Map · W · Weighted wave
A wave model in which companies are placed as points with mass according to their size, connected to each other by springs through sector, shareholding, correlation and live link network links, and how much each company is currently pushed and pulled by its neighbours.
The wave
What it is: Each point is a company; its size shows its mass, which comes from its market capitalisation, and its colour the direction of the net force from its neighbours. The lines are springs: purple shareholding, grey same sector, dashed blue return correlation. Only link types that pass the two-period test are drawn and exert force. The layout is computed from all candidate links: related companies sit close, unrelated ones far apart.
What data feeds it: Company information (Settings > Data): market capitalisation, detailed sector, shareholding structure. Correlation link: two companies are among each other's five nearest neighbours in the residual returns of the last 500 trading days. The spring length is the inverse of the link strength; the coordinates are found by multidimensional scaling of the shortest paths over the links.
What it shows: Mass = (market capitalisation / median)^0.5. Position = the accumulation of residual return decaying with a 20-day half-life. Force = Σ link strength × (neighbour's position − own position) / mass; all neighbours are summed at once. Clicking a point lists, on the right, who pushes and who pulls that company.
What can be concluded from it: This model shows how companies are connected and who is currently pushing whom; it is not a forecast. In the 2020–2026 measurement, the force of correlation and shareholding links significantly predicted the return of the first tradable day, while that of sector links did not; because sector links are by far the most numerous, they used to dilute the other two. Which types are used is measured again at each computation and stated in the table above.
Pushers and pullers
What it is: A neighbour-by-neighbour breakdown of the force acting on the selected company: market capitalisation, mass, accumulated movement, net force and the largest contributions.
What data feeds it: The parts of the same computation: for each neighbour, link strength × position difference / mass. The reason for the link (shareholding, sector, correlation) is shown next to it; hovering over the shareholding label shows which shareholding it is.
What it shows: A positive contribution pushes the company up, because that neighbour has risen more than this company in recent days; a negative contribution pulls it down. The sum of the contributions is the net force.
What can be concluded from it: Here you can see a company being affected from different directions at the same time: opposing contributions can cancel each other out. Because the force carries no measured predictive power, this breakdown should not be used as a buy or sell indicator.
Evidence
What it is: How well the force predicted the return of the following days in the past, in both its mass-weighted and massless forms.
What data feeds it: The force computed each day in the second half of the period. Correlation links are learned from the first half; masses are computed from the capital that could have been published by that day and that day's close. The sector and shareholding structure are as they are today, because their history is not published.
What it shows: Rank correlation with the residual return 1, 2, 5 and 10 days later, and the t statistic. At the top, a plain sentence states which link types exert force. The table below measures each link type separately: the first-tradable-day result in each half of the period separately, the day 1, 2, 5 and 10 results over the whole period, and whether the type is used; hovering over the status label shows the reason. If company information is missing, it also states how many companies have size, shareholding structure and capital history.
What can be concluded from it: A link type is used if its force predicted the return of the first tradable day in the right direction with t of 2 or more in both halves; a relationship that holds in only one period is counted as chance. The last row of the table is the live link network (W · Weighted wave): for each measured day, the links are rebuilt once a month only from returns up to that day, so no half of the history is set aside for learning and later days are never seen. If this row holds in both halves, today's active links replace the fixed correlation links; the same signal goes through the same bucket test as the other models on Summary and can be chosen as the "W · Weighted wave · 5 days" strategy in Auto Player. Because the types are chosen by the result of these two periods, the combined result of the links used is somewhat optimistic. If the t statistic of the second day is below 2, the force knows nothing over a period in which one could trade on it. If the mass-weighted form is not clearly better than the massless one, the contribution of size to propagation cannot be measured.
Companies without links and fillers
What it is: The evidence for the filler links the wave applies to companies that have no active link in the live network (the "Unlinked companies" and "Companies with short history" rows of the evidence table), and the row for supplier–customer links.
What data feeds it: For companies without a link in the network built each day only from returns up to that day, three fillers are tried: weak links (significant co-movement in only one of the last two periods), the 3 nearest peers (the highest positive correlations of the last period), and, for companies without two periods of history, the links of a single period. The supplier row comes from links approved by the administrator.
What it shows: The filler rows are measured only by the ranking of the unlinked companies among themselves; the "Company" column is the number of companies the filler reaches today. A filler is used if it has t of 2 or more on the first tradable day in both periods, and it pulls only that company; it does not change the force on its peer. In the pushers and pullers breakdown, filler contributions are marked "Weak link" or "Near peer".
What can be concluded from it: In the 2020–2026 measurement, the weak link filler held at 0.054 (t 6.6) and 0.049 (t 5.9) on day 2, and the 3 nearest peers at 0.045 (t 5.3) and 0.043 (t 5.2); the prices of unlinked companies also tend to recover when they fall behind similar companies. Companies with short history were never numerous enough on any day for measurement, so single-period links are not used. The fillers are chosen again each time the evidence is refreshed (after the evening link network scan) and also feed into the signals of model W.
The investment information, comments and recommendations given here are not within the scope of investment advisory services. Investment advisory services are provided under an investment advisory agreement to be signed between a client and brokerage houses, portfolio management companies, or banks that do not accept deposits. The signals here are produced from historical data with statistical models, are shown the same to everyone and are not personalised; they may not suit your financial situation or your risk and return preferences. Therefore, making investment decisions based solely on the information given here may not produce results that meet your expectations.