Every time the 50-day average crossed the 200-day since 2010, and what price did in the 180 days after. Green bars ended higher than the cross day, red lower.
−16.85% 50/200 golden → 180 days later on May-22 2025, the last event in this build, checked through Oct-25 2026, measured and published by askthetape.com.
Two averages, one slow and one fast. When the fast one climbs through the slow one, chart people call it a golden cross; the other way, a death cross. We list every cross since 2010 and what price did in the 180 days after it. Crosses younger than 180 days are hollow: they show where price sits so far and are not scored yet. On the last closed day of this build the 50-day sits +12.1% from the 200-day. Pick another pair above: the 20/50 and 100/200 day averages, the 21/55 EMAs, or the price itself crossing its 200-day and 50-week averages; those last two come often, so they are scored 90 days later.
| leg | Mt.Gox daily closes Jul 2010 → Apr 2013, then Bitstamp daily closes · pairs: SMA 50/200, 20/50, 100/200, EMA 21/55, price vs SMA 200, price vs SMA 350 (50 weeks), price vs SMA 1400 (200 weeks), SMA 350/1400 (50-week vs 200-week) · score = close 180 days later / cross-day close − 1 (90 days for the price pairs) · a cross younger than its horizon is hollow, not scored |
|---|---|
| cadence | daily |
| weekly averages | weekly averages here are 350- and 1400-day simple averages of DAILY closes, not averages of weekly closes: measured against the real weekly closes on this build, the 50-week line differs by 0.48 % in the median and 11.25 % at its worst (799 weeks), the 200-week line by 0.22 % and 2.47 % (649 weeks) — the shape is the same, the exact level is not |
| since | 2010-07-17 · first cross possible after the slow average fills (200 days; 350 for the 50-week, 1400 for the 200-week — its first point is May 2014) |
| 🩸 what a «week» is on the two run views | the ISO week, and its close is the close of SUNDAY, UTC — the weekly bar for the week of Oct-25 2026 is the daily close of 2026-10-25. A week is judged only when its Sunday is in the tape and has closed; a week whose Sunday is missing (a hole in the tape, or a tape that has fallen behind) is LOST and is not drawn at all rather than being filled with a Friday or a Saturday: this build carries 0 lost weeks. The 50-week average is the SAME line the price pairs cross — SMA 350 of DAILY closes, not an average of weekly candles — taken on the week's own Sunday, and the side is the sign of close − average at full precision, not from the rounded series. First judged week 2011-07-10: the tape opens 2010-07-17 and the 350-day average has to fill first. |
| 🩸 how thin the closest week is | no week's side in this build is a rounding accident: the smallest margin between a weekly close and the 50-week average over all 799 weeks is 0.104 % (the week to Jul-18 2021, $31,777 against $31,744), which is some ten orders of magnitude above the difference between two honest ways of summing 350 closes. The average is summed here from the raw closes precisely so that the sign is never inherited from a 2-decimal series. |
| the runs, counted | 799 weeks closed, Jul-10 2011 → Oct-25 2026, in 31 runs (15 below, 16 above). Current run: 6 weeks above since Sep-20 2026. Longest below: 62 weeks (to Oct 2015) · 49 weeks (to Mar 2023) · 47 weeks (to Apr 2019). Longest above: 139 weeks (from Mar 2023) · 135 weeks (from Oct 2015) · 113 weeks (from Jun 2012). |
| 🩸 n = 3, and the 40-week cut is ours | «a run this long has ended before» rests on 3 FINISHED observations (62 · 47 · 49 weeks below → 135 · 29 · 139 weeks above). No significance is claimed, no test is run and no forecast is made here. The 40-week line is our own choice, and it is inert rather than tuned: anywhere between 21 and 44 weeks it selects the same runs, because the next longest run below in the whole record is 20 weeks. Read the other view's small bars next to it — the median of all 14 finished runs above is 2.5 weeks. A run below of 44 weeks has just ended and the run above it is STILL RUNNING (6 weeks): it adds nothing to n until it ends. |
| 🩸 what is NOT drawn on the run views | three things, on purpose. The first run of the record (Jul-10 2011, 13 weeks above) started before the 50-week average existed, so how long it had already run is unknowable — it is drawn on the run view but it is not an observation on «how long the first close above held». The run happening right now is not an observation either, for the same reason in the other direction. And the running ISO week is hollow — this build has no running week to draw: it is not a close, it is in no run, no count and no record until Sunday prints. |
| 🩸 what a «return above an average» is on the 12 views of runs in the «…average» groups (the two bitcoin 50-week views follow their own, stricter rule) | a RETURN is the first close ABOVE an average that comes immediately after a run of closes BELOW it. It is dated on the close that crossed — the one a reader could have seen in the moment — and it counts only when that close sits directly next to the close below it in the tape: across a hole nobody watched the crossing, so there is no event. The second view of each pair says how long each return above lasted, and a return that is STILL RUNNING is in no count, no median and no record until it ends. The FIRST run of each record is left-censored (the tape cannot see when it began), so the return after it is not an observation either. On the first view of each pair the height of a bar is the number of that period inside its run, positive above the average and negative below it, restarting at 1 the period the tape changes sides. |
| 🩸 what a «week» is here, and what a lost week does | the ISO week, and its close is the close of SUNDAY, UTC. A week is judged only when its Sunday is in the tape and has closed; a week whose Sunday is missing is LOST: it is not drawn, it is not filled in with a Friday or a Saturday, and — stricter than the two 50-week views above — THE RUN DOES NOT CONTINUE THROUGH IT. A run broken by a hole ends there and a new one starts after it, because a run is a claim about consecutive closes, and a return needs a crossing somebody could see. This build carries 0 lost periods in all (wke50 0 · d730 0 · wkband 0 · eth_wk50 0 · eth_wke200 0). The week we are in is hollow on the weekly run views: it is not a close, and it is in no run and no count until Sunday prints. |
| 🩸 the averages, named exactly | BTC 50-week EMA = EMA 50 over WEEKLY closes (the ISO week's Sunday close), seeded on the first weekly close of the tape, with the first 50 weeks of warm-up not judged · BTC 2-year average = SMA 730 of DAILY closes — the «2-year moving average» — judged only once the full 730-day window has filled · BTC 20-week / 21-week band = a BAND of two lines on weekly closes, SMA 20 and EMA 21, where «above» means the weekly close is above BOTH of them, so the binding line is whichever sits higher · ETH 50-week average = SMA 350 of DAILY closes read on the week's own Sunday — the same construction the bitcoin 50-week views on this page use, not an average of weekly candles · ETH 200-week EMA = EMA 200 over WEEKLY closes, seeded on the first weekly close of the tape, with the first 200 weeks of warm-up not judged. Every one of them is summed HERE from the raw closes at full precision, and the side of a period is the SIGN of close minus average — never inherited from a series already rounded for the chart. An EMA is seeded on the first close of its tape and its first N periods are warm-up, so these numbers begin where the average is honest, not where the tape is. |
| 🩸 the ether tape, and what «since» means on it | The ether views ride a DIFFERENT tape: Binance ETHUSDT daily closes 2017-08-17 →, captured by fetch_eth_daily.py into data/eth_daily_binance.csv. Ether traded before this tape did, so its 200-week EMA is judged only from Jun 2021 (200 weeks of warm-up on top of a tape that opens in 2017) and its 50-week average only from Aug 2018. «10 returns» on the ether views means «since that date», not «in the whole history of ether» — and the hero of each one says the date out loud. |
| 🩸 the saw cut is ours, it is printed, and it does NOT filter the counts | a tape sawing right at its average can mint a «return» every other close, so these views carry a cut: a run below of 60 days / 8 weeks before the return counts. It is printed on every one of them next to the number — and it does not filter the count, because on the 2-year average it would throw away two of the three longest returns on record (1,010 · 829 · 745 days), leaving a chart of 3 bars. So the count is every observed return, and the narrow cut stands beside it with its own numbers. Nothing here is a probability: «6 of 14 died» is a count out of a count, n is small on every one of these views, no test is run, no significance is claimed and no forecast is made. A thermometer, not a forecast. |
| 🩸 how the ten views read on the day of this build | BTC 50-week EMA (wke50, 800 weeks judged from Jul-03 2011, 0 weeks lost): now 6 weeks above · 22 returns, 9 died within 2 weeks, longest 127 weeks, median 6 weeks · 8 of them followed a run below of 8 weeks or longer (7 · 1 · 127 · 1 · 63 · 2 · 22 · 3 weeks) — BTC 2-year average (d730, 5,210 days judged from Jul-21 2012, 0 days lost): now 270 days below · 14 returns, 6 died within 14 days, longest 1,010 days, median 31.5 days · 3 of them followed a run below of 60 days or longer (38 · 166 · 829 days) — BTC 20-week / 21-week band (wkband, 829 weeks judged from Dec-12 2010, 0 weeks lost): now 10 weeks above · 29 returns, 11 died within 2 weeks, longest 68 weeks, median 7 weeks · 13 of them followed a run below of 8 weeks or longer (7 · 26 · 11 · 5 · 49 · 1 · 25 · 6 · 54 · 7 · 2 · 21 · 3 weeks) — ETH 50-week average (eth_wk50, 430 weeks judged from Aug-05 2018, 0 weeks lost): now 1 week below · 16 returns, 6 died within 2 weeks, longest 88 weeks, median 3.5 weeks · 5 of them followed a run below of 8 weeks or longer (6 · 25 · 12 · 18 · 5 weeks) — ETH 200-week EMA (eth_wke200, 280 weeks judged from Jun-20 2021, 0 weeks lost): now 2 weeks below · 10 returns, 3 died within 2 weeks, longest 72 weeks, median 4.5 weeks · 3 of them followed a run below of 8 weeks or longer (3 · 6 · 6 weeks). The daily view of the 2-year average is drawn thinned in the ALL window (each bucket keeps its highest and its lowest bar, so no run's peak is lost); every number printed on this page is computed from the FULL daily record, not from the drawn points. |
| 🩸 the 100-week average of ether, named exactly | ETH 100-week average = SMA 700 of DAILY closes read on the week's own Sunday — 100 weeks × 7 days, the same construction the 50-week ether view on this page uses, not an average of weekly candles. It is summed HERE from the raw daily closes at full precision, and the side of a week is the SIGN of that week's close minus the average, never inherited from a series already rounded for the chart. The tape is ONE venue and one pair: Binance ETHUSDT daily closes, captured by fetch_eth_daily.py into data/eth_daily_binance.csv (3,357 closes in this build) — not a composite spot index, not a perpetual, and nothing spliced in from a second exchange, so nothing here is inherited from anyone's index. The line most charts quote is a DIFFERENT CONSTRUCTION, not a different tape: an average of the last 100 WEEKLY closes. On this same tape that line stood at $2,740.64 on Oct-25 2026 against our $2,740.79 (a gap of $0.15); over the 380 weeks judged here the two constructions never put the close on different sides (0 of 380 weeks), and their two thresholds for the next Sunday are usually a few dollars apart. Ether traded before this tape did: 700 daily closes of warm-up sit on top of a tape that opens in Aug 2017, so the first week this average can be honest about is Jul 2019. «5 returns» on this view means «since that date», not «in the whole history of ether» — and the hero says the date out loud. |
| 🩸 how the 100-week view reads on the day of this build — and why the week we are in is not in it | ETH 100-week average (eth_wk100, 380 weeks judged from Jul-21 2019, 0 weeks lost): now 40 weeks below · 5 returns, 1 died within 2 weeks, longest 103 weeks, median 19 weeks · 3 of them followed a run below of 8 weeks or longer (1 · 68 · 19 weeks). A week counts here only once its SUNDAY has printed: the week we are standing in is hollow on this view, it is in no run, no count, no median and no record until Nov-01 2026 closes. That is the whole point of the view. «The price is above its 100-week average today» and «the week CLOSED above its 100-week average» are different claims, and dating the second one a week early is the mistake this page exists to avoid: every return counted above is dated on the weekly close that crossed — the close a reader could have seen in the moment — and a return that is still running is in no count at all. n is small, no test is run, no significance is claimed and no forecast is made. |
| this build | We have prices through Oct-25 2026; May-22 2025 is the latest scored 50/200 golden-cross date. |
| policy | no backfill · read it here · export: whole history |
| from | to | weeks | side | what followed |
|---|---|---|---|---|
| Sep-20 2026 | Oct-25 2026 | 6 | above | still running |
| Nov-16 2025 | Sep-13 2026 | 44 | below | above: 6 weeks from Sep-20 2026 |
| Mar-19 2023 | Nov-09 2025 | 139 | above | below: 44 weeks from Nov-16 2025 |
| Apr-10 2022 | Mar-12 2023 | 49 | below | above: 139 weeks from Mar-19 2023 |
| Mar-27 2022 | Apr-03 2022 | 2 | above | below: 49 weeks from Apr-10 2022 |
| Jan-02 2022 | Mar-20 2022 | 12 | below | above: 2 weeks from Mar-27 2022 |
| Dec-26 2021 | Dec-26 2021 | 1 | above | below: 12 weeks from Jan-02 2022 |
| Dec-19 2021 | Dec-19 2021 | 1 | below | above: 1 week from Dec-26 2021 |
| May-31 2020 | Dec-12 2021 | 81 | above | below: 1 week from Dec-19 2021 |
| May-24 2020 | May-24 2020 | 1 | below | above: 81 weeks from May-31 2020 |
| May-17 2020 | May-17 2020 | 1 | above | below: 1 week from May-24 2020 |
| May-10 2020 | May-10 2020 | 1 | below | above: 1 week from May-17 2020 |
| May-03 2020 | May-03 2020 | 1 | above | below: 1 week from May-10 2020 |
| Mar-08 2020 | Apr-26 2020 | 8 | below | above: 1 week from May-03 2020 |
| Jan-12 2020 | Mar-01 2020 | 8 | above | below: 8 weeks from Mar-08 2020 |
| Dec-29 2019 | Jan-05 2020 | 2 | below | above: 8 weeks from Jan-12 2020 |
| Dec-22 2019 | Dec-22 2019 | 1 | above | below: 2 weeks from Dec-29 2019 |
| Dec-15 2019 | Dec-15 2019 | 1 | below | above: 1 week from Dec-22 2019 |
| Dec-01 2019 | Dec-08 2019 | 2 | above | below: 1 week from Dec-15 2019 |
| Nov-24 2019 | Nov-24 2019 | 1 | below | above: 2 weeks from Dec-01 2019 |
| May-05 2019 | Nov-17 2019 | 29 | above | below: 1 week from Nov-24 2019 |
| Jun-10 2018 | Apr-28 2019 | 47 | below | above: 29 weeks from May-05 2019 |
| Jun-03 2018 | Jun-03 2018 | 1 | above | below: 47 weeks from Jun-10 2018 |
| May-27 2018 | May-27 2018 | 1 | below | above: 1 week from Jun-03 2018 |
| Oct-25 2015 | May-20 2018 | 135 | above | below: 1 week from May-27 2018 |
| Aug-17 2014 | Oct-18 2015 | 62 | below | above: 135 weeks from Oct-25 2015 |
| Jun-17 2012 | Aug-10 2014 | 113 | above | below: 62 weeks from Aug-17 2014 |
| Jan-29 2012 | Jun-10 2012 | 20 | below | above: 113 weeks from Jun-17 2012 |
| Jan-08 2012 | Jan-22 2012 | 3 | above | below: 20 weeks from Jan-29 2012 |
| Oct-09 2011 | Jan-01 2012 | 13 | below | above: 3 weeks from Jan-08 2012 |
| Jul-10 2011 | Oct-02 2011 | 13 | above | below: 13 weeks from Oct-09 2011 |
The chart itself is drawn in the browser: turn JavaScript on to hover a day, pick a window and draw your line. The numbers above come from the last build, Oct-26 2026 08:25Z.
What the tape says today · what's new · all 111 metrics