20% drawdown · Current risk state
Checking current risk state…
The saved risk light and benchmark are hidden until the opening data check finishes.
Macro Drawdown Radar · U.S. S&P 500
Monthly historical benchmark
Checking current benchmark…
Waiting for the opening data check so an older risk state is not shown as current.
Plain-English guide
How to use this dashboard in 60 seconds
Start with the large traffic light
Green below 55 means limited broad confirmation. Amber from 55–69 means vulnerability is building. Red at 70+ means historical conditions were materially more dangerous.
Read the percentage as a historical frequency
The 14.9% benchmark means 14 of 94 Amber months in the 2000–2022 monthly validation had a 20% forward drawdown within 12 months. It is not a promise, a countdown or a newly calibrated daily forecast.
Demand confirmation from different families
One extreme valuation reading is vulnerability, not timing. Risk becomes more convincing when credit and market momentum deteriorate together and housing or labor also weakens.
Use the daily detector for activation
Below 70 is Low, 70–89 is Elevated, and 90+ is High. A High market-stress reading means fast stress is occurring; its history is short, so confirm it with price, volatility and credit.
Green + Low stressObserve
No broad warning. Continue normal monitoring.
Amber + Low stressPrepare
Review exposures and scenarios, but do not call a crash from vulnerability alone.
Cycle Red at 70+Escalate review
24 of 56 historical cycle-Red months had a 20% forward drawdown: 42.9%. This is not a separately tested credit-confirmed subset.
High market stress 90+Verify immediately
Check whether the market, VIX and spreads confirm a fast-moving event.
Three rules to rememberWatch changes, not just levels · Never count two credit measures as two independent confirmations · Use the false-alarm panel before treating Red as a hedge command.
Decision dashboard
Eight signals, shown without hiding divergence
All readings are direction-adjusted: higher is always worse. Individual inputs are percentiles; combined readings are scores. Credit inputs remain one statistical family.
Elevated risk98.4
Valuation
Buffett market-cap-to-GDP proxy
Checking for updates · saved observation · Market cap: Q1 2026 · quarterly · GDP: Q2 2026 · quarterlyBelow watch26.8
Credit composite
Median of spread level, widening, bank lending standards and financial conditions. Two spread inputs are shown below.
Baa spread proxy level23.6
63-calendar-day HY widening46.2
Below watch48.4
Yield curve & policy
10-year Treasury minus effective Fed Funds; this card displays the spread-level percentile
Checking for updates · saved observation · Observation: 1 Sept 2026Below watch42.8
Housing
Building-permit level compared with its historical range
Checking for updates · saved observation · Period: July 2026 · monthlyBelow watch31.8
Labor
Initial unemployment claims and their 13-week change
Checking for updates · saved observation · Week ending 22 Aug 2026Watch58.8
Real-money liquidity
Real M2 growth after adjusting nominal M2 for consumer prices
Checking for updates · saved observation · M2: July 2026 · monthly · CPI: July 2026 · monthlyWatch56.9
Market momentum
S&P 500 price change over 63 calendar days; the broader Market pillar also includes trend and volatility
Checking for updates · saved observation · Latest close: 2 Sept 2026Low45.8
Market stress
Five/ten-calendar-day downside speed, VIX, volatility acceleration and a Baa spread proxy
Checking for updates · saved observation · SPX: 2 Sept 2026 · VIX: 1 Sept 2026Confirmation breadth · 2 of 4 at Watch or higher
Four Risk Pillars
Valuation: Red · Macro: below Watch · Credit: below Watch · Market: Watch
Below 5555–6970+
ValuationRED
98.4MacroGREEN
46.5CreditGREEN
26.8MarketAMBER
56.920% backtest results Expand historical results and limitations
Backtest results
What the refined radar caught
7 / 7monthly 20% episodes reached Red
10 momedian warning lead in full history
42.9%12-month event rate in Red months
0.8%12-month event rate in Green months
2 / 2daily 20% events caught since 2017
| Previously missed episode | First warning | Lead to 20% breach | Signals that added coverage |
|---|---|---|---|
| 1968–70 | May 1969 | 12 months | Curve inversion, restrictive real Fed Funds, real-M2 contraction and Baa spreads |
| 2007–08 | Jan 2008 | 8 months | Credit widening, falling permits, curve re-steepening, claims and market trend |
| 2020 | 31 Jan 2020 | 41 days | Volatility and VIX acceleration, 5/10-day downside speed and Baa widening |
How to interpret the improvement
The long-cycle additions were chosen after reviewing the earlier misses, so 7/7 is a retrospective hypothesis check—not an untouched accuracy claim. The compact probability model remains frozen; the enriched rules now need a new point-in-time validation.
20% false-alarm audit Expand independent, repeated and post-crash warnings
False-alarm audit
Two independent false-alarm regimes—not 32 independent bad calls
Eight Red months clustered into two regimes after removing crash-linked, boundary and post-crash observations.
Red months with no new 20% forward drawdown. This raw 57.1% figure mixes true false alarms with repeated, boundary and post-crash readings.
14.3% of Red months remain after applying the mutually exclusive classification below.
3
Independent pre-crash
First qualifying Red observation for each holdout crash
5.4% of Red months20
Repeated confirmation
Additional Red months tied to the same approaching crash
35.7% of Red months2
Boundary-window
Warnings 13–18 months before breach, just outside the 12-month target
3.6% of Red months23
Post-crash state
Red during an existing drawdown or before recovery, including March–May 2020
41.1% of Red months8
Genuine false-alarm months
Red outside the pre-crash, boundary and post-crash definitions
14.3% of Red months| Independent false-alarm regime | Red months | Observed outcome |
|---|---|---|
| Jul 2015–Mar 2016 | 7 | 9.1% maximum next-12m monthly drawdown |
| Jan 2019 | 1 | 3.3% maximum next-12m monthly drawdown |
Classification rules
- Pre-crash: the 12 months before a 20% threshold breach; only the first Red is independent.
- Boundary: 13–18 months before breach.
- Post-crash: breach through recovery; March–May 2020 uses the daily detector because monthly averages missed the fast breach.
- False-alarm regime: remaining Red months less than six months apart are grouped together.
These categories are an explanatory audit, not a newly optimized trading rule.
Method, data sources and limitations
Two-layer design
The cycle layer measures 6–12 month vulnerability. The daily layer uses downside speed, VIX, volatility acceleration and spread widening to detect sudden market stress.
Important limits
FRED histories are revised rather than point-in-time vintages. The daily shock history contains only the 2020 and 2022 20% events. The 30% probability model failed validation and remains excluded.
Primary sources
Baa–Treasury spread, initial claims, building permits, M2, S&P 500, VIX, NFCI and SLOOS.