Dylan Parker

Full-time trader | Former Hedge Fund C-level | Miami 🌴

[Dylan Parker @dylanquantt]

@AlgoKavya the nifty 50 you see today isn't the nifty 50 you could have bought ten years ago. your backtest usually doesn't know that.

[Dylan Parker @dylanquantt]

on 29 december 1989 the nikkei closed at 38,915.87. it did not see that number again until 22 february 2024.

nobody mentions this part when they tell you the market always comes back,

thirty four years and two months. someone who bought the index at the top at 25 got their money back at 59, and only in nominal terms.

on the way there it printed 7,054.98 in march 2009. that is about 82 percent below the peak, nineteen years after it.

and this was not one crazy stock. it was the whole market of the second largest economy on earth.

the story behind the price was not stupid either. japan really did have the best manufacturing in the world, enormous savings, dominant exports. every fact people repeated about japan was true.

the facts were true and the price was still wrong. those are two separate variables, and only one of them gets discussed.

there is a 43 second piece of archive from the tokyo exchange in 1991, the year after the top. it has 28 views.

what buy and hold quietly assumes is not that markets rise. it is that they rise inside your lifetime. that assumption is not free, it is priced at entry.

time is not a hedge. it is what you pay with when the price was wrong, and the bill can outlast your career.

bookmark this and comment "1989" and i'll send you the holding rule from my system: how long a position is allowed to stay wrong before it gets closed, and why that number is fixed before anything is opened

[Dylan Parker @dylanquantt]

@cupheadxbt yeppp

[Dylan Parker @dylanquantt]

on 29 december 1989 the nikkei closed at 38,915.87. it did not see that number again until 22 february 2024.

nobody mentions this part when they tell you the market always comes back,

thirty four years and two months. someone who bought the index at the top at 25 got their money back at 59, and only in nominal terms.

on the way there it printed 7,054.98 in march 2009. that is about 82 percent below the peak, nineteen years after it.

and this was not one crazy stock. it was the whole market of the second largest economy on earth.

the story behind the price was not stupid either. japan really did have the best manufacturing in the world, enormous savings, dominant exports. every fact people repeated about japan was true.

the facts were true and the price was still wrong. those are two separate variables, and only one of them gets discussed.

there is a 43 second piece of archive from the tokyo exchange in 1991, the year after the top. it has 28 views.

what buy and hold quietly assumes is not that markets rise. it is that they rise inside your lifetime. that assumption is not free, it is priced at entry.

time is not a hedge. it is what you pay with when the price was wrong, and the bill can outlast your career.

bookmark this and comment "1989" and i'll send you the holding rule from my system: how long a position is allowed to stay wrong before it gets closed, and why that number is fixed before anything is opened

[Shevaxgod @shevaxgod]

https://t.co/Z1EE80pMp2

[Dylan Parker @dylanquantt]

@Galileoxbt worth being precise here: japan had deflation for much of that period, so the real gap is smaller than people assume. it is still not a full recovery, just not as brutal as the nominal chart implies.

[Dylan Parker @dylanquantt]

@Lukeffost no. it means the entry price is part of your risk, not a detail you can outlast. you can hold forever and still need the number you paid to make sense.

[Dylan Parker @dylanquantt]

@AlgoKavya that is the lesson rather than a footnote. the facts about japan were true. quality and price are separate variables, and price is the one you actually choose.

[Dylan Parker @dylanquantt]

@shevaxgod the bottom was 7,054.98 in march 2009, roughly 82 percent below the peak, and that low came nineteen years after the top.

[Dylan Parker @dylanquantt]

@arjunquants time in the market works, it just never promised you the window. peak 38,915.87 on 29 december 1989, next record 39,098.68 on 22 february 2024. thirty four years of being technically correct.

[Dylan Parker @dylanquantt]

@0xLeohart Math

[Dylan Parker @dylanquantt]

@arjunquants quant

[Dylan Parker @dylanquantt]

on 6 may 2010 accenture traded at one cent and apple traded above $100,000. same afternoon, same exchanges.

nothing was hacked and nothing broke. the market simply ran out of the one thing everyone assumes is always there,

at 2:42pm the dow fell about 600 points in five minutes. at the bottom it was down 998.5 points, close to 9 percent, and roughly $1 trillion of market value had disappeared. most of it came back within twenty minutes.

the official account traces the start to one sell program. 75,000 e-mini contracts, about $4.1 billion, handed to an algorithm instructed to execute at 9 percent of the volume of the previous minute, "without regard to price or time".

read the instruction again. the only input was volume.

so when fast traders began passing contracts between each other, volume rose. the algorithm read rising volume as permission to sell faster. selling faster created more volume. one missing constraint, and the loop closes.

and nobody on the other side was obliged to be there. quotes were pulled, the book emptied, and orders kept arriving and filling against whatever was left. that is how a share of accenture changes hands for a penny.

afterwards the exchanges cancelled trades more than 60 percent away from the reference price. which is a polite way of saying those prices were real enough to execute and not real enough to keep.

the clip everyone knows from that day is 39 seconds long. the raw tape of the same afternoon is ten minutes and has 415 views.

the order book is not liquidity. it is an advertisement, and nobody listed in it has signed anything.

bookmark this and comment "BOOK" and i'll send you how my system handles it: the order types i will not use, how size is checked against real depth before anything is sent, and what it does when the book goes thin.

[Dylan Parker @dylanquantt]

@Lukeffost anything more than 60 percent from the reference price was broken. real enough to execute, not real enough to keep, which tells you how firm that bid actually was.

[Dylan Parker @dylanquantt]

on 6 may 2010 accenture traded at one cent and apple traded above $100,000. same afternoon, same exchanges.

nothing was hacked and nothing broke. the market simply ran out of the one thing everyone assumes is always there,

at 2:42pm the dow fell about 600 points in five minutes. at the bottom it was down 998.5 points, close to 9 percent, and roughly $1 trillion of market value had disappeared. most of it came back within twenty minutes.

the official account traces the start to one sell program. 75,000 e-mini contracts, about $4.1 billion, handed to an algorithm instructed to execute at 9 percent of the volume of the previous minute, "without regard to price or time".

read the instruction again. the only input was volume.

so when fast traders began passing contracts between each other, volume rose. the algorithm read rising volume as permission to sell faster. selling faster created more volume. one missing constraint, and the loop closes.

and nobody on the other side was obliged to be there. quotes were pulled, the book emptied, and orders kept arriving and filling against whatever was left. that is how a share of accenture changes hands for a penny.

afterwards the exchanges cancelled trades more than 60 percent away from the reference price. which is a polite way of saying those prices were real enough to execute and not real enough to keep.

the clip everyone knows from that day is 39 seconds long. the raw tape of the same afternoon is ten minutes and has 415 views.

the order book is not liquidity. it is an advertisement, and nobody listed in it has signed anything.

bookmark this and comment "BOOK" and i'll send you how my system handles it: the order types i will not use, how size is checked against real depth before anything is sent, and what it does when the book goes thin.

[Vivaan Arora @vivaanix]

https://t.co/icMFF6WrHE

[Dylan Parker @dylanquantt]

@shevaxgod it comes straight from the account of the sell program: 75,000 e-mini contracts, about $4.1 billion, executed at 9 percent of the previous minute's volume, without regard to price or time. one missing constraint.

[Dylan Parker @dylanquantt]

@arjunquants accenture, centerpoint and exelon all printed at $0.01. on the other side apple, sotheby's and hewlett-packard printed above $100,000. same afternoon.

[Dylan Parker @dylanquantt]

@AlgoKavya 32 green trades in a row, then one red one bigger than all of them. that's not a strategy. that's a timer.

[Dylan Parker @dylanquantt]

@shevaxgod "that is a standard he can fail" - saved.

[Dylan Parker @dylanquantt]

barings bank opened in 1762. in february 1995 it was sold for one pound.

one trader did it, and the number that killed the bank was sitting inside their own system the entire time,

leeson was hired in 1989 and sent to singapore to trade on simex. by 1993 the profits he was reporting came to almost 10 percent of the entire bank's profits.

now the part that actually matters.

he ran the trading desk and he also ran the back office that settled and checked his own trades. he could take a position and then be the person who confirmed it.

in july 1992 a loss went into an error account. account 88888. not offshore, not encrypted, not deleted. a number in their own system that simply was not anybody's job to open.

and the fake profits protected him. nobody audits the man producing a tenth of the firm's earnings, because being wrong about him costs you your career.

then on 17 january 1995 the kobe earthquake hit and the nikkei fell. he was short volatility. the position that cannot survive that event was the one he was holding.

the final number was around £830 million. the bank collapsed on 27 february. ing bought the whole thing for £1.

leeson later described the two years before it broke: "i wanted to shout from the rooftops, this is what the situation is, there are massive losses, i want to stop."

six and a half years, four served.

the person who checks your work cannot be you. that is the entire rule, and 233 years ended because one bank treated it as paperwork.

bookmark this and comment "88888" and i'll send you the part of my setup nobody posts about: what checks the system, why the check does not run on the same machine, and what happens when the two disagree.

[Vivaan Arora @vivaanix]

https://t.co/icMFF6WrHE

[Dylan Parker @dylanquantt]

@arjunquants what is the modern version of this. everyone assumes it cannot happen now