Nobody publishes the failures. Every strategy you see posted worked. The ones that didn't were quietly deleted. That's not evidence, it's a filter. 9 tested. 8 dead. NR-009 is live. Every trade posted before the outcome. Null hypothesis: there is no edge.

Sat through another "brace for the $15B options expiry" week. So I went and checked what those days actually do. Every quarterly expiry since 2019, 29 of them. On the day itself Bitcoin moves about half a percent. A normal day moves 1.4%. Even a normal Friday moves more. 22 of the 29 were calmer than an average day. The only big one was 25 December 2020, up 7.3%. That was Christmas with nobody at the desk, not the options. How I counted: last Friday of March, June, September, December. Move from midnight to midnight, up or down, then the median so one crazy day can't skew it. Coin Metrics closes. Rules set before I opened the data. What this doesn't cover: the hour around settlement at 08:00 UTC. If you got pinned there, that's real and it won't show up in daily closes. Genuine question: does anyone actually remember an expiry Friday that hurt? C-006

There's a price where option buyers lose the most. People say price gets pulled toward it. The last three quarterly expiries settled 5 to 17% below it. Friday settled 10.4% above. Four observations. The pull just reversed.

$351 million walked out of Bitget last night. But nobody stole a key. You read that right. If you keep money on any exchange, the next ten lines are about you, not about Bitget. Here's what happened, in order. At 18:31 UTC on Thursday, Bitget's systems flagged transfers leaving its hot wallets that nobody had ordered. By the time they counted, $351.6 million was gone. Largest exchange hack of 2026. An exchange keeps most money in cold wallets, offline, like a safe. A small part sits in hot wallets, online, so withdrawals are fast. The hot wallets got hit. The safe didn't. How, in plain words: the thieves didn't pick the lock. They broke into the system that writes the transfer requests, faked the paperwork, and Bitget's own approval process signed off and sent the money. Bitget says no private keys were taken. That's true, and it's the scary part. The guard opened the door because the note looked right. Now the sentence every exchange uses: "User funds are safe." Three numbers nobody attached to that sentence. 1. Safe means 76% of the safety net. Bitget's User Protection Fund holds "over $464 million" and will cover the whole loss. $351.6M is 76% of it. One hack and three quarters of the net is spent. A second one this size, and the sentence changes. 2. Safe doesn't mean available. Your balance is correct on screen. Withdrawals are paused, deposits and trading are open, and there is no reopening date. Bitget said it won't promise a deadline it can't guarantee. So "safe" today means: the number is right, and you can't have it yet. 3. September is now the most expensive month of 2026. Liquid Network lost $320M earlier this month. Add Bitget and you're near $672M in one month. For scale: Bybit, February 2025, was about $1.5 billion in one night. Bitget is a quarter of that and still the biggest of this year. One more. Of the 102.97 million XRP taken, more than 99% is still sitting in the attacker's wallet, untouched. That's the number to watch. When it starts moving, so does the recovery odds. Bitget suspects North Korea, based on IP clues. Suspects. A full incident report was promised within 24 hours of the breach. That deadline is [invullen: passed / due at 18:31 UTC today]. So this was not a hack that stole your money. It was a hack that showed you what "your money" on an exchange actually means: a promise, a fund behind the promise, and a withdrawal button someone else controls. What this gives you for the next "funds are safe" headline: Find the fund. Every big exchange publishes a protection fund. Divide the hack by the fund. Under 50%, the promise holds twice. Over 75%, it holds once. Read "safe" as "correct", not "available". Ask when withdrawals reopen. If the answer is a date, note it. If the answer is a review, note that too. Watch the stolen wallet, not the press release. Stolen coins that don't move can still be frozen or traced. Coins that move through a mixer are gone. Registration: Bitget withdrawals reopen within 72 hours of the breach, so before 18:31 UTC Sunday. Scored Monday as C-006. Reply with the exchange where you keep the most, and I'll tell you what its protection fund covers, in hours, in dollars, for every one asked before midnight CET.

Everyone said Bitcoin would get pulled down to $76,000 this Friday. It settled at $83,931. $76,000 was the "max pain" level for Friday's quarterly Bitcoin options expiry on Deribit, about $16 billion in contracts. Max pain is the price where the most options expire worthless. The theory: big option sellers push the price toward it before settlement. Deribit's settlement price: $83,930.84. That's 10.4% above max pain. The three quarterly expiries before this one all traded below max pain into settlement: Dec 2025: max pain $96,000, BTC ~$88,600 Mar 2026: max pain $75,000, BTC ~$71,000 Jun 2026: max pain $70–72,000, BTC ~$60,000 Four quarters. None settled at max pain. The part the "price magnet" headlines skip: max pain for this expiry was $70,000 on August 22, $72,000 in mid-September, $76,000 this week. It moved toward the price. Not the other way around. Before you trade the next max pain headline, check the date the number was pulled. It changes every day. Settlement: Deribit delivery price, 30-minute average of the BTC index before 08:00 UTC. Max pain: Deribit.

Price fell 2.1% on 23 September. Open interest on Binance BTCUSDT fell 7.7% over the same day. Measured in BTC, not dollars, so the price drop isn't doing the work. Positions closed faster than price moved. Somebody was forced.

Bitcoin dropped 2% yesterday, days after an 8-month high. "Buy the dip" is everywhere. I counted every dip like it since 2011. 161 of them. Rules, written down before the count: New high: a close above the highest close of the past 240 days. Dip: a close 2% or more below the day before, within 5 days of that high. Test: is the close 30 days later higher than on the dip day? Data: Coin Metrics daily closes, Jan 2011 to May 2026. Dips after a high: 100 of 161 higher a month later. 62%. Any random day: 57%. Any other 2% red day: 50%. A coin flip. 62 against 57 sounds like an edge. With 161 cases the true rate sits between 54% and 69%, and 57 is inside that. So: cannot be settled. First time this series ends there. What the count does show: a red day by itself is worth nothing. A red day next to a fresh high does a bit better. The information is in the high, not in the dip. 37 of the 161 come from 2017 alone. Strip that year and the edge gets thinner. If someone sells you "buy the dip": ask which dips, and what the other dips did. C-005

Bitcoin buyers put in $2.65 billion this week. The gamblers left the same day. If you bought Tuesday's dip and got stopped out Wednesday, this is why. And it's not bad news. Here's what happened, in order. Bitcoin closed Wednesday at $84,397. Down 2.1% on the day, still up 10.75% on the week. Two groups trade Bitcoin. Spot buyers pay full price and hold the coin. Leverage traders borrow to bet on the price and get thrown out when it moves against them. Wednesday, the two groups went in opposite directions. Spot kept buying. US Bitcoin ETFs took in $347M net on Wednesday (SoSoValue). That's five days in a row: $159M, $433M, $999M, $715M, $347M. About $2.65 billion in a week. Leverage left. Open interest, the total of all open leveraged bets on Binance, fell 7.7% in one day to 98,503 BTC. Funding, the fee longs pay shorts, sits at 0.0057% per 8 hours. Positive, but well under the 0.01% neutral rate. Longs aren't paying up to stay in. So the price fell 2% on a day when the holders bought and the borrowers were leaving. That is what a leverage flush looks like when spot is underneath it. Three numbers nobody attached to the headline. 1. The year flipped. Yesterday I wrote the spot ETFs were still net negative for 2026. That's outdated. Bloomberg has them at roughly +$320M for the year, after about $4.6B came in since August 19. Correction posted, not deleted. 2. Leverage is growing in dollars and shrinking in Bitcoin. Aggregated open interest across all exchanges: $61.6B, +7.5% in 30 days (CoinStats, Sep 22). Binance open interest in Bitcoin: 98,503 BTC, -8.1% in 30 days. Different scope, not a clean comparison. But when the dollar number rises while the coin number falls, price did the work, not new bets. 3. Every other market says this shouldn't be happening. The Fed hiked on September 17, first time in three years. Japan's 10-year yield is above 3%, highest since 1996. The US 10-year is at 5.1%. Stocks topped in mid-August. Gold is struggling. Benjamin Cowen had the whole picture right and gave a 35% chance the Bitcoin low was in. Then Bitcoin broke the May high anyway. His words: "Markets don't have to make sense." His line: $83,000. Above it on a weekly close, the low is in. Below it, his Q4 pullback is back on. Bitcoin is trading at $83,470 as I write this. So this was not a rally losing steam. It was a rally losing its leverage while keeping its buyers, on a week when everything else said sell. What this gives you for the next red day inside a green week: 1. Split the day into spot and leverage. ETF flow plus open interest, same date. One up, one down means a flush, not a reversal. 2. Read funding as a price tag, not a signal. Under 0.01% means longs are getting paid to wait. Over 0.03% means they're paying to stay, and that's when a squeeze starts looking for them. 3. Count open interest in coins, not dollars. Dollars rise when price rises. Coins only rise when new bets do. Reply with what you did Wednesday. Bought, sold, stopped out, or sat still. One word. I'll tell you which of the three checks would have changed it, for every reply before midnight CET.

$72,000 is Friday's max pain. Spot: $81,290, 13% above. You've read "the magnet" five times since December. Five times it settled 2–17% below max pain. None got pulled up. Friday is the first with spot above.

"The crypto winter is over." You'll hear it a lot this week. I tested what that sentence actually rests on: Bitcoin just made its first 8-month high in a long time. Does that signal mean anything? Rules, written down before the count: Signal: the first close above the highest close of the past 240 days, after at least 120 days without one. Test: is the close 90 days later higher? And does the price never close 20% below the signal day in between? Data: Coin Metrics daily closes, Jan 2011 to May 2026. 12 signals. 11 were higher 90 days later. 11 never lost 20%. With 12 cases the true rate sits between 65% and 99%. Pick any other day at random: higher 90 days later 60% of the time. So yes. This one holds. First claim in this series that does. Now the small print. The one miss was October 2021: the breakout to $69k, then the crash. May 2019 counts as a hit, and then fell 50% by March 2020. "Bear market over" and "higher 90 days later" are not the same sentence. This month's signal isn't in the data yet. It gets tested on 21 December. Next time a claim holds, check what window it was tested on. C-004

$70k, $72k or $75k: three max pains for one Friday expiry. You've traded on "price gets pulled to max pain." Pulled to which one? Max pain: the close where option buyers, calls and puts alike, lose most. Friday 08:00 UTC we post the distance to each.

NR-009 · TARGET HIT · +1.63R One winning trade proves exactly as much as one losing trade. Nothing. You still screenshot the wins. ASTERUSDT · 1D · SHORT Registered 19-9-2026, before the outcome. Entry 0.7673 Stop 0.825 Target 0.673 → hit Result +1.63R (12.3% move) What this trade does not show: that head and shoulders works. A single result can't separate an edge from a lucky draw. That takes a sample. It was also the first trade under the amended scope (any pair above $1M 24h volume, any timeframe). The amendment went out before the outcome, and this win doesn't validate it either. Both get judged at the same checkpoint: 30 trades. NR-009 so far: 1 trade, 1.63R cumulative. Kill: −10R cumulative, or 6 losses in a row. Before you screenshot your next win, write down how many times that setup has actually been traded. Under 30, it's a mood, not evidence.

Bitcoin ETFs took in $999 million on Monday. It wasn't Wall Street. If you read "biggest ETF day of 2026" and felt late, keep reading. It was people like you, and they didn't buy where you think. Here's what happened, in order. An ETF is a fund that buys Bitcoin for people who don't want to hold it themselves. On Monday, US Bitcoin ETFs took in $999 million net. About 11,530 Bitcoin. Biggest dollar day since October 2025, biggest Bitcoin day since November 2024. The reflex: "The big money is back. If I don't buy now, I'm buying behind them." Three numbers nobody attached to the headline. 1. It was retail. Bloomberg's Eric Balchunas read the pattern: uneven, jagged inflows spread across six funds. One big player buying would show up as a straight line. Jagged means thousands of small buyers. The "institutions are back" headline is describing you. 2. They didn't buy the top. When you order ETF shares, the fund usually reports the purchase a day later. Balchunas puts most of Monday's $999M as orders placed Friday the 18th, when Bitcoin went from $76,400 to $80,900. Not at Monday's $87,200 peak. The headline said "chasing". The booking says they bought the dip and the headline showed up three days later. 3. The crowd just got its money back. Count everyone who bought through these ETFs since launch. Above $86,000 that group is, on average, in profit for the first time since January. That breakeven line sat at $89,600 in November 2025 and dropped to $79,900 by March as people kept buying lower. Every new dollar at $86K moves it back up. Monday's buyers didn't lift the floor. They moved the line where the crowd stops losing money. And the year is still red. Add every day of 2026 together and the spot ETFs are still net negative for the year. One record day did not get 2026 back to zero. So this was not big money chasing the rally. It was small money closing a hole it dug in the winter. What this gives you for the next "record inflow" headline: 1. Check the year-to-date line before the day line. A record day inside a negative year is a recovery, not a wave. 2. Subtract one trading day. The flow you read about today was bought yesterday or earlier. If price already moved, the headline is late by design. 3. Find the breakeven. When the average holder crosses it, you're not early to a trend. You're standing next to a crowd that just got its money back and is deciding what to do with it. Reply with the ETF flow headline you almost traded on and the date. I'll tell you what the booked price actually was, for every one asked before midnight CET.

$14.7B in Bitcoin options expires Friday. You read it as price getting pulled to $72k. Coinglass says $70k. Binance says $75k. Same expiry, $5,000 apart. "Max pain" is the price where most option buyers lose. Which one is yours?

$14.1B in Bitcoin options expires Friday. You read that as $14.1B in play. At $76,000, $3.2B of it pays anything. The other $10.9B pays nothing at this price. Notional counts every contract at full size, whatever it's worth. Divide the headline first.

Bitcoin made $454 million disappear yesterday. And the price went UP. If you shorted the top of the range last week, some of that was yours. And most of you are about to make the mirror mistake. Bitcoin went from $81,500 to $85,200 in two hours on Monday, then printed $86,000 the same day. Highest since January. A short is a bet that price falls. When price rises fast, the exchange force-closes the bet. A forced close is a buy. Buys push price higher. Higher price closes more shorts. The first push needs buyers. The loop after that doesn't. This was not a rally. It was an eviction. Proof: $454M in BTC shorts liquidated vs $53M in longs (Coinglass, Sep 21, 24h). 8.5 to 1. Glassnode reads it the same way: Monday's leg was driven by short liquidations, not by sellers giving up. Across all of crypto the total was $782M shorts vs $144M longs. What changed under the hood: aggregated BTC futures open interest is up 7.5% in 30 days to $61.6B, funding is positive again, and Santiment has Bitcoin FOMO at its highest since 2024. Leverage didn't leave the market. It switched sides. The 3 checks before you chase any breakout, in this order: 1. Long/short liquidation split. If shorts dwarf longs, the buying was forced, not chosen. The line I use is 5:1. Untested. I'm counting it from today. 2. Who bought: spot or perps. Perps only means borrowed conviction. 3. Did open interest rise with price. Yes means the next squeeze is already loaded, and it's on the long side now. Drop the coin you're about to chase and I'll tell you which of the three checks it fails. Every one asked before midnight CET gets an answer.

Everyone says September is Bitcoin's worst month. I counted all fifteen. It isn't. August is. Rules, written down before the count: A month is red when it ends below where the previous month ended. Worst month: most red years and the lowest median return. Data: Coin Metrics daily closes, Jan 2011 to Dec 2025. 180 months. September: red 9 of 15, median −2.9%. August: red 10 of 15, median −8.4%. Any month at random: red 43% of the time, median +4.4%. September is redder than average. Nine of fifteen. But with fifteen cases the true rate sits anywhere between 36% and 80%, and the "average −2.9%" leans on 2011, when Bitcoin fell 37% from five dollars. Why you've heard it anyway: the Yahoo Finance count starts in 2013 and gets 8 of 13. Different start year, different story. Same data. Not covered: exchange monthly candles close at 23:59 UTC, Coin Metrics prices are taken at 00:00. One day of difference can flip a single month. Next time someone names a "worst month": ask which year they started counting. C-003

3.25%: where the Fed now says rates will settle. You read that as a fixed place. In 2012 it said 4.25%. From 2019 to 2023, 2.5%. Raised five times since 2024. If your plan says "cuts back to neutral", write down which one. The number moved while you waited.

52% on a second hike in October. You read Wednesday as one and done. Six first hikes since 1990. Four were followed at the very next meeting. The Fed's own dot says one more this year. Two meetings left. 52% is CME FedWatch, 17 Sep. Scored 28 Oct.

A golden cross just printed on Bitcoin. You're being told a 10% drop always follows. I counted all 12 since 2012. Then I counted every other day. The claim, reported by Benzinga on 16 Sep: every golden cross in Bitcoin's history was followed by a correction of at least 10%. Rules, written down before the count: Golden cross: the 50-day average of daily closes finishes above the 200-day. Correction: a close at least 10% below the highest close since the cross. Window: 30 days. The two examples given for the claim ran 3 to 4 days and about 10 days. Data: Coin Metrics daily closes, 18 Jul 2010 to 23 May 2026. 12 crosses. 8 corrected within 30 days. 4 did not: May 2020, Oct 2023, Oct 2024, May 2025. With 12 cases the range is 39% to 86%. Now the number you don't get told. Start the same 30-day clock on every other day in the data: 3,480 of 5,548 saw the same drop. 63%. Give it 90 days: 11 of 12 crosses, 92%. Every other day: 92%. Bitcoin drops 10% within three months of almost any day you pick. The cross adds nothing to that. Not covered: intraday wicks. Three of the four misses were close on closes: 9.0%, 8.9%, 7.1%. With wicks some may pass 10%. The base rate rises with them. Next time you hear "every time X, then Y": ask how often Y happens without X. C-002