Pullback and retest entries

It is 11:20 ET (23:20 ICT, winter). ES broke above yesterday’s high at 6,010 at 10:05 ET, was accepted above it, and ran to 6,022 without you. The M6.02 entry was clean, but you hesitated. Now price is drifting back toward 6,010. Is this a second chance at a better price, or the start of a failed breakout? This lesson gives you a written way to tell the two apart before you click, and an honest account of what the better price does and does not buy you.

Key terms

Pullback (hồi lại): after an accepted break or an impulse, price moves back toward where the move started without re-entering the old range Definition.

Retest (kiểm định lại): the first return to the broken level itself after acceptance. It is an event you wait for, not one you predict Definition.

Location family (nhóm vào lệnh theo vị trí): of the four entry families in M6.01, the one that trades the same idea as the breakout entry, only closer to its invalidation.

Term you need: acceptance means time, volume and closes beyond the level, scored with checks written before the probe (M4.01, M6.02). No acceptance, no retest: only a probe.

Chart first: accepted, pulled back, retested

How it works

The retest checklist

A retest entry asks the auction two questions in order. Both are written before the session.

1. Was the break accepted? Your M6.02 acceptance checks passed. If not, there is nothing to retest.
2. Did the return hold? This course’s trigger is a house rule House rule: price trades within 1 point (4 ES ticks) of the broken level, and then two consecutive 5-minute bars close back beyond it.

Everything else you see on the touch, a long wick, the footprint at the low, delta turning your way, goes in the journal, not in the trigger. Delta explains the price move of its own bar, not the next one MeasuredOther market (measured on stocks; M3.04 has the evidence). Whether order-flow confirmation at a level adds anything beyond the level is not yet measured; M4.09 sets out the rule our lab will test.

Retest or failed breakout

An orderly drift back, overlapping bars and shrinking ranges, gives the two closes a chance to print. A vertical bar that slices straight through the level and closes back inside the old range is the auction changing its answer: that is M5.01’s and M6.04’s territory, and the long idea is invalid. The checklist does this sorting for you; whether the sorting produces an edge is not yet measured on ES or NQ Lore.

The stop, and what the better price does not buy

The stop goes a little beyond the retest extreme, and the rule exit is acceptance back inside: a 30-minute close back below a broken high House rule. Because the entry sits close to that invalidation, the stop is small and the reward-to-risk looks large Definition. That is arithmetic about payoff, not about how often you win.

Two things the arithmetic hides:

• A filter changes the sample. You never buy the breaks that run without looking back, and a pullback to the level is also what a failing breakout looks like at first. Retest trades are a different sample from “all accepted breaks”, with an unknown win rate (M5.07 works the numbers for the setup).
• A tight stop beside a level is not protected by the level. In our lab’s recent NQ and ES futures check, prior-day highs, lows and closes were broken through more often than random prices at the same distance Measured: no edgeNQ. A stop 2 points beyond a famous line is still a stop 2 points beyond a price.

The costs: missed moves and adverse fills

Strong breaks often do not come back. How often an accepted ES or NQ break returns to within a point of the level before moving on is not yet measured, so treat any fill rate you hear as a guess and log your own. If you rest a limit at the level instead of waiting for the trigger, a second cost appears: a passive order tends to fill when aggressive orders are pushing through it, which is exactly when the level is breaking PlausibleOther market (the markout idea in M6.06; a market-microstructure mechanism, not yet measured on NQ).

Split entries. Half on the breakout, half on the retest is a legitimate plan only when both sizes, both triggers and one shared invalidation are written before the open House rule. Adding the second half mid-trade because price came back is hope with a limit order.

Why it is still worth knowing

• Shared vocabulary. “Accepted above 6,010, retest held on two 5-minute closes” can be checked after the session; “nice pullback” cannot.
• Framing risk. The retest puts the invalidation near the entry, which keeps the loss per trade small and the size decision honest (M6.08).
• A testable hypothesis. Every part is a written rule, so you can score it on your own sessions against a random-level version (see the drill).

Myth / hypothesis: “Retests only work at real levels: yesterday’s high, the overnight high, the IB edge. A random swing high is just a dip.” Lore The levels themselves were tested, and they held no better than random prices at the same distance (box below). Mapped levels are still worth using, for a different reason: they let you write the plan before the session, so you are not inventing the level after price reacts.

What the data says

The retest entry has not been tested as a rule. What has been tested is the premise behind “retests need real levels”: that well-known levels hold better than other prices.

So the retest is a way to locate an entry and a stop, not a measured edge. Any edge has to come from your rule and your execution, and your journal is where you find out.

Worked example: the long side

Synthetic session (no real date), priced for 1 ES ($12.50 a tick, $50 a point; 1 MES is a tenth of that). It shows the reading, not a signal.

• Before the open. Yesterday’s high 6,010 is on the map with an if-then plan for both sides. Acceptance checks from M6.02: 30 minutes above, 1.5× usual volume, two 30-minute closes above. Retest rule: within 1 point, then two 5-minute closes above.
• 10:05 ET (22:05 ICT), probe. First trade above 6,010. No verdict.
• 10:30–11:00 ET, verdict. 30-minute closes at 6,014.50 and 6,020.75, nearly an hour above, volume about 1.6× usual. Accepted. Price is 10 points above the level: no chase.
• 11:00–11:30 ET, pullback. 6,022 high, then 6,018, 6,014, 6,012.25 on overlapping bars. The 11:25 bar wicks to 6,009.00 and closes at 6,011.00. The 11:30 bar closes at 6,011.50. Touch within 1 point, two closes above: trigger met.
• The plan. Long 1 ES at 6,011.50. Hard stop 6,007.50, 1.5 points under the wick: 4 points = 16 ticks × $12.50 = $200 = 1R. Rule exit: a 30-minute close below 6,010 means flat at market, even if the stop has not traded. Target 2R at 6,019.50 (+8 points, $400) House rule; the 6,022 high is the next mapped price if you trail a runner.

The chaser. Buying the 11:00 close at 6,020.75 with the same invalidation means a stop at 6,007.50: 13.25 points ($662.50) of risk. The same 8-point gain is about 0.6R for the chaser and 2R for the retest. Same idea, same level: the entry location changed the arithmetic. It did not change the unknown win rate.

The mirror and the failure

Short side. Turn the session over. Yesterday’s low 5,990 breaks at 10:05 ET and is accepted below by 11:00. The pullback wicks to 5,991.00, then two 5-minute bars close below the level, the second at 5,988.50. Short 1 ES at 5,988.50, hard stop 5,992.50 (4 points, $200), target 2R at 5,980.50, flat on a 30-minute close back above 5,990.

Failure 1: the slice. Back to the long. Instead of drifting, one 5-minute bar drops from 6,021 straight through 6,010 to 6,004 and closes there. No touch-and-hold, no two closes: no entry. That is a failed breakout, and the long idea is dead.

Failure 2: the retest held, the breakout did not. You are long at 6,011.50. At 12:00 ET the 30-minute bar closes at 6,009.25. The hard stop did not trade, but the rule exit did: flat at about 6,009.25, a loss of 2.25 points ($112.50, about 0.56R). Nothing was misread; the retest described ten minutes, and the close back inside is newer information.

Failure 3: no retest. Price never comes back within a point of 6,010 and closes the day at 6,034. You made nothing. That is the family’s price, and your log should count these days too.

Prop accounts and resting orders. In a prop-firm evaluation your fills are simulated, so a fill at the exact level there does not prove a live order would have filled (MKT2.03). Firms also ban strategies that exploit how the simulator fills orders. Read the firm’s prohibited-conduct rules before you rest limits at levels or automate this entry (PROP1.04).

Common mistake: Calling every dip a pullback entry. A retest needs a prior accepted break; without one you are buying weakness and naming it after a model. Check the break first, every time.

Common mistake: Reading a good reward-to-risk as an edge. 2R from a retest against 0.6R from a chase is arithmetic. Whether retest entries win often enough to beat the breakout entry, or no trade, is not yet measured on ES or NQ.

Common mistake: Trusting a stop because it sits behind a famous level. Our lab found prior-day highs, lows and closes broken through more often than random prices at the same distance. Size the stop for the distance, not for the level’s name.

Drill: retests at mapped levels vs random levels

In market replay (PLAT2.06 shows the ATAS steps), work through 20 recent ES or NQ sessions:

1. Before each session, write the map (prior-day high and low, overnight high and low), your acceptance checks and the retest rule.
2. Also write one random price per session: a level at the same distance from the open as your nearest mapped level, on the other side.
3. For every accepted break of either kind, log: did a retest within 1 point happen, did the two closes come, and what came first, 2R or the invalidation? Log breaks that never retested too.
4. Compare the two columns.

What we don’t know yet

• How often accepted ES or NQ breaks come back within a point of the level before moving on: not yet measured.
• Whether retest entries beat breakout entries, or no trade, after costs: not yet measured.
• Whether a level holds better on its return after a break than on its first touch: not yet measured; the lab’s level test scored first touches only.
• Whether order-flow confirmation at the touch adds anything beyond the level: not yet measured (the M4.09 rule is written for a forward test).

Recap

• A retest entry needs an accepted break first; then a written trigger, a touch within a point and two closes back beyond, separates a retest from a failed breakout.
• The entry sits near its invalidation, so the stop is small and reward-to-risk is large; that is arithmetic, and the win rate is not yet measured.
• The family’s costs are breaks that never come back and resting limits that fill when the level is breaking.
• Prior-day and overnight levels held no better than random prices at the same distance in our lab, so mapped levels are for planning ahead, not proof the retest holds.

Quiz

Further reading

• Footprint Charts: How to Read Order Flow (Jigsaw): reading the touch bar print by print, for your journal.
• How To Trade A Market Profile Ledge (Axia Futures): what holding or failing at a level means in auction terms.
• Our lab: in 721 sessions of QQQ and SPY 1-hour bars and 46 sessions each of NQ and ES futures, prior-day and overnight levels held no better than random prices at the same distance; the retest of a broken level was not scored.

Prerequisites: M6.01, M6.02 and M4.01. Next: M6.04 Rejection entries, the family that trades the failed break this lesson tells you to skip. Connects to: M5.07 (the breakout–retest setup) · M4.07 (the level map and its placebo test).

Educational content only, not financial advice. Futures trading involves substantial risk of loss. Synthetic examples are illustrations, not trade calls.