Does a breakout hold after two closes, a retest or stronger volume—and what does waiting cost? · By CoinNudge Research · Method reviewed 2026-09-15 · Guide updated 2026-09-15 · Historical study updated · Observation range: – · Calculation btc-breakout-confirmation-1.1 · Historical study: daily; latest retained observation: hourly
How to Confirm a Bitcoin Breakout and Spot a Fakeout
Current answer: As of , using Binance BTCUSDT and ETHUSDT spot, closed 1-hour candles; historical replay: Conclusion: judge what happens after the breakout, not just the breakout candle. In 233 eligible BTC events, compare conditions known four hours after the break with failure during the NEXT 24 hours. Sustained 2× turnover: 35.7% later failure with the condition (25/70), versus 60.1% without (98/163). Positive, strengthening MACD: 42.1% later failure with the condition (61/145), versus 70.5% without (62/88). RSI ≥70: 25.7% later failure with the condition (18/70), versus 64.4% without (105/163). Failure means two consecutive hourly closes below the original resistance minus 0.1 prior ATR. These are exploratory historical associations, not trading win rates or a proven independent edge; distance above resistance can explain part of the difference.
Historical study updated
Explore data products · Check Market Events sample · Explore Telegram alerts
Current source-backed snapshot
The key finding: look for momentum and trading activity that continue after the breakout—not just one strong candle. In this historical BTC sample, sustained turnover and strengthening momentum accompanied fewer later failures. A break above resistance, two holding closes or a retest alone did not certify a true breakout. The experiment's percentages, sample sizes and exact failure definition are stated immediately below. Recent-period evidence is weaker and some comparison groups are small; the full-history differences are not established as a reliable current trading rule. These are historical associations, not trading win rates; price already being farther above resistance may explain part of the difference.
Input documentation: Binance Spot Klines · Binance Public Data · CoinNudge frozen research contract · CoinNudge methodology
Observation window: Fixed 24-hour resistance; 20-day resistance sensitivity; four-hour decision landmark
Calculation cadence: Historical study: daily; latest retained observation: hourly
Calculation version: btc-breakout-confirmation-1.1
Download current dataset: JSON · CSV Raw values retain the dataset's published precision. Free fair-use limit: 60 requests per minute per IP, shared across all Research JSON and CSV endpoints.
Look for continuation after the break—not just a strong first candle
Bottom line: continuing momentum and turnover were associated with fewer later failures in this BTC sample. One strong breakout candle, two holding closes or a retest alone did not certify success. Strong RSI was not an automatic sell signal.
Recent evidence is weaker: in the later historical BTC period, strengthening MACD had 46.7% failure (14/30) versus 54.2% (13/24). The 95% difference interval is -33.2 to 20.5 percentage points. It spans zero: a reliably lower recent failure rate is not established. Other recent indicator groups are also small. Do not treat the full-history contrast as a proven current trading rule.
| Condition already observable | With condition: failures / sample | Without condition: failures / sample |
|---|---|---|
| Four hours of sustained ≥2× turnover | 35.7% · 25/70 | 60.1% · 98/163 |
| MACD positive and strengthening | 42.1% · 61/145 | 70.5% · 62/88 |
| RSI ≥70 | 25.7% · 18/70 | 64.4% · 105/163 |
Eligible events were still above the original resistance at the four-hour decision point, without an early confirmed failure. Later failure means two consecutive hourly closes below that resistance minus 0.1 prior ATR. These overlapping conditions are not independent predictions; being farther above resistance can explain part of the difference.
Strong RSI was not an automatic reversal
With RSI ≥70, BTC later failure was 25.7% (18/70), versus 64.4% (105/163) without it. This is a conditional historical frequency—not the probability your next trade succeeds.
Activity continuing matters more than a label
With the next four hours averaging at least 2× pre-break turnover, later failure was 35.7% (25/70), versus 60.1% (98/163). Unlike the original volume test below, this measures sustained activity, not only the trigger candle.
What this establishes: historical associations under one explicit failure definition. What it does not establish: an independent indicator edge, an optimal entry, a profitable trade, or a combined success score. The ten conditions overlap; do not add or multiply their percentages. No failure within 24 hours is not the same as a sustained rally.
The multi-indicator experiment: data, sample and timing
Source: Binance BTCUSDT and ETHUSDT spot hourly OHLC, base volume, quote turnover and valid taker-buy quote volume. Research window: 2021-09-01 00:00:00 UTC to 2026-09-15 00:00:00 UTC, end exclusive; an additional month seeds recursive indicators. Coverage and missing-bar counts appear below. These are reconstructed historical candles, not five years of locally observed alerts.
Keep the original resistance R: the preceding 24 completed hourly highs. Retain the first qualifying close and space events by 76 hours. At T=t0+4h, include only events above R without an early two-close failure, with the full 72-hour post-T path available. This survivor sample is not all initial breakouts. BTC and ETH are analyzed separately. All feature inputs end at T; no later prices enter a feature.
The outcome is two consecutive hourly closes below R−0.1×prior ATR during the next 24 hours after T. The numerator is the number of such failures; the denominator is the available condition-present or condition-absent group. Missing features are excluded from both groups, never interpreted as false. A failed path can later recover; a recovered failure is not erased.
| Condition | With: failures / N | Without: failures / N | Failure difference (pp) | 95% interval (pp) | Evidence |
|---|---|---|---|---|---|
| RSI(14) >= 70 | 18/70 · 25.7% | 105/163 · 64.4% | -38.7 | -50.3 to -26.2 | Exploratory · unadjusted |
| MACD(12,26,9) histogram > 0 and above its value three hours earlier | 61/145 · 42.1% | 62/88 · 70.5% | -28.4 | -40.6 to -15.8 | Exploratory · unadjusted |
| Bollinger(20,2 population SD) relative width rises from previous hour and close > upper band | 16/47 · 34.0% | 107/186 · 57.5% | -23.5 | -38.5 to -7.3 | Exploratory · unadjusted |
| Mean quote turnover in hours 1–4 >= 2 × pre-trigger 24-hour mean | 25/70 · 35.7% | 98/163 · 60.1% | -24.4 | -37.7 to -11.2 | Exploratory · unadjusted |
| T close minus frozen resistance >= 1 prior Wilder ATR(14) | 52/140 · 37.1% | 71/93 · 76.3% | -39.2 | -50.3 to -28.5 | Exploratory · unadjusted |
Each interval is an unadjusted 95% seven-calendar-day block-bootstrap interval, using 2,000 paired draws including empty days. Both groups need at least 20 events; sparse intervals are withheld. This extension followed exploratory inspection: it is not a preregistered confirmatory experiment.
All ten conditions, including weak or inconsistent filters
| Condition | With: failures / N | Without: failures / N | Failure difference (pp) | 95% interval (pp) | Evidence |
|---|---|---|---|---|---|
| RSI(14) >= 70 | 18/70 · 25.7% | 105/163 · 64.4% | -38.7 | -50.3 to -26.2 | Exploratory · unadjusted |
| MACD(12,26,9) histogram > 0 and above its value three hours earlier | 61/145 · 42.1% | 62/88 · 70.5% | -28.4 | -40.6 to -15.8 | Exploratory · unadjusted |
| Bollinger(20,2 population SD) relative width rises from previous hour and close > upper band | 16/47 · 34.0% | 107/186 · 57.5% | -23.5 | -38.5 to -7.3 | Exploratory · unadjusted |
| Mean quote turnover in hours 1–4 >= 2 × pre-trigger 24-hour mean | 25/70 · 35.7% | 98/163 · 60.1% | -24.4 | -37.7 to -11.2 | Exploratory · unadjusted |
| T close minus frozen resistance >= 1 prior Wilder ATR(14) | 52/140 · 37.1% | 71/93 · 76.3% | -39.2 | -50.3 to -28.5 | Exploratory · unadjusted |
| ADX(14) >= 25, rising over three hours, and +DI > -DI | 56/111 · 50.5% | 67/122 · 54.9% | -4.5 | -17.7 to 8.0 | Exploratory · unadjusted |
| T close > EMA9 > EMA21 > EMA50 | 85/173 · 49.1% | 38/60 · 63.3% | -14.2 | -29.2 to 1.0 | Exploratory · unadjusted |
| T close > current UTC session cumulative quote/base VWAP | 113/212 · 53.3% | 10/21 · 47.6% | 5.7 | -16.5 to 29.0 | Exploratory · unadjusted |
| Valid aggregate taker-buy quote / quote turnover in hours 1–4 >= 55% | 6/17 · 35.3% | 117/216 · 54.2% | -18.9 | Insufficient group size | Exploratory · unadjusted |
| Trigger upper wick >= 40% of trigger high-low range | 19/23 · 82.6% | 104/210 · 49.5% | 33.1 | 13.4 to 48.8 | Exploratory · unadjusted |
ADX strength, EMA alignment, VWAP position, active-buy share and upper wicks are deliberately retained in the results rather than hiding less persuasive comparisons. ADX measures strength, not direction; the test additionally requires +DI above −DI. An almost universal condition cannot discriminate well, and tiny groups do not support a reliable percentage.
ETH replication and recent-period BTC/ETH checks
ETH · full history
| Condition | With: failures / N | Without: failures / N | Failure difference (pp) | 95% interval (pp) | Evidence |
|---|---|---|---|---|---|
| RSI(14) >= 70 | 12/66 · 18.2% | 81/147 · 55.1% | -36.9 | -48.6 to -24.1 | Exploratory · unadjusted |
| MACD(12,26,9) histogram > 0 and above its value three hours earlier | 46/131 · 35.1% | 47/82 · 57.3% | -22.2 | -35.4 to -8.8 | Exploratory · unadjusted |
| Bollinger(20,2 population SD) relative width rises from previous hour and close > upper band | 7/39 · 17.9% | 86/174 · 49.4% | -31.5 | -45.4 to -16.8 | Exploratory · unadjusted |
| Mean quote turnover in hours 1–4 >= 2 × pre-trigger 24-hour mean | 19/63 · 30.2% | 74/150 · 49.3% | -19.2 | -31.2 to -5.1 | Exploratory · unadjusted |
| T close minus frozen resistance >= 1 prior Wilder ATR(14) | 44/136 · 32.4% | 49/77 · 63.6% | -31.3 | -45.1 to -16.8 | Exploratory · unadjusted |
| ADX(14) >= 25, rising over three hours, and +DI > -DI | 38/99 · 38.4% | 55/114 · 48.2% | -9.9 | -23.0 to 4.0 | Exploratory · unadjusted |
| T close > EMA9 > EMA21 > EMA50 | 64/161 · 39.8% | 29/52 · 55.8% | -16.0 | -30.8 to -0.6 | Exploratory · unadjusted |
| T close > current UTC session cumulative quote/base VWAP | 79/185 · 42.7% | 14/28 · 50.0% | -7.3 | -28.4 to 13.6 | Exploratory · unadjusted |
| Valid aggregate taker-buy quote / quote turnover in hours 1–4 >= 55% | 6/14 · 42.9% | 87/199 · 43.7% | -0.9 | Insufficient group size | Exploratory · unadjusted |
| Trigger upper wick >= 40% of trigger high-low range | 16/30 · 53.3% | 77/183 · 42.1% | 11.3 | -9.7 to 31.0 | Exploratory · unadjusted |
BTC · later historical period
| Condition | With: failures / N | Without: failures / N | Failure difference (pp) | 95% interval (pp) | Evidence |
|---|---|---|---|---|---|
| RSI(14) >= 70 | 3/11 · 27.3% | 24/43 · 55.8% | -28.5 | Insufficient group size | Exploratory · unadjusted |
| MACD(12,26,9) histogram > 0 and above its value three hours earlier | 14/30 · 46.7% | 13/24 · 54.2% | -7.5 | -33.2 to 20.5 | Exploratory · unadjusted |
| Bollinger(20,2 population SD) relative width rises from previous hour and close > upper band | 3/9 · 33.3% | 24/45 · 53.3% | -20.0 | Insufficient group size | Exploratory · unadjusted |
| Mean quote turnover in hours 1–4 >= 2 × pre-trigger 24-hour mean | 5/14 · 35.7% | 22/40 · 55.0% | -19.3 | Insufficient group size | Exploratory · unadjusted |
| T close minus frozen resistance >= 1 prior Wilder ATR(14) | 8/26 · 30.8% | 19/28 · 67.9% | -37.1 | -59.9 to -13.2 | Exploratory · unadjusted |
| ADX(14) >= 25, rising over three hours, and +DI > -DI | 10/20 · 50.0% | 17/34 · 50.0% | 0.0 | -27.7 to 25.0 | Exploratory · unadjusted |
| T close > EMA9 > EMA21 > EMA50 | 20/35 · 57.1% | 7/19 · 36.8% | 20.3 | Insufficient group size | Exploratory · unadjusted |
| T close > current UTC session cumulative quote/base VWAP | 27/50 · 54.0% | 0/4 · 0.0% | 54.0 | Insufficient group size | Exploratory · unadjusted |
| Valid aggregate taker-buy quote / quote turnover in hours 1–4 >= 55% | 3/7 · 42.9% | 24/47 · 51.1% | -8.2 | Insufficient group size | Exploratory · unadjusted |
| Trigger upper wick >= 40% of trigger high-low range | 5/5 · 100.0% | 22/49 · 44.9% | 55.1 | Insufficient group size | Exploratory · unadjusted |
ETH · later historical period
| Condition | With: failures / N | Without: failures / N | Failure difference (pp) | 95% interval (pp) | Evidence |
|---|---|---|---|---|---|
| RSI(14) >= 70 | 4/14 · 28.6% | 14/34 · 41.2% | -12.6 | Insufficient group size | Exploratory · unadjusted |
| MACD(12,26,9) histogram > 0 and above its value three hours earlier | 11/32 · 34.4% | 7/16 · 43.8% | -9.4 | Insufficient group size | Exploratory · unadjusted |
| Bollinger(20,2 population SD) relative width rises from previous hour and close > upper band | 2/10 · 20.0% | 16/38 · 42.1% | -22.1 | Insufficient group size | Exploratory · unadjusted |
| Mean quote turnover in hours 1–4 >= 2 × pre-trigger 24-hour mean | 5/15 · 33.3% | 13/33 · 39.4% | -6.1 | Insufficient group size | Exploratory · unadjusted |
| T close minus frozen resistance >= 1 prior Wilder ATR(14) | 8/30 · 26.7% | 10/18 · 55.6% | -28.9 | Insufficient group size | Exploratory · unadjusted |
| ADX(14) >= 25, rising over three hours, and +DI > -DI | 8/19 · 42.1% | 10/29 · 34.5% | 7.6 | Insufficient group size | Exploratory · unadjusted |
| T close > EMA9 > EMA21 > EMA50 | 13/31 · 41.9% | 5/17 · 29.4% | 12.5 | Insufficient group size | Exploratory · unadjusted |
| T close > current UTC session cumulative quote/base VWAP | 14/40 · 35.0% | 4/8 · 50.0% | -15.0 | Insufficient group size | Exploratory · unadjusted |
| Valid aggregate taker-buy quote / quote turnover in hours 1–4 >= 55% | 2/6 · 33.3% | 16/42 · 38.1% | -4.8 | Insufficient group size | Exploratory · unadjusted |
| Trigger upper wick >= 40% of trigger high-low range | 5/10 · 50.0% | 13/38 · 34.2% | 15.8 | Insufficient group size | Exploratory · unadjusted |
The later period starts 2025-09-01 and ends before protocol publication on 2026-09-15. Exclude 76-hour paths crossing either boundary. This is a retrospective time split, not untouched forward validation. Small groups and weaker or reversed differences must not be omitted. Correlated BTC and ETH markets are not independent proof.
Does being farther above resistance explain the result?
A price already farther from the fixed failure barrier is mechanically less likely to cross it. RSI, MACD and Bollinger are also price-derived. The table separates distance at T into prior-ATR bins; small cells remain descriptive. This is not a matched comparison controlling volatility, trend and market regime, so incremental predictive value remains unresolved.
| Distance above R · ATR | Condition | With: failures / N | Without: failures / N |
|---|---|---|---|
| [0,0.5) | RSI(14) >= 70 | 0/0 · —% | 39/49 · 79.6% |
| [0,0.5) | MACD(12,26,9) histogram > 0 and above its value three hours earlier | 6/9 · 66.7% | 33/40 · 82.5% |
| [0,0.5) | Bollinger(20,2 population SD) relative width rises from previous hour and close > upper band | 0/0 · —% | 39/49 · 79.6% |
| [0,0.5) | Mean quote turnover in hours 1–4 >= 2 × pre-trigger 24-hour mean | 1/1 · 100.0% | 38/48 · 79.2% |
| [0.5,1) | RSI(14) >= 70 | 0/1 · 0.0% | 32/43 · 74.4% |
| [0.5,1) | MACD(12,26,9) histogram > 0 and above its value three hours earlier | 13/17 · 76.5% | 19/27 · 70.4% |
| [0.5,1) | Bollinger(20,2 population SD) relative width rises from previous hour and close > upper band | 1/1 · 100.0% | 31/43 · 72.1% |
| [0.5,1) | Mean quote turnover in hours 1–4 >= 2 × pre-trigger 24-hour mean | 4/4 · 100.0% | 28/40 · 70.0% |
| [1,2) | RSI(14) >= 70 | 5/14 · 35.7% | 24/45 · 53.3% |
| [1,2) | MACD(12,26,9) histogram > 0 and above its value three hours earlier | 21/45 · 46.7% | 8/14 · 57.1% |
| [1,2) | Bollinger(20,2 population SD) relative width rises from previous hour and close > upper band | 6/12 · 50.0% | 23/47 · 48.9% |
| [1,2) | Mean quote turnover in hours 1–4 >= 2 × pre-trigger 24-hour mean | 4/11 · 36.4% | 25/48 · 52.1% |
| [2,inf) | RSI(14) >= 70 | 13/55 · 23.6% | 10/26 · 38.5% |
| [2,inf) | MACD(12,26,9) histogram > 0 and above its value three hours earlier | 21/74 · 28.4% | 2/7 · 28.6% |
| [2,inf) | Bollinger(20,2 population SD) relative width rises from previous hour and close > upper band | 9/34 · 26.5% | 14/47 · 29.8% |
| [2,inf) | Mean quote turnover in hours 1–4 >= 2 × pre-trigger 24-hour mean | 16/54 · 29.6% | 7/27 · 25.9% |
Exact formulas, field dictionary and reproduction scope
RSI and ADX use Wilder smoothing (14); EMA starts with a simple average; MACD uses EMA12−EMA26 and EMA9 signal. Bollinger uses 20 closes and population standard deviation ×2; width=(upper−lower)/middle. Recursive features require 200 contiguous hourly inputs ending at T. VWAP uses every completed bar in the UTC day of the T bar’s open, with sum quote divided by sum base volume, not typical-price weighting.
The JSON and CSV include technical_observations (observation ID, feature_time, raw values, flags and outcomes), technical_comparisons (counts, rates, missingness, interval and seed), and technical_distance_strata. four_hour_quote/4 divided by prior_quote_mean defines sustained relative volume; four_hour_buy_quote/four_hour_quote defines the valid active-buy share. distance_atr=(decision_close−reference_level)/atr_prior. Null means unavailable. feature_time is an as-of cutoff in a historical replay, not original local capture time.
The frozen input CSV now includes base volume v so VWAP can be reproduced. The original verifier checks the original breakout protocol; the additional technical verifier checks raw feature calculations, flags, all group counts/rates and distance-bin counts. Bootstrap rules and seeds are published in contract.technical_extension. Features are exploratory; 4h-EMA, ATR-percentile and pre-break squeeze extensions are not claimed as completed tests here.
What should a trader actually watch?
Evidence of continuation
Price holds beyond the original level, momentum strengthens after the initial candle, and turnover remains elevated. Treat these as questions for a monitoring checklist—not independent votes that guarantee success. High RSI during a breakout is not, by itself, a sell instruction.
Evidence of weakening or failure
Price remains close to resistance while momentum and turnover fade. Under this study’s rule, two later closes below the buffered level establish failure; subsequent reclaim is a separate outcome. A wick or one dip alone is not this defined failure.
Funding, open interest, liquidations and cross-venue spot flow can add market context, but their aligned local history does not support this same five-year experiment. They are not silently filled or combined into these percentages. A proper matched and forward study is still needed before presenting a trading rule.
First separate a breakout, a hold and a failure
A breakout candidate
An hourly BTC close above the highest high of the preceding 24 completed hours. A wick above that price is not enough. Freeze that resistance, called R, when the bar closes.
Evidence of holding
Ask whether the next two closes stay above R plus a small volatility buffer, or whether a defined retest holds. These describe an observed path; neither certifies a “true” breakout.
A defined failure
Two consecutive later hourly closes below R minus the same buffer. One dip back into the range is recorded separately. A later reclaim does not erase an earlier failure.
Confirmation has a price
All comparisons start at T, four hours after the trigger. Waiting may avoid an early failed move, but may also mean a higher reference price or missing a move without a retest.
What the BTC history actually says
Primary scope: Binance BTCUSDT spot, hourly candles, 24-hour resistance, 2021-09-01 00:00:00 UTC to 2026-09-15 00:00:00 UTC (end exclusive). ETH and 20-day resistance are separate checks, not extra BTC evidence.
The comparison excludes 121 early failures and 22 paths not above resistance at T from the 376 complete paths. They remain in the downloadable ledger. 1 retained event(s) have a pending or boundary-crossing window; 2 have an input gap. These are not counted as successful breaks.
How often did the breakout fail later?
24 hours after the common decision point| Condition | With: failures / N | Without: failures / N | Failure difference (pp) | 95% interval (pp) | Evidence |
|---|---|---|---|---|---|
| Two closes above the level | 103/200 · 51.5% | 20/33 · 60.6% | -9.1 | -26.9 to 9.4 | Descriptive, not causal |
| Retest held within four hours | 44/65 · 67.7% | 79/168 · 47.0% | 20.7 | 6.7 to 34.4 | Descriptive, not causal |
| At least 2× relative volume | 62/126 · 49.2% | 61/107 · 57.0% | -7.8 | -20.3 to 5.1 | Descriptive, not causal |
| At least 55% aggressive-buy share | 67/117 · 57.3% | 56/116 · 48.3% | 9.0 | -3.1 to 21.7 | Descriptive, not causal |
| Two closes + 2× volume | 55/114 · 48.2% | 68/119 · 57.1% | -8.9 | -20.9 to 3.9 | Descriptive, not causal |
Two closes: later failure is 51.5% with the condition and 60.6% without it. The difference interval is -26.9 to 9.4 percentage points. It spans zero: this sample does not establish a reliably lower failure rate.
A held retest: 67.7% versus 47.0%. In this definition and sample, retested paths fail more often later. A return to resistance selects a different price path; it must not be sold as automatic confirmation. This is a test of a narrowly defined four-hour retest, not every chart pattern called a retest.
What does waiting cost?
The next table is not a trading backtest. It measures the price paid for waiting and the subsequent path, using T close as a common reference. Positive waiting change means a higher reference price than at the initial break. MFE / MAE are the greatest favorable / adverse price excursions after T.
| Condition present | N | Median wait change | Median T→24h return | Median MFE / MAE | Extended without condition |
|---|---|---|---|---|---|
| Two closes above the level | 200 | 0.35% | -0.14% | 1.27% / -1.44% | 19/33 |
| Retest held within four hours | 65 | 0.22% | -0.81% | 0.99% / -1.96% | 119/168 |
| At least 2× relative volume | 126 | 0.30% | -0.39% | 1.38% / -1.39% | 57/107 |
| At least 55% aggressive-buy share | 117 | 0.13% | -0.36% | 1.10% / -1.46% | 81/116 |
| Two closes + 2× volume | 114 | 0.32% | -0.39% | 1.44% / -1.41% | 65/119 |
“Extended” means a later close reached R + 2 prior ATR within T+24h, whether or not it also failed later. The absent-condition count measures missed qualifying paths, not missed profitable trades. Costs, fills, stops and leverage are not modeled.
Does the pattern survive a later period?
Historical validation: 2025-09-01 to 2026-09-15 UTC, excluding any full 76-hour path crossing the boundary. The protocol was first published on 2026-09-15: this is a retrospective chronological split, not a previously untouched or prospectively preregistered test.
| Condition | With: failures / N | Without: failures / N | Failure difference (pp) | 95% interval (pp) | Evidence |
|---|---|---|---|---|---|
| Two closes above the level | 22/47 · 46.8% | 5/7 · 71.4% | -24.6 | Insufficient group size | Exploratory · unadjusted |
| Retest held within four hours | 9/14 · 64.3% | 18/40 · 45.0% | 19.3 | Insufficient group size | Exploratory · unadjusted |
| At least 2× relative volume | 10/27 · 37.0% | 17/27 · 63.0% | -25.9 | -50.9 to 1.0 | Descriptive, not causal |
| At least 55% aggressive-buy share | 21/35 · 60.0% | 6/19 · 31.6% | 28.4 | Insufficient group size | Exploratory · unadjusted |
| Two closes + 2× volume | 9/26 · 34.6% | 18/28 · 64.3% | -29.7 | -54.2 to -4.4 | Descriptive, not causal |
Development period, yearly differences and bootstrap sensitivity
| Condition | With: failures / N | Without: failures / N | Failure difference (pp) | 95% interval (pp) | Evidence |
|---|---|---|---|---|---|
| Two closes above the level | 81/153 · 52.9% | 15/26 · 57.7% | -4.8 | -24.6 to 16.0 | Descriptive, not causal |
| Retest held within four hours | 35/51 · 68.6% | 61/128 · 47.7% | 21.0 | 5.5 to 37.1 | Descriptive, not causal |
| At least 2× relative volume | 52/99 · 52.5% | 44/80 · 55.0% | -2.5 | -18.4 to 12.6 | Descriptive, not causal |
| At least 55% aggressive-buy share | 46/82 · 56.1% | 50/97 · 51.5% | 4.6 | -10.7 to 19.6 | Descriptive, not causal |
| Two closes + 2× volume | 46/88 · 52.3% | 50/91 · 54.9% | -2.7 | -17.7 to 12.3 | Descriptive, not causal |
Yearly figures are descriptive and often sparse. 2021 and the current year are partial. Changing the block length checks interval sensitivity, not whether a threshold was optimal.
| Year | Condition | With / without N | Failure difference (pp) |
|---|---|---|---|
| 2021 | Two closes above the level | 19 / 0 | — |
| 2021 | Retest held within four hours | 4 / 15 | 66.7 |
| 2021 | At least 2× relative volume | 7 / 12 | -7.1 |
| 2021 | At least 55% aggressive-buy share | 11 / 8 | -26.1 |
| 2021 | Two closes + 2× volume | 7 / 12 | -7.1 |
| 2022 | Two closes above the level | 32 / 6 | -4.2 |
| 2022 | Retest held within four hours | 9 / 29 | 48.3 |
| 2022 | At least 2× relative volume | 17 / 21 | 2.8 |
| 2022 | At least 55% aggressive-buy share | 9 / 29 | -10.0 |
| 2022 | Two closes + 2× volume | 17 / 21 | 2.8 |
| 2023 | Two closes above the level | 38 / 4 | -22.4 |
| 2023 | Retest held within four hours | 13 / 29 | -12.5 |
| 2023 | At least 2× relative volume | 30 / 12 | -5.0 |
| 2023 | At least 55% aggressive-buy share | 8 / 34 | 40.4 |
| 2023 | Two closes + 2× volume | 28 / 14 | -3.6 |
| 2024 | Two closes above the level | 40 / 10 | 0.0 |
| 2024 | Retest held within four hours | 16 / 34 | 36.8 |
| 2024 | At least 2× relative volume | 30 / 20 | -8.3 |
| 2024 | At least 55% aggressive-buy share | 31 / 19 | 29.7 |
| 2024 | Two closes + 2× volume | 24 / 26 | -8.0 |
| 2025 | Two closes above the level | 41 / 7 | -10.8 |
| 2025 | Retest held within four hours | 13 / 35 | 8.1 |
| 2025 | At least 2× relative volume | 28 / 20 | 5.0 |
| 2025 | At least 55% aggressive-buy share | 34 / 14 | 7.1 |
| 2025 | Two closes + 2× volume | 24 / 24 | -4.2 |
| 2026 | Two closes above the level | 30 / 6 | -16.7 |
| 2026 | Retest held within four hours | 10 / 26 | 10.0 |
| 2026 | At least 2× relative volume | 14 / 22 | -39.6 |
| 2026 | At least 55% aggressive-buy share | 24 / 12 | 29.2 |
| 2026 | Two closes + 2× volume | 14 / 22 | -39.6 |
| Condition | Block days | Failure difference: 95% interval (pp) |
|---|---|---|
| Two closes above the level | 3 | -27.6 to 9.2 |
| Retest held within four hours | 3 | 6.6 to 34.1 |
| At least 2× relative volume | 3 | -20.6 to 4.7 |
| At least 55% aggressive-buy share | 3 | -3.5 to 21.3 |
| Two closes + 2× volume | 3 | -21.4 to 3.7 |
| Two closes above the level | 14 | -25.6 to 9.6 |
| Retest held within four hours | 14 | 6.2 to 34.0 |
| At least 2× relative volume | 14 | -21.0 to 5.2 |
| At least 55% aggressive-buy share | 14 | -3.5 to 21.9 |
| Two closes + 2× volume | 14 | -22.2 to 4.3 |
Forward evaluation begins on 2026-09-16 at 00:00 UTC, the first complete day after publication. The publication day is deliberately excluded from both retrospective validation and forward evaluation. Forward groups need complete 76-hour paths and enough observations before uncertainty can be assessed.
What if you require one or four closes, or change the confirmation threshold?
These BTC full-period sensitivity checks change only the confirmation flag: one or four closes, a zero or 0.2 ATR confirmation buffer, or 1.5× volume. The population still starts at the same four-hour T and later failure still uses the primary 0.1 ATR buffer. This is not a test of entering earlier, a different stop, or the best parameter.
| Condition | With: failures / N | Without: failures / N | Failure difference (pp) | 95% interval (pp) | Evidence |
|---|---|---|---|---|---|
| One close above R + 0.1 ATR | 113/211 · 53.6% | 10/22 · 45.5% | 8.1 | -13.1 to 30.1 | Descriptive, not causal |
| Four closes above R + 0.1 ATR | 91/185 · 49.2% | 32/48 · 66.7% | -17.5 | -32.3 to -1.2 | Descriptive, not causal |
| Two closes above R (no buffer) | 107/207 · 51.7% | 16/26 · 61.5% | -9.8 | -29.4 to 10.8 | Descriptive, not causal |
| Two closes above R + 0.2 ATR | 96/188 · 51.1% | 27/45 · 60.0% | -8.9 | -23.7 to 7.9 | Descriptive, not causal |
| At least 1.5× relative volume | 92/175 · 52.6% | 31/58 · 53.4% | -0.9 | -16.0 to 14.1 | Descriptive, not causal |
ETH replication — separate asset, same protocol
ETH is not pooled with BTC. A similar direction is a replication check; correlated crypto markets do not provide independent proof.
| Condition | With: failures / N | Without: failures / N | Failure difference (pp) | 95% interval (pp) | Evidence |
|---|---|---|---|---|---|
| Two closes above the level | 75/179 · 41.9% | 18/34 · 52.9% | -11.0 | -27.8 to 6.9 | Descriptive, not causal |
| Retest held within four hours | 23/45 · 51.1% | 70/168 · 41.7% | 9.4 | -7.9 to 26.3 | Descriptive, not causal |
| At least 2× relative volume | 66/146 · 45.2% | 27/67 · 40.3% | 4.9 | -9.4 to 19.2 | Descriptive, not causal |
| At least 55% aggressive-buy share | 42/100 · 42.0% | 51/113 · 45.1% | -3.1 | -16.5 to 9.8 | Descriptive, not causal |
| Two closes + 2× volume | 55/131 · 42.0% | 38/82 · 46.3% | -4.4 | -17.4 to 9.5 | Descriptive, not causal |
ETH later-period check
| Condition | With: failures / N | Without: failures / N | Failure difference (pp) | 95% interval (pp) | Evidence |
|---|---|---|---|---|---|
| Two closes above the level | 17/45 · 37.8% | 1/3 · 33.3% | 4.4 | Insufficient group size | Exploratory · unadjusted |
| Retest held within four hours | 5/9 · 55.6% | 13/39 · 33.3% | 22.2 | Insufficient group size | Exploratory · unadjusted |
| At least 2× relative volume | 13/37 · 35.1% | 5/11 · 45.5% | -10.3 | Insufficient group size | Exploratory · unadjusted |
| At least 55% aggressive-buy share | 7/22 · 31.8% | 11/26 · 42.3% | -10.5 | -35.6 to 14.9 | Descriptive, not causal |
| Two closes + 2× volume | 12/34 · 35.3% | 6/14 · 42.9% | -7.6 | Insufficient group size | Exploratory · unadjusted |
20 complete UTC days of resistance — a different question
This secondary level uses highs from 20 fully completed UTC days. Qualifying hourly closes above it are spaced by 76 hours; they are not necessarily distinct multi-day chart formations. The smaller samples limit conclusions.
| Condition | With: failures / N | Without: failures / N | Failure difference (pp) | 95% interval (pp) | Evidence |
|---|---|---|---|---|---|
| Two closes above the level | 19/54 · 35.2% | 4/8 · 50.0% | -14.8 | Insufficient group size | Exploratory · unadjusted |
| Retest held within four hours | 6/14 · 42.9% | 17/48 · 35.4% | 7.4 | Insufficient group size | Exploratory · unadjusted |
| At least 2× relative volume | 16/39 · 41.0% | 7/23 · 30.4% | 10.6 | -16.3 to 34.6 | Descriptive, not causal |
| At least 55% aggressive-buy share | 15/34 · 44.1% | 8/28 · 28.6% | 15.5 | -8.7 to 39.3 | Descriptive, not causal |
| Two closes + 2× volume | 14/37 · 37.8% | 9/25 · 36.0% | 1.8 | -25.7 to 26.0 | Descriptive, not causal |
Six historical paths — including the failures
These are actual hourly candles, not schematic drawings. For each named class we choose the unused event nearest that class’s median 24-hour return; ties use earliest event time. Class membership overlaps, so cases illustrate mechanisms rather than their frequency.
Confirmed then extended
Trigger: 2022-03-18 17:00:00 UTC · Reference resistance: $41072.96 · Selected from 91 eligible class observations.
Known by T: two closes held; retest not confirmed; volume 2.83×; aggressive-buy share 47.6%.
Observed later: 24h return 1.13%; 24h failure no; 72h reclaim after failure yes. These outcomes were not known at T.
High volume still failed
Trigger: 2022-07-18 06:00:00 UTC · Reference resistance: $21684.54 · Selected from 62 eligible class observations.
Known by T: two closes held; retest not confirmed; volume 2.91×; aggressive-buy share 53.2%.
Observed later: 24h return -1.51%; 24h failure yes; 72h reclaim after failure yes. These outcomes were not known at T.
A retest held
Trigger: 2023-06-16 16:00:00 UTC · Reference resistance: $25759.01 · Selected from 21 eligible class observations.
Known by T: two closes held; retest confirmed; volume 2.86×; aggressive-buy share 49.9%.
Observed later: 24h return 0.35%; 24h failure no; 72h reclaim after failure no. These outcomes were not known at T.
No qualifying retest, then extension
Trigger: 2023-12-13 16:00:00 UTC · Reference resistance: $41588.00 · Selected from 119 eligible class observations.
Known by T: two closes held; retest not confirmed; volume 2.37×; aggressive-buy share 54.4%.
Observed later: 24h return 0.84%; 24h failure no; 72h reclaim after failure no. These outcomes were not known at T.
Two closes held, then failed
Trigger: 2022-10-13 19:00:00 UTC · Reference resistance: $19238.31 · Selected from 103 eligible class observations.
Known by T: two closes held; retest not confirmed; volume 1.64×; aggressive-buy share 51.2%.
Observed later: 24h return -1.40%; 24h failure yes; 72h reclaim after failure yes. These outcomes were not known at T.
Failed, then reclaimed
Trigger: 2026-07-18 18:00:00 UTC · Reference resistance: $64387.99 · Selected from 111 eligible class observations.
Known by T: two closes held; retest not confirmed; volume 1.54×; aggressive-buy share 59.3%.
Observed later: 24h return -0.49%; 24h failure no; 72h reclaim after failure yes. These outcomes were not known at T.
Latest retained BTC observation — not a live trade instruction
The hourly status below is separate from the frozen daily study. It shows the latest event retained under 76-hour spacing, not every current breakout. A historical replay is not proof an alert was delivered at the time.
Observed through 2026-09-15 11:00:00 UTC · Hourly observation
Trigger 2026-09-14 17:00:00 UTC · R $78712.29 · Decision 2026-09-14 21:00:00 UTC.
Above resistance at T. Two closes: held; retest: confirmed. Subsequent 24h outcome: pending.
Coverage, missingness and reproducible evidence
The five-year inputs are reconstructed from historical Binance candles, not five years of locally recorded point-in-time signals. No funding, open interest or order-book history has been fabricated to fill this study. Missing hourly bars invalidate affected event windows; invalid aggressive-buy fields are unavailable, never zero selling pressure.
| Market | Expected / present 1h bars | Missing / duplicates | Valid buy-flow bars | UTC coverage end (exclusive) |
|---|---|---|---|---|
| BTCUSDT | 44160 / 44157 | 3 / 0 | 44156 | 2026-09-15 00:00:00 UTC |
| ETHUSDT | 44160 / 44157 | 3 / 0 | 44156 | 2026-09-15 00:00:00 UTC |
The first month before the research start is included in the input file for ATR warm-up. It is not included in these research-window coverage counts. The source manifest identifies archived partitions and checksums. Source file corrections can change a later release; cite the release ID, not only this moving page.
Release: 20260915-4c4f03983894
Input CSV SHA-256: 8875ea3b255387b25fed97f9864b0f2e63f6cd4b386b66f5972270c9d44d4241
Model: btc-breakout-confirmation-1.1
The input download is larger than the summary: it contains both assets and warm-up data. The page JSON / CSV also expose every retained event and excluded candidate. Run the public verifier against a saved frozen release; the script checks the input hash, all resistance candidates, spacing, confirmation flags and outcome summaries. Method rules below define the uncertainty calculation.
Observed confirmation groups are not randomized. Historical replay and reference price changes are not delivered alerts or executed returns.
How to read this page
- Define resistance before the break and use completed candles, not an unfinished wick.
- A held retest is a hypothesis to test—not proof of a safer trade.
- Compare later outcomes from the same four-hour decision point and show the discarded early failures.
- Keep the historical replay, frozen daily evidence and hourly observation timestamps separate.
What can this page tell you quickly?
- Primary question
- Which observable confirmation conditions accompany fewer later failures in BTC history?
- What qualifies
- An upside 1h close above prior 24h resistance; no volume prerequisite.
- Decision time
- Four hours after the trigger; later returns begin here, not four hours earlier.
- Evidence schedule
- Full historical study refreshes after each UTC day; latest retained BTC observation refreshes approximately hourly.
- Trading boundary
- No condition guarantees follow-through. A non-failure may still have a negative return.
What did the expanded technical experiment test?
The 1.1 extension adds ten condition-present versus condition-absent tests at the original four-hour decision point: RSI, strengthening MACD, expanding Bollinger bands, sustained quote turnover, ATR-normalized distance, ADX with DMI, EMA alignment, UTC VWAP, valid active-buy share and trigger upper wick. The calculation version is btc-breakout-confirmation-1.1; technical rules are breakout-technical-landmark-1.0. The original resistance, spacing, eligibility and outcome rules remain unchanged.
The first experiment tables show BTC counts, failure percentages and unadjusted calendar-block intervals, followed by ETH and retrospective later-period checks. All ten tests are exposed, not only favorable results. The downloadable technical_observations contain raw feature values and flags; technical_comparisons expose denominators, missingness, intervals and seeds. technical_distance_strata reports descriptive comparisons within four prior-ATR distance bins. These are not matched controls or a causal adjustment.
The clearest descriptive pattern is continuation of momentum and activity, but price-based indicators can simply encode distance from the same failure barrier. The study therefore does not claim an independent predictive edge or a combined success score. The extension was developed after exploratory inspection, and the later historical period was not untouched. Additional funding/OI and four-hour-trend experiments are outside this extension's completed evidence scope.
Verify: Frozen rules and technical observations · Technical verification example
How should you check a Bitcoin breakout in practice?
Start with a level you could have drawn before the move. In the primary test that means the highest high in the 24 completed hours preceding the trigger candle. Do not move resistance upward after the price rises or downward after it fails. A different level—weekly resistance, an all-time high or a discretionary trend line—requires its own evidence and cannot inherit these percentages.
First wait for the hourly close. Next ask whether the next two closes remain clearly above the original level. A small buffer, 10% of prior hourly ATR(14), distinguishes a marginal touch from a sustained close. ATR is calculated before the trigger, so the breakout candle cannot inflate its own tolerance. This is a fixed research convention, not an optimized trading stop.
If price comes back, distinguish a shallow retest from a breakdown. Our retest is deliberately strict: an hourly low enters the frozen buffer band in the first three hours, and that candle plus the following candle close above the upper band by hour four. A deeper dip, a recovery after six hours or a wick without closing confirmation does not qualify. It may still lead to a profitable move, but it is a different event.
Finally check when the idea has failed. Two consecutive hourly closes below the lower band count as failure. One close inside the old range is a warning recorded separately, not automatically the same outcome. If price subsequently reclaims the level, report both facts rather than rewriting the failed observation as an uninterrupted success.
Does high volume or buying pressure confirm a real breakout?
Volume tells you how much traded, not that every participant agreed with the direction. We divide the trigger hour’s USDT quote turnover by the average of the preceding 24 hours and test a fixed 2× threshold. The candidate pool itself does not require high volume. Otherwise we would remove the low-volume comparison before asking whether volume helped.
Aggressive-buy share uses taker-buy quote volume divided by total quote volume, when both are valid. A 55% reading means 55% of that candle’s measured quote turnover was buyer-initiated; it is not 55% more buyers, net wallet deposits or proof of institutional accumulation. Invalid source fields remain missing and are excluded from that specific comparison, not silently assigned to the low-buy group.
A large candle can include both genuine demand and late chasing. Strong volume may accompany a failed break as well as an extension. Read the high-volume failure example alongside the aggregate counts. This first version does not claim 2× volume or 55% buying is optimal; different thresholds, timeframes, exchange coverage and market regimes require separately reported sensitivity tests.
Verify: Binance Kline field definitions
Why compare every condition four hours after the break?
A retest cannot be known at the original breakout close. If an analysis labels a move as confirmed using the next four hours and then credits it with the return during those same hours, it is using future information. Our common decision point T is the fourth hourly close after the trigger. All reported later returns start from that price.
This answers a specific question: among paths still above resistance and without an early confirmed failure at T, does an already observed condition accompany different subsequent outcomes? It does not directly answer whether entering immediately outperforms waiting. We show the excluded early failures and the change in price during the wait so this conditioning is visible.
For a fair side-by-side horizon comparison the primary pool requires all 72 hours after T, even though the principal failure outcome is measured over 24 hours. A newly completed 24-hour result can therefore remain outside the common comparison until the longer path matures. The row-level export still exposes each horizon’s own pending, gap or complete status.
The groups are condition-present and condition-absent observations, not randomized treatments or a matched control design. A retest group is selected partly because price returned toward resistance; those paths can already be weaker. Lower or higher failure rates are associations that must not be translated into the causal benefit of waiting for a retest.
How much historical evidence is enough to trust the pattern?
One thousand candles are not one thousand independent breakouts. This study retains the first qualifying close, then suppresses other qualifying closes for 76 hours. Every suppressed candidate keeps its reason and blocking observation ID in the export. Different definitions and assets are processed separately. For 20-day resistance, repeated qualifying closes may belong to the same broader chart formation, so even spacing does not make them perfectly independent.
Market events cluster in time. The primary uncertainty calculation resamples seven-day blocks of the full UTC calendar, including days with no events. The same draw is used for the condition-present and condition-absent group so their difference is evaluated jointly. Three- and fourteen-day blocks are sensitivity checks. At least 20 usable observations in each group and 1,900 valid draws out of 2,000 are required for a displayed interval.
The data are split chronologically at September 2025, and any event whose complete 76-hour path crosses a boundary is excluded from that period. This protocol was first specified in September 2026; the earlier split is a retrospective check, not a claim of prior preregistration. Forward observations accumulate after the protocol date and should be assessed separately once mature.
Inspect the yearly rows and the ETH replication instead of selecting whichever filter has the best full-period number. Intervals crossing zero leave the direction of the difference unresolved under the resampling model. Even an interval excluding zero is not a guarantee: this is a small family of observational comparisons without multiple-testing-adjusted significance claims, and structural market changes remain possible.
Verify: Full study contract and release
Use the research to ask better questions—not to automate a promise
For an individual trader, the useful output is a repeatable checklist: identify the original level, wait for completed observations, separate strong activity from proof, and define what invalidates the idea. CoinNudge’s Telegram service can surface its supported price and technical conditions so you do not have to watch every chart. Check the current alert list before subscribing. This research’s exact four-hour retest protocol is not a promise that an identical Telegram alert is currently delivered.
For a researcher, the value is the reproducible chain from hourly inputs to resistance, candidate exclusion, observable confirmation and later outcomes. Download a frozen release and the matching hashed inputs before testing a different threshold or entry rule. Keep outcome fields out of predictive inputs and distinguish reconstructed historical events from genuinely locally observed PIT records.
Public study downloads remain available independently of a data subscription. Paid data products have their own catalog, coverage and delivery terms; do not assume this article includes an unrestricted historical data license or that every study variable is present in every paid package. Review the sample and documented API scope before purchasing.
Verify: See the Telegram alert list · Compare subscriptions · Research data products · Inspect data samples
What is measured, and what is not?
| Measured claim | Evidence on this page | Boundary |
|---|---|---|
| A close above resistance does not certify a successful breakout. | The public candidate ledger, later failures and six deterministically selected historical paths. | Applies to the stated hourly Binance spot definitions; not all technical chart patterns. |
| A confirmation condition can be evaluated without crediting it with earlier returns. | Every main comparison uses T=t0+4h and reports early exclusions and waiting price changes. | Conditioning on survival changes the population; the comparison is observational. |
| The published results can be reconstructed from identifiable inputs. | Frozen JSON release, full hourly input CSV, SHA-256 manifest and public verification script. | Historical replay is not proof of five years of live signal delivery or original point-in-time capture. |
Take the next step with this evidence.
Reading about how to confirm a bitcoin breakout and spot a fakeout? Choose ongoing notifications or a dataset you can inspect in your own research workflow.
Receive qualifying events in Telegram.
Use the ready-made alert checklist, message explanations and daily briefing. Check the supported categories and coins before subscribing; every Research page is not a separate alert category.
Check the alert list · View alert plans
Inspect the data before choosing a plan.
Market Events v1 connects recorded events and comparison rows with context, later outcomes and quality flags. Paid access provides customer API keys and daily JSON, CSV and Parquet files.
Check Market Events sample · Explore the full data catalog · Compare data plans
Researcher and Research Desk currently include Market Events v1 only, for its specified pairs and published history. This page's full dataset is not automatically included. Other data families require coverage, date, field and source-use confirmation before purchase. Telegram Pro and data access are separate subscriptions.
Method and data boundary
btc-breakout-confirmation-1.1. Binance BTCUSDT spot 1h closes above the preceding 24 completed hourly highs; secondary resistance is the preceding 20 complete UTC daily highs. Freeze R and prior Wilder ATR(14), using a 0.1 ATR buffer. Space retained events by 76 hours. Two-close confirmation uses hours 1 and 2; a retest requires a low within R ± buffer at hour 1..3 and that close plus the next close above R+buffer by hour 4. All comparisons start at T=t0+4h, only if T close > R and no early pair of consecutive closes <R-buffer. Later failure is two consecutive closes <R-buffer strictly after T. Compare condition present versus absent, not matched or causal controls. Primary outcome T+24h; additional T+4h/72h paths in exports. Require the complete 76h path for the common comparison pool. Calendar-day circular moving-block bootstrap includes empty days: 2000 shared resamples, 7-day blocks with 3/14-day sensitivity. Minimum 20 per group; linear 2.5/97.5 percentiles. xorshift32; first four big-endian SHA256 bytes of version|symbol|definition|period|block_size seed (zero becomes one); block starts use state modulo calendar-day count. No multiple-testing-adjusted significance claims. Historical chronological split at 2025-09-01; first protocol date 2026-09-15; forward observations thereafter; reject 76h paths crossing boundaries. Missing bars and pending outcomes are excluded with reasons, never imputed. Extension 1.1 adds ten exploratory hourly technical conditions at T; full row-level features, BTC/ETH chronological summaries and prior-ATR distance strata are downloadable. Technical seed is SHA256 first4 big-endian of breakout-technical-landmark-1.0|symbol|period|7. The existing 7-day 2000-draw interval and minimum-group rules apply; no multiple-test adjustment or independent predictive-edge claim. Base volume is included for complete UTC-session quote/base VWAP; see contract.technical_extension for exact formulas.
This is an observational historical replay of upside Binance spot breaks, not prospective PIT history or a profitable trading-system claim. Conditioning on survival to T changes the population. Retest and non-retest paths are not randomized or matched. The 2025–2026 split is retrospective, not proof that its data was never previously inspected. No fees, slippage, funding or execution simulation is included.
Sources and verification
- Binance Spot KlinesHourly OHLC, quote turnover and taker-buy quote fields.
- Binance Public DataOfficial historical archive formats and SHA-256 checksum convention.
- CoinNudge frozen research contractOwn calculations, model version, daily release and full observation ledger.
- CoinNudge methodologySite-wide missingness, provenance and research boundaries.
Frequently asked questions
Does RSI above 70 mean a Bitcoin breakout is false?
Not by itself. In this historical breakout sample, RSI at or above 70 at the four-hour decision point accompanied fewer later failures than RSI below 70. Strong RSI can describe continuing momentum rather than an immediate reversal. This is an exploratory association, partly confounded by distance above resistance; recent groups are small. It is neither an automatic sell signal nor a reason to chase a trade.
What is the difference between breakout volume and sustained volume?
The trigger-volume test compares the breakout hour's quote turnover with the mean of the preceding 24 hours. The sustained-volume test compares mean hourly quote turnover in the NEXT four hours with that SAME pre-trigger mean; both use a 2x threshold. The second condition is only known four hours later. Its later outcomes start after that decision point, so the waiting period is not counted as a subsequent gain.
Why is a lower breakout failure rate not a higher trading win rate?
Failure here means two consecutive hourly closes below frozen resistance minus 0.1 prior ATR during the 24 hours after the four-hour decision point. A price can avoid that barrier and still finish below an entry price; a failed breakout can later recover. Entry, exit, fees, slippage and position sizing are not simulated. One minus the failure rate therefore is not a profitable-trade probability.
How can I tell if a Bitcoin breakout is real?
You cannot know its future with certainty. Define resistance before the move, require completed closes, inspect whether it holds or retests under a fixed rule, and specify a failure condition. This study measures later failure under those rules rather than assigning a guaranteed true-breakout label.
Does a retest make a Bitcoin breakout safer?
Not automatically. The study compares a narrowly defined retest completed within four hours with otherwise eligible paths lacking that retest. The dynamic table shows failure counts, denominators and uncertainty; retested paths are not randomized and can be weaker to begin with.
How many candle closes confirm a breakout?
The principal hold condition requires the first two hourly closes after the trigger above frozen resistance plus 0.1 prior ATR. This is a testable convention, not evidence that two is universally optimal. All conditions are compared at the same four-hour decision point.
What is a false breakout in this study?
Two consecutive later hourly closes below the original resistance minus 0.1 prior ATR. One close back inside and a later reclaim are separate fields. Failure rates therefore cannot be compared with another study unless its threshold and horizon match.
Do I need high volume for a breakout?
High volume is tested as an additional condition, not required to enter the candidate pool. The fixed threshold is trigger quote turnover at least twice the preceding 24-hour mean. Large volume can accompany failures too; the study publishes both sides.
Are these five years of point-in-time trading signals?
No. Earlier events are reconstructed from historical hourly candles under the published rule. That is distinct from records actually observed and stored by CoinNudge at the time. No historical alert-delivery or executable-profit claim is made.
Can I download the full data and reproduce the result?
Yes. The page links current JSON/CSV evidence, a frozen research JSON release, the corresponding hourly input CSV with SHA-256, and a verification script. Record the release and calculation version so later updates do not silently change your citation.
How often does this study update?
The full historical comparison is materialized after each UTC day by the existing archive worker. The latest retained BTC observation refreshes approximately hourly and has a separate timestamp and stale flag. Missing or pending outcomes remain explicit.
Turn the evidence into a better monitoring checklist
See which price and technical conditions CoinNudge can deliver to Telegram. Compare coverage before subscribing; a signal is not a promised return.
Explore the alert list