The two models trade different hours. The trend model is New York only, 09:30 to 16:00 New York time, and he will not touch the first half hour. The reversion model is London, 03:00 to 09:30 New York time — that is 08:00 to 14:30 in London, and it does not shift with the clocks because both cities change on their own dates but stay five hours apart. The replay starts where the model you have selected actually trades.
Most sessions do not trade. Step one throws out any day where price never leaves the previous session's value area, so about 4 sessions in 10 have no set-up at all on the trend model and 7 in 10 on the reversion one. That is the model working, not the page being broken — but use NEXT SET-UP to skip to a session that actually has one.
- Step one, the state. Only trade when the market is out of balance — outside the previous session's value area. "with this model we want to transact here. We are not interested in this market condition."
- Step two, the location. Profile the leg that broke out and find the low volume node inside it. "what you search is low volume node."
- Step three, the trigger. Wait at that level for aggression — a print of 30 contracts or more in your direction. "when you see a big red ball, you can jump in."
- The stop goes behind the swing the aggression defends, one or two ticks inside the extreme rather than past it: "you are taken out before everyone, before acceleration takes place."
- The target is the point of control of an older balance area, and the whole position comes off there: "the probability is that the market will reverse from it 70% of the time."
- Stop to break-even as soon as the trade proves itself, half off there. Flat at the session end, one position at a time. Real costs.
- The reversion model is the same machinery inverted, in London: price leaves balance, comes back inside, and only the second push is taken. "I don't take the first movement."
| what is switched on | trades | win | a trade | total |
|---|---|---|---|---|
| his model, both filters | 883 | 7.6% | −$6.13 | −$5,417.00 |
| without the big print | — | 7.6% | −$7.92 | — |
| without the low volume node | — | 5.7% | −$6.45 | — |
| neither — the session filter alone | — | 5.7% | −$5.74 | — |
By the time of day. He will not trade the open — "the direction gets really clear from 15 minutes to 30 minutes inside the opening". The worst entry window is 10:00-11:00, at −$9.31 a trade.
| entry window | trades | win | a trade | total |
|---|---|---|---|---|
| 10:00-11:00 | 186 | 4.8% | −$9.31 | −$1,732.00 |
| 11:00-13:00 | 328 | 8.5% | −$2.74 | −$899.00 |
| 13:00-15:00 | 248 | 7.3% | −$7.99 | −$1,982.00 |
| 15:00-16:00 | 121 | 9.9% | −$6.65 | −$804.00 |
By target distance. He warns that reaching for distance costs you: "the more you seek to go above the ATR daily, the more the probability will get lower". The trades reaching beyond 8x the risk make −$10.08 each over 440; the band that loses the most in total is over 8R.
| target distance | trades | win | a trade | total |
|---|---|---|---|---|
| 2-3R | 66 | 18.2% | $2.44 | $161.00 |
| 3-5R | 185 | 10.8% | −$2.99 | −$553.00 |
| 5-8R | 191 | 12.0% | −$3.08 | −$588.00 |
| over 8R | 440 | 2.7% | −$10.08 | −$4,434.00 |
Every setting, not the one that suits us. The table below changes with the model you have selected above, and it is the same table those pickers drive on the chart.
The reversion model. In London the reversion model takes 75 trades at −$9.17 each, against −$3.56 with its filters removed, so its filters do not help.
Everything else worth saying
The control is the model against itself. There is no random entry here. Each arm takes the SAME set-ups and removes one of the two things he says is the edge, so the comparison is his model with a part missing rather than his model against something unrelated.
This is a floor on his model, not a measurement of it. He is explicit that step three cannot be automated: "it's really sensitive to market... you cannot just automate it".
What an earlier version of this page got wrong. It moved the stop to break-even at 1R and took half off there. He does neither: break-even comes "when you see that you get an additional breakout", and at the target "we are going to take out not half the position ... the full position". His minimum is 2.5 to 1, not 2. All three are now his.
Nothing was searched. There is no grid of settings on this page and so nothing to correct for. His filter is 30 contracts because that is the number he gives; the sessions, the stop and the target are his too. Where a choice was ours it is named in the limits below.
Micros carry a heavier cost in points. $3.00 a round turn on MNQ is 1.50 points; $4.00 on NQ is 0.20 points. Gross of commission the two contracts run within a point of each other, so on the full-size contract the same model bleeds far less to costs. That is an argument for the size you trade, not for the model.
What this test does not do
We tested the mechanical skeleton of what he describes: the session, the state, the node, the print, the stop, the target and the break-even. We did not test his judgement, and he says that is the part that matters.
- Step three is discretionary and we made it a rule. "a lot of aggression" became "a print of 30 contracts or more within two minutes". That is a reading, and a different reading gives different trades.
- Our choices, named. A low volume node is a price level under 30% of the average volume in the leg, which he does not define as a number. The stop sits two ticks inside the swing, as he says. Break-even is his additional breakout: price taking out the extreme it pulled back from.
- He trades NQ, this is MNQ. 21 sessions of NQ tick data exist against 376 of MNQ. Over the shared dates they behave the same, which is why the micro is used for the long sample.
- No cumulative delta. He also uses CVD building in his favour as a reason to move to break-even early. Only the breakout trigger is used here.
- The replay is a sample. Every number is scored over all 376 sessions of tick data; the chart ships the most recent 116 of them so the page stays light.
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.
Chart Fanatics publish his model in writing and he walks through it on their channel. Top three in the Robbins World Cup futures division, about 500% over twelve months, audited. He trades NQ. Their write-up names the same five steps tested here — market state, location, execution trigger, risk, target.
Read it in their own words first: Auction Market Theory Trading Strategy by Fabio →