How limacro forecasts Nasdaq

1. The core idea

The world's major central banks collectively control the supply of risk-asset money. When they expand their balance sheets, liquidity flows into financial markets. When they shrink them, it drains out.

Equity prices — especially growth-heavy ones like the Nasdaq — respond to that liquidity with a lag. Historically that lag has been roughly three to four months. So if we measure global central-bank liquidity today, we can make a directional statement about where Nasdaq tends to be a few months from now.

That's the foundational idea. Everything else is the engineering required to convert noisy central bank data into a clean, usable signal.

2. Data sources

All inputs come from official, publicly available institutional sources. No paywalled feeds, no anonymous data brokers, no proprietary insider channels.

Data freshness varies by source. Most US data updates within one trading day. Some European and Asian central bank series have a one- to four-week reporting lag — the model handles these timing differences with appropriate interpolation and freshness checks. Internal sanitization logic flags and smooths data anomalies (rare in practice).

The specific series identifiers, weighting scheme, FX-conversion logic and freshness rules are part of the model's implementation and not published.

3. The four signal classes

Every day, the model evaluates current macro conditions and assigns one of four signal classes. The classification is driven by two inputs: the rate of change in global liquidity, and consumer sentiment as a confluence filter for the top tier. (Market volatility and equity trend are shown for context but no longer drive the classification.) Specific thresholds and weighting are part of the model and not published.

4. Lead time & timing

Today's liquidity reading implies something about Nasdaq several months from now — not next week. The exact lead time is calibrated empirically against decades of historical data; we use a horizon of approximately one quarter to one third of a year, tuned for optimal signal strength.

Concretely: if the model classifies today as a High Conviction Bull, that's a directional statement about Nasdaq's trajectory over the next several months. The implied movement comes from historical statistics — in past High Conviction Bull periods, Nasdaq moved on average +6.4% over the relevant forward window.

This is a directional indicator, not a price target. The model says "Nasdaq tends to move in this direction over the next several months when these conditions hold" — not "Nasdaq will be at exactly X by date Y".

5. Backtest results

We classified every month from January 2010 to early 2026 using the model rules, then measured what Nasdaq actually did over the relevant forward window for each classification.

78.6% overall directional accuracy across the non-Neutral signals. On High Conviction calls specifically, the model was right in 49 of 58 historical cases.

"Hit" for Bullish tiers means Nasdaq was directionally positive over the forward window. For Bearish, it means Nasdaq was negative. Every monthly classification with a complete forward window is included — no cherry-picking.

6. What we deliberately don't do

• We don't force a directional call when the data is ambiguous. If liquidity sits in the neutral corridor, we say so. A meaningful share of the time the model has no high-conviction view. That's honest — most services fill the silence with marketing noise.
• We don't give price targets. The model makes directional claims, not point predictions like "Nasdaq will hit 32,000 by August". Anyone giving you a precise number is overselling.
• We don't add weak indicators just to look sophisticated. We tested numerous additional macro inputs during model development and rejected most — they added complexity without meaningful accuracy improvement.
• We don't hide the results. Every signal class's hit rate, sample size and historical performance is documented transparently. You can verify the numbers against public market data.

7. Limitations & honest caveats

The model is not magic. Things you should know:

• Bearish signals are harder than bullish. The Strong Bearish sample is only 16 signals (50% accuracy) — lower confidence than the bull tiers. When this tier is right, the decline is real and sizeable (-6.6% average); when wrong, the market moves about as far the other way (+9.8% average). Real bear markets are rarer than bull markets, so treat this tier as a lower-confidence, asymmetric-payoff signal, not a high-conviction trade.
• Lead time is an average, not a fixed delay. The historical lag from liquidity to Nasdaq movement is "roughly several months" — sometimes faster, sometimes slower. The model is best read as "within the coming quarter or two" rather than a precise calendar.
• The High Conviction tier depends on sentiment data. High Conviction requires a fresh consumer-sentiment reading (UMICH), which is monthly and published with a lag. When that reading is stale or unavailable, a genuine High Conviction setup is reported as Strong Bullish instead — conservative by design, but it can understate conviction at the margin.
• M2 rises structurally with inflation. The index combines central-bank balance sheets with broad money supply (US and China M2). M2 adds breadth, but it also trends upward over time as prices rise — so the index is read as a rate of change, not a level, to strip out that structural drift.
• Data dependencies. If a central bank changes its reporting cadence or methodology, the model is affected. We monitor data quality continuously, but no real-time system is perfect.
• This is not investment advice. limacro is a data and analytics tool. It tells you what historical patterns and current liquidity suggest. What you do with that information is your own decision. Past accuracy does not guarantee future results.