Currency Indexes

HOW THE INDEX IS BUILT: for each major currency, the geometric mean of its value against each of the other seven (equal weights, no trade weighting). It is computed from the seven USD pairs — by triangular arbitrage every cross is the ratio of two USD pairs, so this is exactly the same index, and it keeps stale or junk prints on thin crosses out (averaging the 28 crosses' own prints was up to 0.2% off at some Friday closes). 100 = the close of 2 Jan 2026 (UTC). Candles are built from finer bars so a high or low is a real index value (summing pair highs that happened at different times would overstate every range); wicks are read at those finer bars' opens and closes, a little shorter than a continuous index. Built candles are stored in the database exactly as built, so charts load from it and only the recent candles are recalculated on each request; a page older than the store is built live. Candles are bid-side aggregates like every forex candle here; the live last value uses mid quotes. WHAT IT IS AND IS NOT: the eight indexes are one way of splitting the 28 pairs — the gap between two indexes is exactly 8/7 of the move of the pair between them, so an index carries no information the pairs do not; it shows which SIDE of a pair is moving. Description of the recent tape, not a forecast.

EFFICIENCY = |net move| ÷ total day-to-day travel over the last 21 completed UTC days (1 = a straight line, 0 = back where it started), on the index's daily closes. Bands against a Gaussian random walk over 21 daily moves (400k-draw simulation): ranging under 0.09 (its bottom quarter), choppy 0.09–0.19, mixed 0.19–0.32, trending 0.32–0.5 and one-way 0.5 or more (together its top quarter; 0.5 or more about 1 random walk in 15). The Trending list holds the indexes at 0.32 or more and the Ranging list those under 0.09, so on random tape about 2 of the 8 would land in each list by chance. Every index outside the list is named under the table with its efficiency. The 5-day efficiency is context only: over 5 days a random walk reaches 0.5 on 41% of windows. Today = the last completed 4H close against the last completed UTC daily close, in the index's daily ATR (ATR = the index's own daily ATR throughout). What it means: a description of the path the index took. Measured on the pairs (study 25, 28 fiat pairs, 2021–26): a one-way month did not predict a one-way next month (2.3–2.7% vs a 4.1–4.8% base rate) — trendiness had no month-to-month memory; trading whole currency baskets by index trend or momentum lost after costs (study 37). Not measured as a signal on the indexes. Description of the recent tape, not a forecast.

HOW IT IS BUILT: each currency index's states on the chosen timeframe, distances in that index's own ATR on that timeframe (daily index ATR on Daily, 4H index ATR on 4H) — AT SUPPORT / AT RESISTANCE = inside, within 0.2 index ATR of (the next zone beyond included), or pushing through by under 0.75 index ATR (no breaking close yet) a support zone below / resistance zone above (the At support / resistance tab's rule); a currency at both is listed on both sides but left out of the support-vs-resistance group, unless a rejection decides its side (next); THE ZONE THAT LAST TURNED IT DECIDES THE SIDE (your rule, added 2026-10-02): for a currency at a zone, the newer of its rejections in the last two trading days (the rejection below; newer = the latest 4H candle that reached each zone) decides which side it is listed on, as long as the last close is still within 1.5 ATR of that zone (the zone engine's 'approaching' band) — even when it sits inside, at or pushing through a zone on the other side; otherwise, or when one 4H candle reached both zones, where it sits decides — so USD back at a small pivot zone a day after the November 2025 resistance zone turned it down is listed at resistance, not support. A rejection never adds a currency to the listing or removes one from it, only decides its side (so its pair rows can move between groups, and a currency at zones on both sides is then listed on one side only): on the last 150 daily closes it changed the side on 5 of the 196 index-days at a zone, on the last 600 4H closes on 93 of the 772 index reads at a 4H zone (counted with the earlier break rule, before the false-break exception) (the rejection window is the same 12 4H candles on both timeframes — 2 daily bars, but 12 4H bars). It is a rule, not a measured ranking of zones; UPTREND / DOWNTREND CLEAR OF ZONES = efficiency 0.32 or more over the last 21 closes with no zone, the next zone beyond included, within 0.75 index ATR on either side and not pushing through any zone (the Trending, clear of zones tab's rule). Each group pairs two currencies by YOUR rules (the side at support or in the uptrend up, the side at resistance or in the downtrend down) and names the market pair and what the rules imply for it. Daily = the index zones on study 35's daily bars and 21 daily closes; 4H = the same zone engine and measure run on 4H index candles and 21 4H closes — 4H zones are unmeasured on any timeframe (the pair charts use daily zones only). The pair's move = 7/8 of the gap between its two indexes over the same 21 closes (exact in log terms). Every reason line also names the zone on the OTHER side (the resistance zone above a currency listed at support, the support zone below one listed at resistance) and any REJECTION in the last two trading days — a 4H candle that opened on the near side of, or inside, the nearest zone of that role beyond the last close reached it (after the zone took that role, and for 4H zones after the zone was confirmed) and every 4H close since is back on the near side (index wicks from 15-minute index values; the pairs' own 4H candles on the pair line) — so a row never hides a bigger level next to the one that listed it; beyond deciding the side as above, rejections and the other side do not change which group a currency is in or the verdict. Card = the pair's state on the aligned-pairs card (week and month 4H push/pull, the same on either timeframe). THE PAIR ITSELF: the same zone engine, the same listing rule (the zone that last turned it decides the side) and the same measure run on the pair's OWN candles (Daily = the same daily zones the pair page draws; 4H = the engine on the pair's 4H candles, unmeasured), priced at the same 4H close as the index read; its distances and the same 0.2 / 0.75 thresholds are in the pair's own ATR on that timeframe (daily pair ATR on Daily, 4H pair ATR on 4H). The pair page reads those zones at its latest 5-minute close, so its chip can differ by up to one 4H candle's move. The 'Pair' column is the indexes' figure (7/8 of their gap, exact in logs); the pair-itself line uses the pair's own candles — usually within a few hundredths of a percent, further apart when a cross carries an off-market print at either end of the 21 closes. Its zones and its efficiency are its own. AGREES = the pair is itself at the zone the rule implies (a support zone for an up rule, a resistance zone for a down rule) or trending that way clear of zones; OPPOSITE = at the other zone, or trending the other way clear of zones; NEITHER = otherwise. Agreement is the same geometry seen on a second series — not a measured edge (on the pairs, at fresh arrivals a zone held 5 days 33% of the time vs 35-36% at a made-up zone of the same width 1.5 ATR further, and already-touched zones were no different from the made-up one — study 34). BY CONSTRUCTION: the two indexes carry no information the pair does not — their gap IS the pair's own path; a pairing can run with or against the pair's recent direction, and the 'same way / against it' line under each move says which. WHAT IS MEASURED (none of it in favour): with an ALIGNED pair's strong currency index inside, pushing through or at a resistance zone (or its weak index at a support zone), the pair went 1 daily ATR with the trend before 1 against 48% of the time vs 44% when neither of its two indexes had a zone within 1.5 index ATR (49% otherwise), and its week turned against the trend 20% vs 20% of the time — none of the gaps held in both halves of the data (study 35, earlier break rule, aligned-card readings only; pairs not aligned on the card are outside what it measured); on the pairs, at fresh arrivals a zone held 5 days 33% of the time vs 35-36% at a made-up zone (study 34, earlier break rule); trendiness had no month-to-month memory (study 25) and trading whole currency baskets by index trend or momentum lost after costs (study 37). Rows sharing a currency move together — several rows can be one bet. A screen of your rules, not a recommendation. Description of the recent tape, not a forecast.