Interview noteThis page restates a public interview in new words. Dollar results, win rates, and payouts are claims made by the speaker or the host. They are not this desk's record, and they are not a result to expect.
Matteo Conti, answers the question he says retail traders keep asking an institutional trader: which one strategy would you run to pass a prop challenge. The host introduces him as a seven-year market maker, formerly at Nordea Markets, a member of World-Class Edge, and chief investment officer at SQR Capital. The caption also hears the first name as Mattio, Mateusz, and Michio. The strategy is the one his community calls the prop-firm golden ticket, a systematic VWAP pullback he built for the mathematics of a challenge rather than for a personal account. The host says the historical test produced over 300 percent across more than 4,000 trades. Matteo's own later figures are a 64 percent profitability rate, a single-challenge pass rate of 49.8 percent, and a 93.6 percent chance of passing at least one challenge inside four attempts. None of it is offered as advice.
The misconception, then the reason VWAP is crowded
The mistake he thinks retail makes is hunting the next trade and needing that trade to work. Institutional trading, in his contrast, hunts a profitable strategy, which means hundreds of trades over years. Knowing whether an idea has an edge is a test. Gather the data, encode the rules, run them on history, then run simulations that speak to whether the thing can make money later. Retail, he says, often skips that. He brought this particular system because the context is a prop firm. It is built to maximize the chance of getting funded, not to be the strategy he would run on his own capital.
The microstructure claim is about execution traders inside banks and brokerages. Their job is to take the portfolio manager's order and get it done. He says 90 to 95 percent of those orders are worked by VWAP algorithms, targeting the execution around the volume-weighted average price. VWAP, as he defines it, is the average price of everything traded up to that moment, weighted by volume: the price where the largest amount of money changed hands. Because so many algorithms are trying to finish near it, a pullback into the VWAP is where their activity intensifies. They are more certain of a fill closer to the benchmark, and the imbalance between selling algorithms and buying algorithms gets revealed as a sharp move, down if sellers are in control, up if buyers are. That push is what the strategy is trying to take. The VWAP he uses is anchored at the cash open, 9:30 a.m. Eastern.
Three conditions, then the first pullback candle
The chart is Nasdaq 100 futures. He loads a five-minute and, for the trend question, a fifteen-minute. He also has a fifty-minute graph up at the start. The VWAP that matters is calculated on the fifteen-minute and drawn on the five-minute. Every fifteen minutes the system asks whether a trend exists. For a long, three things have to be true together. Price is above the VWAP. The VWAP itself has been rising over the past fifteen minutes, which is what a green cross on his indicator means. And over the past hour, four fifteen-minute bars, the Nasdaq has increased. He first says that increase is 1 percent. When he states the short and then restates both sides, the threshold is 0.1 percent: higher than 0.1 percent for the long, lower than minus 0.1 percent for the short. The 0.1 percent is the figure he says was the optimal parameter. For a short the mirror is price below the VWAP, the VWAP falling over the past fifteen minutes, and the hour down at least 0.1 percent. The three rules are level, direction, and speed, and he wants all three at once. The first hour, 9:30 to 10:30 Eastern, is not traded. The VWAP needs that hour to show where money actually changed hands.
The trigger is the first pullback after those conditions are true, and it does not have to tag the VWAP closely. For a long, the first red candle pulling toward the VWAP. For a short, the first green candle. The order is a market at the open of the next bar, sent as the signal candle closes. A long example on the tape has price above a rising VWAP, the hour up more than the threshold, a pullback, and a long at the open of the next bar that finishes in the money. A short example is a falling VWAP, a very small first positive bar, and a short that also finishes in the money. Because the rules are coded, the trades are the computer taking every signal, not a manual backtest of the pretty ones.
He came to the rules as a market maker who has watched VWAP execution for seven years. Closes that keep printing under VWAP, in that experience, come with heavy selling. A VWAP that keeps falling is a drift. The 0.1 percent was an optimization on in-sample data, chosen to limit overfitting. Development used 2020 through 2024. He says the same rules have been used over the prior six months to pass challenges, and that the reason they fit a prop account is in the statistics rather than in a story.
Exits that lose more than they make, on purpose
The long risks 80 points to make 40. The short risks 80 to make 50. He calls that a negative reward-to-risk in the way retail uses the phrase. The guardrails that make it mechanical are one position at a time, a maximum of four trades a day, and a stop after two consecutive losers. No new trade after 3:30 p.m. Eastern. Everything is flat at 3:55 p.m., before the cash session ends. Four trades and two losses were the in-sample numbers that maximized the chance of passing a challenge. Sniping the exact moment of the pullback is hard, which is why the count is capped rather than left open. He shows a day with the cap on: four trades, three wins, one loss.
The equity curve he puts up runs since 2021. Research stopped at 2024, so from 2024 through 2 August 2026, the day before the conversation, the curve is out of sample. His explanation of why that split matters is the standard one, said plainly. Overfitting can draw a straight line up by adding rules and cherry-picking the take-gain and the stop that looked best on the past. Unseen data then destroys it. Split the history, change parameters only on the first half, freeze them, and look at the second half. Here the research was through 2024 and the same inputs were applied forward. Both halves rose. He calls that the first check, and he says the probability of overfitting is then small. It is not a proof.
Why a bad payoff passes challenges
The number he points at is the 64 percent profitability rate, which is high because the payoff is low. The average trade is small. On a private account he does not think this system makes money. Later he describes the same profile as an average winner of about $866, an average loser of about $1,300, and a profitability rate of about 65 percent. Those averages are a losing payoff. The win rate is what makes the combination, in his simulations, unusually good at passing a challenge and unusually poor as a personal-capital strategy. He downloaded the trades from 2020 and 2021 onward, wrote a Python simulation, and ran 20,000 paths. A single challenge, traded as one NQ or ten micro NQ, passes 49.8 percent of the time, in about 3 days, or 3.4 days in the more precise reading, inside a 20-day window. Two attempts raise the chance of at least one pass to 74.8 percent. Three attempts, 87.3 percent. Four attempts, 93.6 percent. If a challenge costs $50, four of them are $200, and that is the arithmetic behind the nickname. He is explicit about the next problem. Getting funded is this strategy. Getting paid, once funded, is a different strategy, and he leaves it for another conversation. The host's reaction is that a genuine 50 percent pass rate is already far above what he sees, and that the fantasy of a 90 percent win rate at 10 to 1 does not exist.
Manual trading of the same rules is possible and fragile. His recommendation is to keep the signal encoded and enter by hand when it fires. Eyeballing the three conditions will sometimes match and sometimes not, and it will not carry the 49.8 percent. Full automation is the further step, and at that point he would not run one system. He would run three or four, on a prop account or on personal capital. The job, as he describes the institutional version, is about 90 percent research, development, and validation, and about 10 percent watching the risk and the performance of a computer that is already in the trade. The anecdote is a follower who had blown 47 challenges, decided it was impossible, then passed the first one with this approach while he was at work, after years of sitting ten hours for entries that a machine could wait for.
Trust the test, and know when the test is no longer describing the live run
What still stops someone who understood the interview is the urge to out-trade the research. Ten hours to find something that looks profitable, then automation, then a hand on the take-gain, the stop, or one extra trade that will surely improve it. His instruction is to stay out of that urge. The hard part of systematic trading is trusting work that is already done. Discretion can stay. Testing what you do still adds something, even if nothing is automated. Simplify a discretionary idea until it can be run on history and you can see whether those rules had a positive drift. The same tooling answers a question like whether five years of trades at a 20-point target would have looked different at 100 points.
Telling a normal drawdown from a broken or overfit system is what the Monte Carlo is for. It produces yardsticks. His example of the shape of one is a 5 percent probability of a drawdown larger than $10,000 over ten trades. Five thousand dollars is inside that yardstick. Fifteen thousand is not, and the strategy comes off. The thresholds come from the validation of that specific system. Inside them, he does not want the trader to worry. Outside them, stop, and do not keep it live. The process from idea to deployment, and the monitoring after deployment, is the skill. His last line is shorter than the model. If you do not know what you are doing, do not trade. If you want to trade, test it first, and put capital on it only after there is statistical evidence it is likely to make money.