Map · Link network
The live links between companies: which ones persist, which have just started, are strengthening, weakening or have ended; and a company's 1st, 2nd and 3rd level links.
Summary and hit rate of the states
What it is: Today's number of live links, the number expected by chance, direct and indirect links, those explained by sector or shareholding, and the links that started, ended, strengthened or weakened in the last period. Below that, how well each state stated in the past held in the following period.
What data feeds it: Daily returns; each day's market average is subtracted so that a move affecting the whole market links no one. The data are split backwards from today into periods of 250 trading days. Sector and shareholding information comes from Settings › Data › Company information.
What it shows: Two companies moved together in a period if the correlation of their returns is positive with t of 2 or more. A link is live if it held in both of the last two periods, or if it held in most of its history (at least 5 periods and 60%) and held again in the last period. Just started: live for at most two periods. Strengthening, weakening: the correlation of the last period differs significantly from the previous three periods (Fisher z of 2 or more); a live link that did not hold in the last period also counts as weakening. Ended: did not hold for two periods in a row. The chance count is computed from the co-movement rate obtained when one company's days are matched with the days of a different period.
What can be concluded from it: Relationships live and change: suppliers change, business volume changes, shareholdings end. In the 2015–2026 measurement, of the 207 links alive in 2017–2019, only 71 were still alive in 2024–2026. The hit rate of the states is shown on the screen; the probability that a just-started link also holds in the next period is lower than that of a persisting or strengthening link, but several times that of a pair with no link. Co-movement shows that a link exists, not its direction (who affects whom).
Changing links
What it is: The list of links that started, ended, strengthened and weakened; filtered with the buttons at the top.
What data feeds it: The same computation as the summary: the correlation of daily returns with the market average subtracted, over periods of 250 trading days.
What it shows: Each square on the strip is a period, from oldest to newest left to right: a filled green square is a period in which they moved together, a grey square one in which they did not, a dashed square one with insufficient data; the depth of green is the strength of the correlation. The correlations of the last period and the previous three periods, the known reason explaining the link (same sector or shareholding), and whether the link is direct or comes through which company are shown.
What can be concluded from it: A link without a known reason (inorganic) may be a supplier–customer relationship, a shared investor base or a common economic influence. A link that comes through another company disappears once that company is taken into account; the real relationship is with the intermediary.
Scan log
What it is: The list, in date order, in which the link network scan records the links whose state has changed since the previous scan.
What data feeds it: The LINKS job, which runs every business day at 20:00 (Europe/Istanbul, after the evening price and company data); the link_edge table holds the latest state of the links, and the link_event table holds every state change and the day it was seen. The job can also be started by hand from Settings › Scheduler.
What it shows: On which day which link started, strengthened, weakened, ended or is no longer tracked. The first scan records only the initial state and counts no changes.
What can be concluded from it: Links are a living structure; this list is the record of how the network changes over time. Because periods are 250 trading days, a link's state changes slowly from day to day; a one-day change is not conclusive on its own, and whether it persists in later scans should be checked.
Companies without links
What it is: Companies that have no active link at all in the live network, the reasons, and whether W · Weighted wave applies force to them through a filler.
What data feeds it: The same link network computation: the company's past links (ended or weakening), how many periods its own price history covers, and the fillers the wave chose on evidence.
What it shows: Reason: "Link ended or weakened" (with which peer, when, last-period correlation), "Short history" (two periods, 500 trading days, not yet complete), "No lasting link found". The W · Weighted wave column shows the filler applied or "No force".
What can be concluded from it: A company without links has a price that moves in its own way; this is not an error. The filler is applied because it was measured that these companies also recovered in the past when they fell behind similar companies. The same explanation appears on the Stock screen when W is selected.
Supplier–customer links
What it is: Commercial relationships that come from sources rather than prices, and next to them whether the two companies move together in prices.
What data feeds it: The supplier_link table. The administrator enters them by hand or has the AI search for them with web search; the AI proposes only relationships between followed companies for which it can show a source (KAP special situation disclosures, notes to financial reports, annual reports, SEC EDGAR for companies traded in the US). Proposals enter no computation until the administrator approves them. Only free sources are used.
What it shows: Supplier → customer, relationship, source link, status (proposed, approved, rejected), origin (by hand or AI) and its counterpart in the link network (moving together, ended, not seen). If continuous scanning is on, each evening only companies that have published something new are searched: a new business relationship, tender or order disclosure on KAP, a new financial report or annual report; companies never scanned are searched once.
What can be concluded from it: A relationship not visible in prices is not wrong; it may be a small share of the company's revenue. Approved links are measured as a separate type in the W · Weighted wave evidence table and join the wave only if they hold in both periods. The cost of scanning is paid from the AI choice of the administrator who turned scanning on.
Company links
What it is: The selected company in the centre; directly linked companies in the 1st ring, and those reached through them in the 2nd and 3rd rings. Next to it, the company's live links and its weakening and ended links.
What data feeds it: Shortest path on the network built only from direct, live links. After a scan it is read with a Cypher query from the Apache AGE graph (bag_agi) in PostgreSQL: companies are :Sirket nodes, live direct links are :DOGRUDAN edges, other links are :BAG edges. If there has been no scan yet, the same path is found from an on-the-fly computation.
What it shows: Hovering over a company in a ring shows the path from the centre to it (e.g. THYAO → SISE → FROTO). Clicking a company brings it to the centre.
What can be concluded from it: Something happening at a large company is expected to show first in the 1st-level companies, then in the companies linked to them. In the measurement, next-day co-movement weakens as the level increases; this is a direct indication of propagation as the sheet theorem describes, but on its own it is not a buy or sell indicator.
Relations from the news (administrators)
What it is: Company pairs named in the same item by the followed news sources, all exchanges together; how many items named each pair and what the AI found.
What data feeds it: The relation_candidate table. Companies are found in news by name (legal suffixes removed), by tickers written like "(NASDAQ: NVDA)" and "$TGT", and on Borsa İstanbul by their upper-case code; an item naming more than five companies counts as a list. Every day at 21:30 the most mentioned pairs (at least two items) are asked of the fast model, ten at a time.
What it shows: Candidate pairs, pairs ready to be asked and checked pairs, the thirty most mentioned pairs and their results: supplier–customer, partnership, competitor, ownership or no relation.
What can be concluded from it: Appearing in the same item is not a relation by itself; the AI only reports a relation it can show with a source. A sourced supplier–customer relation becomes a proposed supplier link awaiting an administrator's confirmation. The cost is paid by the administrator who turned on the supplier scan and stops at the operator's budget warning line.
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.