LessonThis is a lesson restated in new words. Any figure in it is an example from the source, not this desk's record.
Gianluca Bruni says that a single morning is too small to prove a mechanical edge. Picture a sixty percent win rate at one-to-one, he says, and only after the series has survived checks for overfitting. Uncertainty, in the formula he uses, is one over the square root of the number of trades. Ten observations are about a thirty-one point swing, so a true sixty can print as thirty or as ninety. A thousand observations shrink that to about three points, a band he places near fifty-seven through sixty-three. The sketch is his, not an audit.
Why it matters
He treats about one hundred trades as the first sample quiet enough to mean something. No morning produces that count, so one stopped trade is uninteresting and a slightly worse later signal is preferable to a sample of one. Ten fills are still not a verdict. The host adds the condition underneath: the win rate has to keep being the win rate. Emre Gams, sharing the room from the options side, says the same thing about contests. One competition is one observation, not a skill score. The host is also explicit that Gams's open option gain, north of twenty thousand dollars, was not made while the cameras were on.
The idea in plain words
The introduction calls Bruni a Robbins World Cup champion who executes a statistical expectancy by rule. Gams is introduced as a former hedge-fund manager reading options flow. Different products, different jobs. Bruni wants Nasdaq downside momentum because a few hours will not give his main playbook enough signals to outrun variance. Gams is positioned in S&P futures where, on his account, the heaviest business has already traded, with options opened about a week before.
Bruni skips the story of the break. Once a session looks short, expectancy is positive only after the trigger, so the work is seeing the trigger. The marked zone can be called supply or demand. He does not care. The break is what flips him. Fifteen minutes is home. Five minutes is today's compromise, chosen so more triggers can appear. What he shows is two candles, each at least sixty ticks, both aligned with a trend that two layers of structure agree on. He calls the layers actual and peripheral. Up candle, down candle, repeating, and there is no setup. If the layers disagree, that is chop and a pass. Picking the single prettiest tick cutoff, he says, is how a backtest overfits.
This screen model is a prop tool, not the main strategy. Buying power is the prop advantage he names. Frequency, in his framing, carries the expectancy. Something with zero point two expectancy and three signals a week can lose, under prop risk limits, to a smaller expectancy that fires six times a day. The sheet reads zero point one two R, which he calls deliberately thin. A luck run on the same data peaked around forty R. Their path reached about fifty R, and he treats the gap as the edge getting past luck. Net on the page is fifty-nine point eight percent at one percent risk, a figure he discounts at once because the sample is too short to be realistic. Still being in a loss after one hundred trades has about a sixteen percent chance on his test. After four hundred six trades that chance is about two point three percent. He does not claim a risk-adjusted win against a benchmark, and without the prop wrapper he would not run this version himself. Clearing the prop constraints is the only claim.
How to use it
The stop is chosen before the target. It sits above the two signal candles, at the high of the first. Reward is always one and a half times that distance. Size is set so the stop is about one thousand dollars. If the fifteen-minute later prints the same pattern, he adds. A fresh valid pair while he is already short is a new position. Red alone is not enough. Below sixty ticks, the candle is invalid. He takes the five-minute short anyway, ten contracts, because the hour needs repetitions. That one fill does not earn his confidence. He says the positive expectancy appears across a thousand repetitions. Five or ten say nothing.
That first short loses. The level he was leaning on breaks, so he reverses long toward a marked zone and is willing to do it against the value area. Uncommon, he says, and closer to a fifty-five percent coin than sixty, still paid at one and a half. He is late, keeps the contract count, and puts the stop under the low of the first signal candle. An add requires the next candle to travel at least sixty ticks, and the combined stop then moves under the new structure. Pyramiding is his word for the add. Twenty-two contracts appear because the stop distance changed, not because he dropped the dollar cap. Near a two-thousand-dollar stop, one and a half times risk is about three thousand. They are still about nine hundred down from the loser, so hitting the target would leave roughly two thousand one hundred up on the session. The second down pair never prints. The long works. About three hours later the target fills: about three thousand on the winner and about two thousand for the session after the loss. The comparison the host wants is visual. Gams's twenty thousand was mostly already open and off camera. Bruni's roughly two thousand one hundred happened under stated rules. Bruni refuses to treat the comparison as the experiment.
Gams will not inherit Bruni's bearishness. At the inflection he is flat, after being both short and long on the opening drive inside a tight range. Guessing the next few seconds, he says, is how a trader gets chopped. The options position wants price to remain inside a hundred-point S&P range, seven thousand eight hundred seventy down to seven thousand seven hundred seventy, chosen where he thinks options are mispriced. Leaving the middle, he hedges with futures and calls the hedge gamma scalping. He will not predict the winner. One gamma level near seven thousand eight hundred thirty-two matches a thin overnight pocket from about seven thousand eight hundred thirty to seven thousand eight hundred thirty-six. He will sell against that pocket only after defense appears. Balance, on the bell curve he sketches, is the usual state. The violent residue is about thirty to thirty-five percent of sessions, and being right on about thirty percent of those breaks is, he says, enough for compounding. Options do not get a futures-style stop. He asks how much of the option gain he will surrender. A futures short down about three hundred dollars is noise next to that book.
What usually goes wrong
Bruni's 2020 education was to feel the tape. Discretion, he says, took money and composure because the feeling changed by the morning. Gams turns down human delta-one funds for the same reason. Next year's instinct can be a different instinct. A friend running an algorithmic fund, in his telling, made about one hundred percent, or about eighty after fees, with a peak-to-trough hole well into the double digits. The gain is real as told. The hole exceeds what Gams will accept. Futures can, on his account, turn twenty thousand dollars into a few hundred thousand and can also lose twenty thousand repeatedly. A positive expectancy can still be the wrong business. Scoring one day is the error both men are avoiding.
Questions people ask
When does Bruni say a win rate starts to mean something?
Around one hundred trades. Ten trades, on the square-root sketch, can still disguise sixty percent as thirty or ninety. A thousand trades narrows the band to about fifty-seven through sixty-three if the rate does not decay. A single loss is not evidence.
Is zero point one two R the strategy he lives on?
No. He presents it as a prop-frequency tool, not his main book, and he discounts the fifty-nine point eight percent page return for a short sample. Personal capital, he says, would not run this version. The prop constraints are the only place he claims it fits.
Why ignore the larger gain on the screen?
Because the host says the options gain above twenty thousand dollars was already open before this session. Bruni's roughly two thousand one hundred dollars is the rule-following result the room could watch. He does not call that the test, and Gams calls a contest a single data point.