14	not reported	not reported	not reported	not reported	not reported	held
Will the 10-year Treasury yield rise to at least 4.985 % by 2026-09-20?	2026-09-20	0.75	not reported	not reported	not reported	not reported	not reported	held
Will Bitcoin (BTC-USD) fall to at most 76,197 USD by 2026-09-20?	2026-09-20	0.67	0.13	not reported	not reported	not reported	not reported	held
Will WTI-Brent stay between -13.61 and -1.94 $/bbl through 2026-09-22?	2026-09-22	0.80	not reported	not reported	not reported	not reported	not reported	held

scalars

field	value
n open	6
n resolved	5

Ledger A — macro thresholds (5 resolved of 11) Das of 2026-09-19

ReadHistoryData

bet	our p	outcome	brier	resolved
Will Bitcoin (BTC-USD) fall to at most 77,487 USD by 2026-09-11?	70%	YES	0.090	2026-09-12
Will the 10-year Treasury yield fall to at most 4.746 % by 2026-09-11?	30%	NO	0.090	2026-09-12
Will the VIX index fall to at most 13.44 points by 2026-09-11?	22%	NO	0.048	2026-09-12
Will Gold (LBMA PM fix) rise to at least 4,552.05 USD/oz by 2026-09-11?	34%	NO	0.116	2026-09-12
Will WTI front-month crude fall to at most 89.52 USD/bbl by 2026-09-11?	70%	NO	0.490	2026-09-12

mean Brier 0.167 vs base-rate benchmark 0.307 (lower is better; beating the base rate is the bar that matters).

win rate: 4/5 (80%) resolved bets hit (beats the base rate on Brier) — market baseline 1/1 (100%) on the 1 bet that carried a market price. A hit means the stated side of 0.5 matched the outcome; with this few resolutions the rates are a running tally, not a track record.

11 macro threshold bets have been registered in all; 5 have resolved and 6 are open. The short ledger is not a filter on this page — it is the whole ledger, and the first bet was written 2026-09-04. The other two ledgers below are where the rest of the record is.

16 stored builds since 2026-09-04 — every numeric leaf of this panel is written to signal_history at each build, so a repeated value can mean the upstream had not published, not that nothing moved.

open › 0 › base rate

2026-09-04 0.60 → 2026-09-19 0.44

open › 1 › base rate

2026-09-04 0.39 → 2026-09-19 0.64

open › 2 › base rate

2026-09-04 0.48 → 2026-09-19 0.14

open › 3 › base rate

2026-09-04 0.51 → 2026-09-19 0.75

open › 4 › base rate

2026-09-04 0.27 → 2026-09-19 0.67

recently resolved › 0 › base rate

2026-09-12 0.60 → 2026-09-19 0.60

recently resolved › 1 › base rate

2026-09-12 0.39 → 2026-09-19 0.39

recently resolved › 2 › base rate

2026-09-12 0.48 → 2026-09-19 0.48

rows · 5 rows

id	created at	template	question	threshold	operator	baseline	deadline	base rate	market price
5	2026-09-04T15:34:55Z	btc	Will Bitcoin (BTC-USD) fall to at most 77,487 USD by 2026-09-11?	77,487.00	crosses_below	79,379.22	2026-09-11	0.27	not reported
4	2026-09-04T15:34:01Z	ust10y	Will the 10-year Treasury yield fall to at most 4.746 % by 2026-09-11?	4.75	crosses_below	4.77	2026-09-11	0.51	not reported
3	2026-09-04T15:33:10Z	vix	Will the VIX index fall to at most 13.44 points by 2026-09-11?	13.44	crosses_below	14.32	2026-09-11	0.48	not reported
2	2026-09-04T15:32:27Z	gold	Will Gold (LBMA PM fix) rise to at least 4,552.05 USD/oz by 2026-09-11?	4,552.05	crosses_above	4,467.15	2026-09-11	0.39	0.49
1	2026-09-04T15:31:17Z	wti	Will WTI front-month crude fall to at most 89.52 USD/bbl by 2026-09-11?	89.52	crosses_below	90.41	2026-09-11	0.60	not reported

win rate

key	value
our_hits	4
our_n	5
market_hits	1
market_n	1

scalars

field	value
n resolved	5
mean brier	0.17
base rate brier	0.31
registered	11
open	6
first registered	2026-09-04T15:31:17Z
resolved with market price	1

Ledger B — the resolved Kalshi book Das of 2026-09-06

ReadData

This is the market's record, not ours. It is the resolved Kalshi book scored the way we score ourselves, and it is on this page because it is the benchmark our own ledger has to clear — not because the n belongs to us.

resolved markets

3,722,314

trades scored

121,671,888

Brier

0.1685

at trade time

calibration error

0.0072

expected, ECE

Traded price against the realised win rate of everything that traded at that price, over the resolved book. · TrevorJS/kalshi-trades · available_at 2026-09-06

Bubble area is the number of trades in the bucket. The expected-value asymmetry a reader can see for themselves: the cheap buckets carry most of the volume and realise below their price. Source: TrevorJS/kalshi-trades, as of 2026-09-06.

calibration · 20 rows

price	win rate	n
2.70	0.02	15,527,762
8.00	0.07	11,196,789
13.00	0.12	10,382,666
18.00	0.18	10,281,017
23.00	0.24	10,585,305
28.10	0.28	11,055,486
33.00	0.35	11,745,784
38.00	0.38	12,827,019
43.10	0.43	13,501,720
48.00	0.48	16,287,905
53.00	0.53	15,854,010
58.00	0.59	13,242,620
63.00	0.63	12,601,610
68.00	0.66	11,630,980
73.00	0.73	10,914,749
78.00	0.77	10,537,856
83.00	0.84	10,193,288
88.10	0.89	10,559,501
93.10	0.94	11,384,540
97.80	0.98	13,033,081

scalars

field	value
name	Kalshi resolved book
whose	the market's record, not ours
resolved markets	3,722,314
trades	121,671,888
brier	0.17
ece	0.01
log loss	0.50
computed at	2026-09-06T06:31:54.179864+00:00

Ledger C — 28-day signal snapshots Das of 2026-09-19

ReadHistoryData

signal	states	hits	pending
petrofx	29	no reading	0
regime	9	no reading	9
regime_model	16	no reading	16
spread_coint	27	no reading	1
vrp	9	no reading	9

separate from the bet engine: terminal signals that make a mean-reversion or persistence claim are snapshotted and scored 28 days later — no exceptions, no deletions.

recorded

signal states, all cohorts

scoreable

states that make a claim

scored

claims past their horizon

cohorts

weekly snapshots since 2026-09-03

Nothing in this ledger has been scored yet and that is the schedule, not a gap: the first cohort was recorded 2026-09-03 on a 28-day horizon, so the first claim is due 2026-10-01. Until then the honest reading of this table is the count of claims standing, not a hit rate.

16 stored builds since 2026-09-04 — every numeric leaf of this panel is written to signal_history at each build, so a repeated value can mean the upstream had not published, not that nothing moved.

horizon days

2026-09-04 28.00 → 2026-09-19 28.00

rows · 5 rows

signal	n	pending
petrofx	29	0
regime	9	9
regime_model	16	16
spread_coint	27	1
vrp	9	9

scalars

field	value
recorded	90
with claim	35
scored	0
hits	0
cohorts	9
first recorded	2026-09-03
horizon days	28
first due	2026-10-01

In-sample against the lockbox Das of 2026-09-19

ReadData

window	Sharpe	Brier	what it is
training	-3.04	the research loop scores a return series, so neither row has a Brier	the data the optimiser was allowed to fit
lockbox	-5.12	the research loop scores a return series, so neither row has a Brier	2024-01-01 to 2025-03-05, never shown to the optimiser

Both rows are negative, which is the point of publishing them: the rule was rejected, and a research loop that has never rejected anything is not a research loop. 1 trial in the ledger. Source: evolution/vrp_target/{trials.jsonl,lockbox_result.json}.

scalars

field	value
n trials	1
training sharpe	-3.04
training why	not reported
lockbox sharpe	-5.12
lockbox window	2024-01-01..2025-03-05
lockbox dsr	0.00
brier why	the research loop scores a return series, so neither row has a Brier

Evidence grades. Every figure on this site carries a grade assigned by the validation layer, not by the author: D a descriptive fact computed from the record; C a historical distribution with its sample count; N a hypothesis tested and unable to beat chance under the stated control; T a tested-positive claim under the full protocol — none exists as of this build, which is why no forecast appears. Nulls are published in the null-result registry. MacroBessent is a research publication and is not investment advice; nothing here is a recommendation to buy or sell anything. Sources and as-of dates are on every chart.

How the figures on this page are constructed is proprietary. For methodology questions, write to e@finkfi.com.