How limacro forecasts Nasdaq
An overview of the model — what we measure, how we classify it, what the backtest showed, and where the limitations are. Specific implementation parameters are kept internal to protect the integrity of the signal.
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.
| Component | Type of source |
|---|---|
| Federal Reserve liquidity | US central bank reporting (weekly) |
| European Central Bank | ECB weekly financial statements |
| Bank of Japan | BoJ monthly balance sheet |
| People's Bank of China | International banking statistics |
| Bank of England | International banking statistics |
| FX rates (EUR, JPY) | Daily international exchange rates |
| Nasdaq 100 index | Daily equity market data |
| Market volatility (VIX) | CBOE volatility index, daily |
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 considers global liquidity changes, market-stress indicators, and equity trend confirmation. Specific thresholds and weighting are part of the model and not published.
The model's highest-conviction signal. Triggered when global liquidity is meaningfully expanding above an empirically tuned threshold. The other model inputs (volatility, trend) do not need to confirm — historically, strong liquidity has dominated other macro factors.
When this signal has fired historically, Nasdaq has followed in 42 of 43 cases.
Liquidity is contracting, but other macro inputs have not confirmed a bear case. The bear formation is incomplete. We surface this as a softer call rather than forcing a Strong Bearish classification — a deliberate design choice to avoid premature signals.
Historical hit rate: 62%. Useful as "watch" signal, not high-conviction trade.
Liquidity is flat. The model has no high-conviction view. We deliberately do not force a guess here. Most competing services fill this silence with marketing noise — we don't.
Historically Nasdaq drifts modestly positive during these periods (~78% baseline), but that's general market behavior, not a model prediction.
A triple-confirmation signal: contracting liquidity, elevated market stress, and a confirmed equity trend reversal. All three must align — a deliberately strict standard.
Historical hit rate: 75% over 8 signals. Bearish signals are inherently harder to forecast than bullish — the strict triple-confirmation design reduces false positives.
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 GREEN LIGHT, that's a directional statement about Nasdaq's trajectory over the next several months. The implied movement comes from historical statistics — in past GREEN LIGHT periods, Nasdaq moved on average +12.1% 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 2018 to early 2026 using the model rules, then measured what Nasdaq actually did over the relevant forward window for each classification.
| Signal class | Sample size | Hit rate | Avg move |
|---|---|---|---|
| Strong Bullish | 43 signals | 97.7% | +12.1% |
| Yellow Zone | 37 signals | 62.2% | +2.4% |
| Neutral | 7 signals | n/a | +4.9% |
| Strong Bearish | 8 signals | 75.0% | -5.7% |
94.1% overall directional accuracy across the 51 non-Neutral signals. On Strong Bullish calls specifically, the model was right in 42 of 43 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 is flat and stress signals are mixed, we say so. About half 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 8 signals (75% accuracy) — statistically meaningful but smaller than the bullish sample. Real bear markets are rarer than bull markets.
- 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.
- Yellow Zone has high variance. 62% hit rate with a wide range of outcomes. Useful information but not a high-conviction actionable signal.
- Broad money supply (M2) is intentionally excluded. Some liquidity tools include M2 alongside central bank assets. We tested this extensively — adding M2 doesn't improve correlation meaningfully and structurally biases the index upward (M2 always rises with inflation). We focus on the active policy signal.
- 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.