Memory 35% of MII
0.559
Do patterns stick around? Measures persistence in price changes and in how large those changes are.
Market Intelligence Index
How organized is the market?
Memory, shared patterns and connections — three views of market structure, combined into one research score.
Saved market-data snapshotData through 29 Sep 2026 · not a live feed
ELI5 · TL;DR
The saved MII reading is LOW. The author’s research favors mean reversion in this regime. ELI5: look for stretched moves returning toward their recent average.
Think of markets as a crowd: this saved score suggests less organization than usual. LOW is not bearish; HIGH is not bullish.Market Intelligence Index
0.319
0–1 relative score · not a probability
Relative regime
LOW
Below the lower cutoff of 0.427
Last full calculation
15 Sep 2026
129 scheduled estimates in this run
Price table through
29 Sep 2026
2,951 daily price observations
Why two dates? The model recalculates every 21 sessions using the latest 252 daily returns. The price table can extend beyond the calculation date; the last scheduled reading is held between updates.
Prices meet market structure
SPY & QQQ, through each regime
Both start at 100 in the selected window. Background colors show MII at that day’s close.
Read the percentages as strategy research. They are the paper-reported strategy win rates in each regime, not the chance SPY or QQQ goes up. The colors describe structure, not buy/sell signals. Author’s table · page 22 ↗
SPY · S&P 500 ETFQQQ · Nasdaq-100 ETFMII
Adjusted close · saved through 29 Sep 2026Hover or tap · arrow keys to explore · blank bands = warmup
What happened after each regime?
A different question: how often did simply holding each ETF produce a positive return?
Correlation compares the prior-session MII score with the later ETF return: +1 means they rise together perfectly, 0 means no linear relationship, and −1 means they move oppositely.
How to compare: “All eligible dates” is the baseline for the same period. A positive-return frequency of 65% means the ETF finished higher in 65 out of 100 observed starting dates. N counts overlapping starting dates, so many observations share the same market move. The paper’s 63% for LOW mean reversion answers a different question.
See all paper rates, simple strategy definitions and how this comparison works
| Strategy | LOW | MEDIUM | HIGH |
|---|---|---|---|
| Mean reversion | 63% | 55% | 48% |
| Breakout | 46% | 57% | 51% |
| Trend following | 44% | 53% | 61% |
Mean reversion: expect an unusually stretched move to drift back toward its usual level. Breakout: look for price to leave an established range. Trend following: expect an established move to continue. The paper reports these patterns; this page does not run those trading rules or establish their future success.
Timing. The colored bands and hover show that date’s closing MII label. Outcome tables use the previous session’s MII score and label, then measure adjusted close on the displayed starting date to adjusted close 5, 20 or 60 ETF trading sessions later. This keeps the signal earlier than entry. The range selector filters starting dates. An outcome is included only when its full horizon is observed by the end of this saved dataset.
Fair baseline. Both baseline and regime groups use only dates with a known prior MII label and a complete forward return. Warmup and unfinished outcomes are excluded. “Positive” means strictly above zero. N is the number of overlapping starting dates, not independent trades. Tables describe association, with no trading costs, position rules or proof of an edge.
Two separate datasets. MII remains the existing nine-asset proxy calculation above; SPY and QQQ are a price comparison and do not replace its inputs. ETF prices are Yahoo Finance adjusted closes (splits and dividends), retrieved 30 Sep 2026 at 03:14 UTC, with no price filling. The paper rates use the author’s study, not this proxy dataset. ETF prices, outcomes and provenance are kept with the project’s saved data.
MII through time
Explore the same saved result, at different scales.
MIIHistorical cutoffs
September 2023 – September 2026Hover or tap for values · arrow keys to explore
Read the chart: the gold line is MII. The dotted cutoffs mark the middle historical range: below it is LOW, inside it is MEDIUM, above it is HIGH. Cutoffs update monthly using earlier observations only.
The three pieces, in plain English
Component values below belong to the latest saved scheduled reading.
Information 40% of MII
0.302
Do markets share patterns? Measures order and statistical links between moves. It is not a news or social-media score.
Network 25% of MII
0.011
How connected is the group? Measures how assets move together and how their connections are arranged.
Relative to history, not a percentage chance. Each measurement is scaled against its own earlier range. A Network score near zero means its ingredients are near their historical lows in this run; it does not mean the assets have zero correlation.
ELI5 guide to the numbers
Raw versus scaled. A raw value is the original measurement, such as average correlation. Scaling puts different measurements on a 0–1 range so they can be combined. A scaled 0.30 does not mean 30% correlation or a 30% chance of profit.
Calibration. The model learns the earlier low and high of each measurement. New extremes can widen that range. The first valid reading uses 0.5 while the range begins forming; that is a starting value, not proof of neutral conditions.
Warmup. A calculation first needs 252 daily returns. Regime labels then need 252 earlier daily MII observations and the next monthly cutoff update. These observations include scores held between scheduled calculations.
LOW / MEDIUM / HIGH. The labels compare the score with its recent past. Proposed links to trend following, breakouts and mean reversion remain research hypotheses, not demonstrated trade recommendations.
Snapshot, method and source details
Saved dataset. Yahoo Finance daily unadjusted Close, 5 Jan 2015 through 29 Sep 2026. Retrieved 30 Sep 2026 at 00:18 UTC. This is a public proxy panel, not the paper’s original licensed dataset.
Markets. SPX, NDX, EURO STOXX 50 and E-mini S&P futures supply Memory and Information. The Network group also includes VIX, XLK, XLF, XLV and XLE.
Carried prices. 83 price cells use explicitly recorded earlier closes. Some final inputs are from 28 Sep even though the table ends 29 Sep. Local exchange dates do not imply simultaneous information availability.
Calculation. Nine original features; 252-return window, 21-session schedule, expanding scaling. Monthly historical terciles use up to 756 prior daily MII rows, with a fixed 252-row warmup.
Provenance is the receipt. The saved prices, source dates, fill report and checksums are kept with the project’s saved data. Fingerprints detect file changes; they do not prove a data provider is correct.
Original implementation. MarketMind by Layan Oraidi / Layan Aloreidi, version 0.2.0. Independent implementation; no author endorsement. The validation boundaries are documented with the project.
This page is a view-only saved snapshot. The controls explore the existing results; they do not fetch prices or recalculate the index. The full Python project provides calculations, CSV uploads and exports. No login or installation is needed to view this page.