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, former market maker of seven years and chief investment officer at SQR Capital, returns with three systematic strategies built for a prop challenge. The host is Brandon. Months earlier Conti had shown one VWAP approach the host calls the golden ticket, and the host says that version had historically produced up to a 93.6 percent probability of passing a challenge. This session is not that strategy. It is three new ones, each with a reason, an entry, and an exit, shown first on a whiteboard and then as code on Nasdaq futures. Nothing in it is advice. The position size used to explain them is one contract. How he would actually run them on a challenge is a different number, and he saves it for the end.
He builds automated strategies for prop accounts because that is the question retail traders kept asking him: how an institutional trader would attack this specific test. Passing and getting paid are different problems. A challenge strategy is built to touch the gain target before the maximum drawdown. A funded strategy is built to maximize what can be withdrawn from that one account. The three in this hour are aimed at the first problem, with one of them better suited, in his view, to the second. Trading all three by hand gets complicated. Once they are code, a machine can watch every condition. He also says many prop firms block automated execution and justify it as a preference for discretionary traders. He treats that as a weak excuse, because automating the execution is an advantage against the field.
The vault break: a noise band, then the drift
The first strategy he calls the vault break. The caption also says bolt break. It is Nasdaq-100 mini futures, a 30-minute chart, and long only. The hypothesis is the positive drift of the market. It resembles an opening-range breakout, with one difference. A plain opening-range breakout trades often, close to every day. This one refuses the noise. The first hour can push and still not be enough. He wants a clear directional push before a long is allowed.
The noise idea, he says, goes back to academic work, the earliest paper he cites from 1987. The construction is simple. Take the Nasdaq price at midnight Central Time. That is the session open for this model. Take the average true range of each of the past 15 sessions, from that midnight to the end of the trading day, which he places at 5:00 p.m. Eastern, 4:00 p.m. Central. Average those 15 ranges. It is a moving average of the session ATR, about three trading weeks of volatility, not a single 15-bar ATR on the 30-minute chart. Keep 30 percent of it and add that distance to the midnight price. That line is the noise-up barrier, and it is fixed at midnight. It does not walk up through the day. The 30 percent was optimized in sample over more than three years and, he says, held up out of sample. A full ATR would trade too rarely. Thirty percent was the compromise between the width of the band and how often the strategy fires.
From 10:00 a.m. Central, if a bar closes above the noise-up barrier and price is also above the session VWAP, he buys at the open of the next bar. VWAP here is the volume-weighted average price, the reference where the most money changed hands, and on the chart he calculates it from midnight as well. The window runs from 10:00 to 14:30 Central, four and a half hours. No trade outside it. Inside it he will take up to three trades. The cap was chosen in sample. There is no separate news filter. If the drift is real and the VWAP agrees, he does not care what headline produced it.
The exit is a fixed target, a fixed stop, or the clock. On the whiteboard the stop is stated at $1,500 and the target is garbled. On the later trade statistics he is specific: an $800 target and a $1,400 stop. Those are the figures attached to the tested version. If neither is hit, the position is closed at 14:30 Central. If the target is hit and price is still above the noise barrier and still above VWAP, he takes another trade, then a third at most. A stop in the middle of that sequence does not cancel the next one, as long as the two conditions and the window are still true. The geometry is intentional. A reward smaller than the risk raises the win rate, and a high win rate is what he wants inside a challenge, where a long losing streak ends the account before the target arrives. He does not claim the edge is enormous. He claims the shape fits the test.
The coded chart uses data back to 2018, with the out-of-sample period beginning in 2023. One historical example closes above the orange noise line, sits above the blue VWAP, takes the long, hits the target, and opens a second trade because the conditions are still true. That second trade is closed by the clock. A later signal outside the window is ignored. The equity curve, as he reads it, is making new highs, which is his requirement before he will point a challenge at a strategy. Average winning trades are smaller than average losing trades, which follows from an $800 target, a $1,400 stop, and a time exit that cuts both short of those prices. The average trade is still positive. He reads it at $112 on the Nasdaq. The statistic he cares about is the win rate: over 60 percent, almost 67, at about 0.6 to 1.
The VWAP pullback: armed by the opening range, gated by ADX
The second strategy is the VWAP pullback he has shown before, with an ADX gate added. It is again the Nasdaq, long only, one contract, and this time a one-minute chart. The hypothesis is microstructure. A large share of institutional execution is a VWAP algorithm, benchmarking fills to that average. When price returns to it, those algorithms interact. If the session has a positive drift, the pullback can be the entry into the push they create.
The opening range starts at 8:30 Central, the cash open at 9:30 Eastern. A one-minute bar that closes above the opening-range high arms the strategy. That close is not the entry. It is evidence of buying. He then waits for price to touch the VWAP. A wick is enough. The entry is a close back above the VWAP, at the open of the following bar. One candle can do the touch and the close. The last gate is the average directional index, the indicator Welles Wilder Jr. published alongside RSI and ATR. ADX measures how strong a trend is. It does not say which way. Conti knows a hedge fund will not use ADX as the academic definition of trend. Autocorrelation and the rest of that literature exist. He still uses it because it is practical, and because simple technical inputs are still fed into black-box models. His rule is ADX above 20, and not increasing versus the previous bar. Flat or falling means the pullback is losing its momentum. If price has already closed back above VWAP and ADX is the only condition missing, he waits. If ADX is under 20 and later rises through 20 and stops rising, the trade is still allowed. On the coded example he also describes the arming step as happening after 10:00 Central. The whiteboard sequence itself is the close above the range, the touch, the close back above VWAP, and the ADX gate.
The target is the high of the past five bars. The stop is the low of the past twenty. If neither prints, he is out at 3:55 Central. The five-bar high can be the high of the signal candle itself. He says that happens less than 5 percent of the time and does not break the strategy. Targets are limit orders. Stops are market stops. The same time exit sits on every prop strategy he builds, because most challenges cannot be held through the close. He did not know, until this conversation, that the host's firm allows an overnight position for a swap fee. His own rules still flatten.
This one trades less often than the vault break, so the average trade is higher. The host reads the prior strategy's average near $121 and this one's reward near 0.7 with a similar win rate. Conti agrees with the shape: reward-to-risk under one, and a high win rate, which is what he wants for a challenge. Development is again 2018, out of sample from 2023, and the equity curve is recently at new highs. In the example on screen, price trades up to the opening-range high after 10:00, pulls back, closes back above VWAP, and the entry waits until ADX stops rising. The five-bar target is hit. The twenty-bar low comes close and does not stop the trade.
Overnight bias: a third of the night, one trade, three to one
The third strategy is the overnight-bias opening-range breakout. Fifteen minutes, Nasdaq, and this one is long or short. The first two were long only because of the drift. This one uses the overnight discovery as the direction. On the whiteboard he records the overnight range from 11:00 p.m. Central to 8:30 a.m. Central, the participants who can trade before the cash open. On the coded chart he describes that same overnight range as starting at midnight Central. The two clocks are both in the session. They are not the same rule, and he does not reconcile them on camera.
After the open he marks the first fifteen-minute bar, 8:30 to 8:45 Central, as the opening range. The bias comes from where the 8:30 price sits inside the overnight range. Top third, longs only. Bottom third, shorts only. Middle third, no trade. The middle means the overnight auction never showed a real push. Once the bias is set, a dip during that first bar back into the middle third does not cancel it. A long bias buys a break of the opening-range high. A short bias sells a break of the opening-range low. The break candle also needs ADX above 20. Rising, flat, or falling does not matter. ADX is only the strength. The side comes from the overnight third. The fill is the open of the next bar. One trade a day. A long-biased day that first breaks the opening-range low, and only later breaks the high, is still a valid long. He will not add filters that cut an already low frequency. The overnight direction is the gate. The intraday break is the confirmation, and a violation on the way there does not erase it.
The stop is 30 percent of the same object as the vault break: a 15-session moving average of the session ATR, not a traditional ATR with a 15-bar lookback. The target is three times that distance. He calls the reward three to one. The frequency is lower than the other two, and he says the win rate is still high even with that reward. The trade is not held overnight. If neither stop nor target prints, he exits at 2:30 Central. He first says 4:30 and corrects it. The host's guess is right: most of these exits are the clock. A prop daily loss limit, he gives $1,000 as the kind of figure firms use, can also close the trade before the strategy stop is reached. Because of the geometry, he wants the first two strategies on evaluations and this one on a funded account, where the larger directional move is the point.
On screen, a day that opens in the top third breaks the first fifteen-minute range on the following bar, with ADX well above 20 and rising, and the long is the three-to-one version. The equity curve after 2023 is out of sample and not in a drawdown. The average trade is high because the reward is high and because there are fewer trades. The screen figure he accepts is 1.4 with 53 percent. The long and short average trades are close to symmetrical. The long side wins a little more often, which he attributes to the drift. The short side, as the host reads it, is about a 50 percent win rate with 1.6, on about 200 trades. He considers that important precisely because the market's upward drift cannot bail out a short. For the period loaded, one Nasdaq contract, he reads the result at about $200,000. Outliers contributed $88,000 across 53 trades, which he expects, because so many positions die on the time stop rather than on a full target. A challenge window of about 21 trading days is too short for this strategy alone. Sized properly and run by itself, many challenges would expire without a target or a stop. On a funded account it can sit in the arsenal for the larger moves, including accounts that require some activity without requiring this model to be the only one.
Bench it at the 75th percentile, and do not marry it
Someone trading these by hand and quitting after three losses has not seen the strategy. The edges are not strong in the sense of a huge expectancy. They have a positive expectancy, and they work in a challenge because of how often they trade and because of the geometry. Publishing them, he says, does not end the edge by itself. Passing, and getting paid, depends on how the strategy was built and how it is applied, including taking every trade the rules produce. In the group he works with, a strategy already in a drawdown keeps failing challenges, because losing streaks cluster the way winning streaks do. A strategy printing new equity highs is the one to run. A strategy in a drawdown is benched until it makes a new high-water mark. He does not put a moving average on the equity curve. That is an older method the host remembers. Conti's version is the new high, or it is nothing. He will not force a flat or a broken curve back into live trading. On a trading floor, he says, you do not marry a strategy. If it stops working it is benched. If it starts again, it can be reconsidered.
Two failed challenges in a row are the practical alarm. Each of these, run alone, passes about 40 percent of the time, so two failures back to back are a reason to look, not an automatic deletion. He compares the current drawdown with the historical one and asks which percentile it sits in. The test can be as strict as the trader wants. The reason to hold at least three strategies is that one of them will break and the account still needs a substitute. These three, in the tests he has run, correlate at less than 0.25. They can be on the same challenge because they are not the same bet.
The tool he uses to separate variance from a real drawdown is a simple Monte Carlo. MultiCharts will not run it. It will export the trades. From there he uses Python. For people who ask, he points them at Google Colab and at having a model write the small script: upload the trades, build the distribution of drawdowns. His example of a yardstick, not a measured result of these three systems, is a simulation that says the strategy should not be beyond $20,000 at the 75th percentile. A live drawdown larger than that $20,000 is the point where he benches it. The way back is still a new high in the equity curve.
All three are coded, each on its own Nasdaq chart, with the levels drawn as indicators so the rules are visible. He would not trade them by hand. He connects the challenge, and the platform watches all three. When a condition set is true, it opens the position. Running three at once improves the chance of passing relative to running one. The sizing he calls optimal is about five micros of each. At that size he puts the average time to pass at about five trading days. The work is the development. The session itself is the button. On the day of the interview none of the three is armed. The vault break would need a close through the noise zone, and price does not look likely to get there. The overnight open is in the middle third, so that model stands down. The pullback is the only one that could still fire: after 10:00, a touch of the top of the opening range, a return to VWAP, and the ADX gate.