SifuBacktest

1/8 When should you switch off a trading strategy that keeps losing? Horizon's framework on X gives two fixed thresholds: cut the position in half at the first, pause the strategy at the second. Simulated, those rules flag good strategies almost as often as bad ones. 🧵 #Trading #Backtesting

7/8 What helps instead: → deflate the backtest Sharpe for every variant you tried → few checks, planned in advance, with a known false-alarm rate → judge against buy and hold, not the raw curve → validate new parameters only on new data

1/8 When should you switch off a trading strategy that keeps losing? Horizon's framework on X gives two fixed thresholds: cut the position in half at the first, pause the strategy at the second. Simulated, those rules flag good strategies almost as often as bad ones. 🧵 #Trading #Backtesting

5/8 Horizon on overfitted strategies: "These die fast, usually within weeks." They do not. A strategy with no edge loses slowly, and in year one its drawdowns look like a healthy one's. Even a clean significance test spots it after a year only 22 % of the time.

1/8 When should you switch off a trading strategy that keeps losing? Horizon's framework on X gives two fixed thresholds: cut the position in half at the first, pause the strategy at the second. Simulated, those rules flag good strategies almost as often as bad ones. 🧵 #Trading #Backtesting

3/8 5,000 strategies, 3-year backtest at Sharpe 1.0, then live: → intact, 3 years half size 81 %, paused 33 % → edge halved, 3 years half size 90 %, paused 51 % → no edge, 1 year half size 72 %, paused 18 % The rules react to everything a little and to nothing in particular.

1/8 When should you switch off a trading strategy that keeps losing? Horizon's framework on X gives two fixed thresholds: cut the position in half at the first, pause the strategy at the second. Simulated, those rules flag good strategies almost as often as bad ones. 🧵 #Trading #Backtesting

1/6 On Monday bitcoin seemed to be drifting toward max pain at 78,000 ahead of Friday's $10.9bn options expiry. Two days later it has turned away. The gap grew from 2.5 % to 4.5 %. The strike stayed put, and the price moved up. 🧵 #bitcoin #options

5/6 Where the contracts sit (02:00 UTC, bitcoin at 80,835): → calls 10 % or more above 68 % of calls → largest band, 95.0–97.0k 22,058 calls → puts 10 % or more below 56 % of puts Most of the book only pays out if bitcoin moves 10 % or more by Friday.

1/6 On Monday bitcoin seemed to be drifting toward max pain at 78,000 ahead of Friday's $10.9bn options expiry. Two days later it has turned away. The gap grew from 2.5 % to 4.5 %. The strike stayed put, and the price moved up. 🧵 #bitcoin #options

3/6 Max pain has now read 78,000 in nine daily snapshots in a row, since 20 Oct. A pull toward the strike would show up as a gap that keeps shrinking into Friday. It shrank from Saturday to Monday and has widened again since.

1/6 On Monday bitcoin seemed to be drifting toward max pain at 78,000 ahead of Friday's $10.9bn options expiry. Two days later it has turned away. The gap grew from 2.5 % to 4.5 %. The strike stayed put, and the price moved up. 🧵 #bitcoin #options

1/6 $166 billion flowed into money market funds in one week of October, the most since April 2020. Is that cash waiting to buy stocks? 🧵 #Markets #Stocks

5/6 Much of it did not stay: → week to Oct 14: −$40.5B → week to Oct 21: +$21.5B, to $7.94T Hartnett's line on Oct 9: "No rate cuts, no cash cuts."

1/6 $166 billion flowed into money market funds in one week of October, the most since April 2020. Is that cash waiting to buy stocks? 🧵 #Markets #Stocks

3/6 US money market funds took in $72.3B that week, to $7.96T (ICI). → institutional funds: +$58.3B → retail funds: +$14.0B Four fifths came from institutions, mostly into government funds.

1/6 $166 billion flowed into money market funds in one week of October, the most since April 2020. Is that cash waiting to buy stocks? 🧵 #Markets #Stocks

1/8 Could you have seen the biggest crypto scams coming? In 9 of the 15 biggest cases, yes. The warning sign was in the offer itself, before anyone paid in: a fixed return or money for recruiting others. 🧵 #Crypto #Bitcoin

7/8 Today the fraud finds you. Crypto losses reported to the FBI in the US: → 2020 $246m → 2025 $11.37bn The largest item is investment fraud that starts with a chat contact. Of the people the FBI warned in 2025, 78% did not know they were being scammed.

1/8 Could you have seen the biggest crypto scams coming? In 9 of the 15 biggest cases, yes. The warning sign was in the offer itself, before anyone paid in: a fixed return or money for recruiting others. 🧵 #Crypto #Bitcoin

5/8 At an exchange, the warning sits in the balance sheet. On 2 November 2022 CoinDesk published Alameda's: 5.8 of its 14.6bn dollars in assets were FTT, the token FTX issued itself. Nine days later FTX was bankrupt.

1/8 Could you have seen the biggest crypto scams coming? In 9 of the 15 biggest cases, yes. The warning sign was in the offer itself, before anyone paid in: a fixed return or money for recruiting others. 🧵 #Crypto #Bitcoin

3/8 What the 15 biggest cases had in common: → fixed return or pay for recruiting 9 of 15 → no way to check the money was there 14 of 15 → a regulator warned beforehand 5 of 15 You did not need to read code. The warning was in the advertising.

1/8 Could you have seen the biggest crypto scams coming? In 9 of the 15 biggest cases, yes. The warning sign was in the offer itself, before anyone paid in: a fixed return or money for recruiting others. 🧵 #Crypto #Bitcoin

1/9 What is Zcash, and why does its privacy cost ten times as much as Monero's? ZEC went from $74 to a record close of $1,653 in a year and became the tenth-largest cryptocurrency. Privacy explains the rise. It does not explain the price level. #Zcash #Privacy

8/9 "The ETF is only a small part. Binance alone trades $316m of ZEC a day." That is true, but nothing is known about those buyers: not how long they hold, nor at what price they bought. What is known is that the visible buyers stopped and a visible seller sells every day.

1/9 What is Zcash, and why does its privacy cost ten times as much as Monero's? ZEC went from $74 to a record close of $1,653 in a year and became the tenth-largest cryptocurrency. Privacy explains the rise. It does not explain the price level. #Zcash #Privacy

6/9 On the other side is a seller with a name. Fortitude Mining, part of DCG like Grayscale, mined about 21.5 % of new ZEC this year. Its merger filing says: "Fortitude currently sells or otherwise monetizes all the digital assets that it mines, including ZEC."

1/9 What is Zcash, and why does its privacy cost ten times as much as Monero's? ZEC went from $74 to a record close of $1,653 in a year and became the tenth-largest cryptocurrency. Privacy explains the rise. It does not explain the price level. #Zcash #Privacy

4/9 The gap shows up next to Monero, where every transaction is private: → Zcash $20.8bn, about 5,600 private tx a day → Monero $10.3bn, about 27,700 private tx a day For each private transaction a day, the market values ZEC about ten times as highly as Monero.

1/9 What is Zcash, and why does its privacy cost ten times as much as Monero's? ZEC went from $74 to a record close of $1,653 in a year and became the tenth-largest cryptocurrency. Privacy explains the rise. It does not explain the price level. #Zcash #Privacy

2/9 What Zcash can do that Bitcoin cannot: hide the amount from everyone. → Bitcoin on-chain amount always visible → Lightning hidden on-chain, visible to routing nodes → Zcash shielded sender, receiver and amount encrypted Shielding is optional. 29 % of all ZEC is shielded.

1/6 On Oct 12, about three quarters of S&P 500 weight sat in a buyback blackout, the weeks before earnings when companies stop buying their own stock. When does that buyer return, and does the market rise with it? The calendar, and a test with a date. 🧵 #Stocks #Buybacks

5/6 The test: the S&P 500 closes higher on Dec 15 than on Oct 23. A win alone proves little. Oct 23 to Dec 15 was positive in 52 of 76 years since 1950, average +2.2 %. In midterm years: 12 of 19, +2.6 %. The margin decides.

1/6 On Oct 12, about three quarters of S&P 500 weight sat in a buyback blackout, the weeks before earnings when companies stop buying their own stock. When does that buyer return, and does the market rise with it? The calendar, and a test with a date. 🧵 #Stocks #Buybacks

3/6 Share of S&P 500 weight with an open buyback window, Citadel estimate: → Oct 12: 24 % → Oct 31: about 55 % → Nov 3, midterms: about 75 % → Nov 9: about 85 % Goldman counts companies: 48 % of S&P 500 members open before end of October.

1/6 On Oct 12, about three quarters of S&P 500 weight sat in a buyback blackout, the weeks before earnings when companies stop buying their own stock. When does that buyer return, and does the market rise with it? The calendar, and a test with a date. 🧵 #Stocks #Buybacks