Fabio Valentini’s auction models, tested on Nasdaq and gold without his order flow

Source

What the stream says

The method

Step by step, from chart to risk

His trend model as he teaches it, in order. The quotes are his own words.

1. Trade one market in its best session

   Nasdaq futures in the New York session. Don’t trade before the open; the direction usually becomes clear 15–30 minutes in. Close everything the same day.

   “The best session to use this model it’s 100% New York session, specifically for equities.”

2. Read the market state from a profile

   Plot the volume profile of the previous day (or of the last compression): value area high, value area low and point of control (POC). Inside the value area the market is in balance; leave it alone. The trend model only trades once price is out of balance.

   “If you wait for the market to get to a condition of out of balance, your win rate will jump up by at least 20 to 30%.”

3. Skip the first drive

   The first push out of balance can be a fake-out. Wait for it to break, then for the pullback.

   “Always trend following or reversal, wait for the second drive. Don’t take the first drive.”

4. Find the location low-volume node

   Plot a profile over the breakout impulse and find its low-volume node, the thinnest-traded price. Set an alert just before it; don’t leave a limit order there.

   “I don’t put a limit order. I put an alert a little bit below.”

5. Wait for aggression

   At the level, wait for big aggressive market orders in the trade direction, filtered to at least 30 contracts on the 1-minute chart. Enter with them. Alternative trigger used live: a full-body 1-minute close through the range.

   “Your ability to predict is zero, but your ability to read is 100.”

6. Stop just behind the aggression

   The stop sits just beyond the big orders, one or two ticks inside the extreme to avoid slippage. If they fail, you’re out at once.

   “If I’m wrong, I want to be wrong immediately.”

7. Target the previous balance 1:2.5 minimum

   Aim for the previous balance area’s POC or the previous day’s high or low, and take the whole position off there. 1:2.5 is the minimum reward to risk, 1:5 is common.

   “The probability is that the market will reverse from it 70% of the time.”

8. Manage fast and build the day

   Move to break-even when the next breakout confirms. Take partials. Extra, riskier trades are paid for only with that day’s profit.

   “Building profit for the day, building profit for the day, building profit for the day.”

9. Cap the risk

   0.25% per trade on his own account, at most 2% loss a day, and stop early on days when the model isn’t working (compression).

   “I have a maximum stop loss for the day, like in percentage term, that is 2%.”

Implementation

What could be tested, and how

He says the model also works with price action alone (“It’s working also for price action”), which is what this test leans on. Each part of his model, against what the data can show:

His market

Nasdaq (US100) results

Both models together, with his 2% daily loss limit, at futures-like costs (0.5 point per round trip). The trend model’s median stop was 16 points, about $320 per NQ contract, in line with the $160–650 he risked in the stream, and its median hold was 3 minutes.

Robustness

His claims, checked on both markets

Each row switches one part of the model off or changes it, keeping everything else the same, at each market’s real broker spread. Only one claim holds on both markets.

R per trade. Many of these differences are within the noise of 1,000–2,000 trades; the consistent signs matter more than the exact numbers.

Diagnosis

Cost or edge?

Running the same trades with zero spread separates the edge from the cost. On gold the trend model had an edge before costs that the spread ate. On Nasdaq there’s nothing for costs to eat: the trend model is negative even at zero cost, and bigger stops don’t help (stops of 20 points or more: −0.13R per trade).

The gold result now looks like the exception. Across 26 versions on two markets, one positive result just above the significance line is what chance alone produces now and then, and it was on the market he doesn’t trade.

Second market

Gold results

The same playbook on XAUUSD, October 2021 to September 2026, at gold’s real spread. Gold’s median stop was only $1.62 against a ~$0.35 spread, so costs took about a fifth of every trade’s risk.

All versions

Every version tested

“Win” counts trades that ended above zero; break-even exits count as 0R, not wins. The Nasdaq 2× volume row has only 4 trades. Nasdaq in-sample is mostly winter fragments of 2021–2023 plus 2024.

Conclusion

Verdict: no edge without his order flow

Caveats

Limits of this test

• Tick volume counts price updates from the broker, not contracts traded, and says nothing about who was aggressive. The trigger here is a stand-in for his, not the same signal.
• US100 is a broker CFD on the futures. Its price follows NQ, but its spread (median 1.25 points) is wider than futures costs; the futures-cost run scales it down to 0.5 point.
• Orbi’s US100 history skips mid-March to mid-December in 2021–2023, and has gaps in March–April 2024. The 837 usable sessions are mostly December 2023 to September 2026.
• The impulse, the 20–80% node band and the previous-day value area are my definitions. He sometimes uses the profile of the last compression instead of the previous day.
• His live management (partials, re-entries funded by the day’s profit, reading CVD divergence) isn’t modelled; break-even at +1R approximates it.