- One thesis. “I always want to see price rebalance into 50% of the range and then continue with the trend.”
- Power of three, fractal. Accumulation, manipulation, distribution — on the daily, the H4 and the H1 at once.
- The manipulation is a run of the previous period's high or low that fails and closes back inside.
- 09:15 to 11:30 New York, and 10:00 above all: he expects the 10:00 candle to run the 09:00 high and reverse.
- The trigger is SMT divergence — the Nasdaq makes the high and the S&P fails to. This test cannot include it; see below.
- The target is 50% of the range, not the extreme: “I don't need to trade from here to here, because this is a risky move.”
- Stop beyond the swept level.
- Entry at the inversion. “I will either limit a retap into this area or I'll sell stop this” — the gap left by the push up, once price trades back through it.
- Break-even on his trigger. “As soon as this hourly candle flips bearish”, or “as soon as we take out the 15-minute low” — whichever comes first.
His break-even helps, as he says it does. Take the same trades and leave the stop where it was, and both markets do worse. He is explicit that people who call the model unprofitable are not managing it the way he does, and on this data he is right about that.
| market | his model | without his break-even | his break-even is worth | random entry |
|---|---|---|---|---|
| Nasdaq | $1.59 | −$1.37 | $2.96 | −$14.83 |
| Bitcoin | $303.60 | $163.00 | $140.60 | −$212.26 |
One of his rules does not help. The manipulation — requiring price to run the previous high first — removes trades that would have made money. Without it the Nasdaq makes $6.20 a trade ($2.54 to $9.73) and Bitcoin $444.66 ($267.22 to $623.49); both intervals clear zero. The entry at the inversion, the 50% target and his break-even appear to be doing the work.
His 10:00, directly. Split by the hour the model entered: 09:00 $10.54, 10:00 −$0.75, 11:00 −$1.72. On this data the earlier hour is the strongest and his 10:00 is roughly flat.
No number of ours changes it. We had to set two limits he never gives: how long to wait for the entry and how long to hold. The longest Nasdaq trade lasted 15 bars and the slowest fill took 5, so neither limit ever came into play. Doubling or halving them gives the same result.
| market and setting | trades | win | scratched | a trade | 95% range |
|---|---|---|---|---|---|
| Nasdaq · his model | 662 | 59.7% | 15.1% | $1.59 | −$3.97 to $6.92 |
| Nasdaq · the same set-up at any hour | 6,048 | 48.4% | 21.2% | −$0.45 | −$1.73 to $0.89 |
| Nasdaq · without the manipulation | 1,259 | 65.0% | 9.0% | $6.20 | $2.54 to $9.73 |
| Nasdaq · without his break-even | 662 | 67.2% | 0.0% | −$1.37 | −$8.56 to $5.52 |
| Nasdaq · none of it | 10,295 | 60.2% | 0.0% | $0.95 | −$0.10 to $2.05 |
| Bitcoin · his model | 589 | 59.4% | 19.5% | $303.60 | $9.54 to $592.94 |
| Bitcoin · the same set-up at any hour | not applicable — no session window on this market | ||||
| Bitcoin · without the manipulation | 1,228 | 65.9% | 12.4% | $444.66 | $267.22 to $623.49 |
| Bitcoin · without his break-even | 581 | 69.5% | 0.0% | $163.00 | −$220.05 to $546.94 |
| Bitcoin · none of it | 1,168 | 71.4% | 0.0% | $351.07 | $128.98 to $574.69 |
The same table, live. This one follows the two switches above, and it is what the pickers on the chart drive.
Everything else worth saying
Scratches are counted separately. A break-even exit comes back at minus the commission. Counting it as a loss would hide exactly what his break-even is for, so the table shows wins, scratches and losses apart.
What an earlier version of this page got wrong. It entered on a close back inside the range, stopped above the high, and moved to break-even at a fixed fraction of the risk — none of which is what he says. On those rules it concluded that his break-even was what cost him. Rebuilt on his entry, his stop and his break-even trigger, the result reversed. The numbers on this page are from his rules only.
How a bar is scored when it touches both. Most trades here resolve within one or two bars, so a single bar can reach both the stop and the target. Every such bar is scored as a loss. That is the harshest possible reading, and his model is positive despite it.
What this test does not do
We tested the mechanical skeleton: the power-of- three sweep, the session, the 50% target, the stop and the break-even. He says plainly that the rest is his: “I can't teach my intuition, and part of my edge is my intuition.”
- His trigger is missing, and it matters. He enters on SMT divergence between the Nasdaq and the S&P. This archive holds 1,304 days of Nasdaq futures and 28 of S&P, so there is nothing to compare against. No substitute was invented. Divergence is a filter that REJECTS set-ups, so this test takes trades he would have passed on — which works against him.
- One timeframe at a time, not three. He stacks the daily, H4 and H1 power of three and waits for all of them to align. This test uses one higher timeframe per market, which lets through set-ups he would skip.
- He trails winners by feel. On strong days he moves the target to the extreme. A fixed 50% target does not capture that, and it can only have cost him.
- The replay is a sample. Every number is scored over the whole period; the chart ships 40 set-ups on the Nasdaq and 40 on Bitcoin, spread across the whole test.
If you use a pair that is not in the grid, or a filter you think matters, say so and we will test it the same way and publish it.
The largest payout in prop firm history and $2.3m in total payouts, and $130,000 taken to over $1.4m in about a month. He is candid that part of the model is his: “I can’t teach my intuition, and part of my edge is my intuition.” And emphatic about the rest: “the model is profitable — you’re just not managing it correctly.”
Read it in their own words first: