The actual edge
Most market commentary watches price and news. Bedrock watches the plumbing, the balance sheets, and the cross-confirmation across funding markets, credit, and currencies.
The inputs that matter most when conditions are shifting:
- Fed and Treasury plumbing – Overnight Reverse Repo (RRP), Standing Repo Facility (SRF), Treasury General Account (TGA), and Fed reserve balances. These show whether liquidity is draining or building.
- Fed balance sheet – Weekly changes in total assets and bank reserves. Quantitative tightening or easing shows up here first.
- Real yields and breakevens – The 10-year real yield and breakeven inflation expectations. These determine whether bonds are competing with equities and whether inflation is priced in or being repriced.
- Credit spreads and price action – High-yield spread levels, HYG versus LQD relative performance, and credit spread slope. Credit deterioration usually precedes equity stress.
- Dollar and FX – Dollar index (DXY), USD/JPY for carry dynamics, USD/CNY for China stress, AUD/USD for commodity sensitivity. Currency moves often lead or confirm regime shifts.
- Commodity relationships – Gold/copper ratio for stagflation or deflation signals, oil for inflation pressure, copper for growth proxy. Commodity divergences matter more than absolute levels.
- Leadership ratios – HYG/LQD for credit health, GDX/Gold for miner sentiment, KRE/XLF for regional bank stress, RSP/SPY for breadth. These ratios show what is actually working.
The edge comes from watching all of these together, not one at a time. When liquidity is draining, credit is tightening, the dollar is strengthening, and commodities are diverging, that combination matters more than any single signal. Bedrock tracks cross-confirmation across inputs to avoid false signals from noise in any one series.
Why these inputs are better together than alone
A single input can lie. Multiple inputs confirming the same story are harder to dismiss.
- If credit spreads are widening but HYG is outperforming LQD, the spread widening may be noise.
- If the dollar is strengthening but AUD/USD is not breaking down, dollar strength may not be driven by risk-off flows.
- If gold is rallying but the gold/copper ratio is flat, gold may be reacting to currency moves instead of real stress.
- If liquidity is draining (TGA rising, RRP falling) but credit is stable and equities are holding, the drain may not be material yet.
Bedrock waits for cross-confirmation before calling a regime shift. This reduces whipsaw and improves the signal-to-noise ratio.
How Bedrock handles lagging series
Not all inputs update at the same speed:
- Live (real-time during market hours): Equity prices, VIX, credit ETFs, commodities, dollar index, Treasury yields
- Daily with lag: Credit spreads from FRED (usually end-of-day, may lag 1-2 days), TGA from Treasury (updates after market close)
- Weekly: Fed balance sheet and reserve balances (updates Thursdays)
When a series is lagged, Bedrock:
- Uses the most recent available data point
- Notes the lag explicitly in data quality metrics
- Reduces confidence if the lag is material (more than 2 days)
- Uses live price action from related instruments as a cross-check (e.g., HYG/LQD price action when credit spread data is lagged)
How confidence changes when inputs are stale or missing
Confidence is a function of data quality and signal alignment:
- High confidence: All core inputs are live and recent (less than 1 day old), and multiple factors are aligned in the same direction.
- Medium confidence: Some factors are mixed, or one core input is lagged by 1-2 days, but the weight of evidence still supports the current classification.
- Low confidence: Multiple inputs are missing or stale (more than 2 days old), or signals are contradictory across factors.
When a key input is missing entirely (e.g., FRED API failure), Bedrock:
- Falls back to related live price action when available (e.g., HYG/LQD for credit instead of spread data)
- Marks the read as Medium or Low confidence depending on how many inputs are affected
- Notes the missing input explicitly in the trust section
The goal is to never hide data quality issues. If inputs are stale or missing, the read says so clearly.
What Bedrock does not do
Bedrock does not predict. It does not forecast where the market will be in a week or a month. It does not provide price targets or trade recommendations.
It translates current cross-asset conditions into a current read. That read can change quickly if conditions change. Skepticism is appropriate.