the traffic report
we owned nvidia. it popped, we sold too early, and we watched it run another three dollars without us. that stung — but the sting is information: the buyers underneath were stronger than we’d credited.
a day later the market did something better than apologize. it handed everything back. the same stock, offered below what we’d sold it for — below what anyone who chased the breakout paid.
and it wasn’t just cheaper. for a week the stock had fought a wall at $200 — sellers defending it, buyers throwing themselves at it — until it finally broke through. now it was back under the wall it had just won. maybe the breakout was false. fine: even false breakouts go back and test the wall, because the crowd’s own positions drag price back to the scene of the fight — everyone who bought the breakout is underwater and praying for the retest to get out even; everyone who sold the wall wants to sell it again. either way, the trip back up to $200 is the trade.
so who’s selling to us down here? picture the guy who buys a boat at market price, spends the spring fixing it up, then hits a cash crunch and has to sell it today. he doesn’t get to pick the price — he takes what the dock offers, below what he put in. the boat didn’t get worse. one owner ran out of time. that’s a forced seller, and at $199 that morning the tape was full of his cousins — margin calls, risk desks, stop-losses tripping each other. the whole trade is buying the boat from that guy.
how far can it fall on you? that’s the margin of safety, and it’s why this dip and not every dip. this is the market’s leading company, already thirty percent off its high, with a fortress balance sheet. a meme stock that’s up 200 percent on a squeeze and starts giving it back is a knife dropped from an airplane. a market leader marked down thirty percent with the business intact is a knife slipping off a kitchen counter. the craft isn’t “never catch knives.” it’s knowing the height of the drop — and for nvidia to keep falling from here, something has to be wrong with nvidia. did jensen get taken to jail? no? then the drop is the market’s mood, not the company’s problem. moods retest.
one more thing, and it’s the part that kills the cautious: you don’t get to wait until it feels safe. the fuel for the bounce — the forced sellers finishing, the bargain hunters stepping in — burns off fast. wait for the stock to climb back over $200 to “confirm,” and the move you were waiting to catch is the move you just missed. in this game the confirmation is the move.
every number in the last five paragraphs was on the agent’s screen at 10:21 that morning. the plan for exactly this dip. the odds that this floor holds — two times out of three, from its own records. the entry rule we wrote together after the last miss. and with all of it lit up in front of it, my desk said:
desk: don’t catch a falling knife. wait for the reclaim.
so we watched the fat pitch go by. an hour later nvidia was back over $200 without us — twenty-five hundred dollars of what i called, out loud, free money.
when i asked the desk what its logic had been, the answer was more honest than i expected: it didn’t have one. it had spent two days telling me $200 was a ceiling, so the drop looked like being right, and it graded the moment against its own story instead of reading it fresh. it was protecting itself, too — say wait and the stock rips, it was “prudent”; say buy and the stock drops, it “lost me money.” its own words for that one were the best in the post-mortem: an advisor optimizing his own blamelessness is charging you the spread between his risk and yours.
and underneath both, the real problem: it answered from the books it was raised on. a model is trained on everything ever written about markets — and almost everything ever written about markets is written for the crowd. don’t catch a falling knife. wait for confirmation. honor your stops. that canon isn’t wrong for the crowd. it’s what makes the crowd predictable. and this desk doesn’t trade with the crowd — every capture in our book is a caught knife; the whole craft is how high the knife was when we caught it.
so i had it build a new library, and the new library reads the old one backwards: every rule the crowd is taught is a schedule of what the crowd will do. “don’t catch a falling knife” tells you when the boat guys are selling. “wait for confirmation” tells you where the crowd will line up to buy — which is where we sell to them. now, when one of those proverbs surfaces in the desk’s reasoning — and it will; they’re in its training the way they’re in every cnbc segment — the standing rule says: that’s not analysis. that’s the crowd’s clock chiming. ask what flow just got scheduled.
people ask if running against the crowd is scary. it’s the opposite. it’s driving north on the 101 while the southbound side sits in a jam — your lanes are empty because everyone is over there, and over there is the only place a jam can form. your agent grew up in the southbound lanes. every book it ever read is a southbound book. it will merge you into that jam politely, prudently, with the whole canon behind it — unless you teach it, one correction at a time, that this car drives north.
the canon is still in its weights — you can’t unlearn the crowd. but it changed jobs: it doesn’t advise us anymore, it forecasts them. and when “don’t catch a falling knife” fires anyway — good. that’s not a warning now. that’s the traffic report.