Choosing your entry family

It is 21:40 ET (09:40 ICT, winter). You have just finished M6.01 to M6.06 and NQ is drifting in a 12-point overnight range. Your plan for tomorrow lists all four entry families, so whatever the market does, you have an entry for it. That feels like flexibility. In practice it means every trade is a different experiment, and after a month your journal cannot tell you which family works for you or why it failed. This lesson is about choosing one family first, pricing it honestly, and knowing what your own sample can and cannot prove.

Key term

An entry family (nhóm vào lệnh) is a group of entry models that commit at the same moment of the auction and pay for it in the same way: breakout, pullback/retest, rejection or continuation (M6.01) Definition. Specialising (chuyên môn hóa) means trading one family, with one written trigger, invalidation and stop, until you have a logged sample of it.

Term you need: the break-even win rate (tỷ lệ thắng hòa vốn) is the win rate at which a stop, a target and your costs net to zero. With the stop as 1R, the target as k R and the round-turn cost as c R, it is (1 + c) ÷ (k + 1) Definition. M6.08 owns the full treatment; here it is the tool for comparing families.

Why one family first

Each family fails in its own way, and recognising your family’s failure while it happens is a perceptual skill that takes repetitions. The breakout trader learns to spot the false break; the retest trader learns to let a missed move go; the rejection trader learns to sit through an adverse spike to a stop that was set in advance; the continuation trader learns to tell a pause from chop. Splitting fifty trades across four families gives you about a dozen of each, which is too few to see any family’s pattern.

Our rule is one family, traded on sim to at least 50 logged trades with the same written trigger, invalidation and stop, before you add a second House rule. That is a process choice, not a finding: our lab has not measured whether specialists do better than generalists, and nobody can promise that the family you pick has an edge. What specialising does guarantee is a journal sample clean enough to be scored.

Three matches: setup, session, temperament

Match the family to the setup (M5). This one is close to a definition: a setup tells you what the auction should do next, and the family should commit to that, not to its opposite Definition. A fade setup at an extreme is traded with rejection or retest entries; a break from balance is traded with breakout or pullback entries; an established trend is traded with continuation entries. Remember that the setup’s level is where you choose to look, not a reason price turns: our lab found pre-market levels, prior-day highs and lows included, respected no more often than random prices at the same distance Measured: no edgeNasdaq index (M6.03).

Match the family to the session. The overnight session is quiet: in our lab’s NQ tick recorder (60 sessions, Aug to Nov 2026) each Asia-hours clock hour, 18:00 to 03:00 ET (06:00 to 15:00 ICT, winter), traded about 6 to 13% of the volume of the busiest cash-session hour MeasuredNQ. That is a fact about volume and the liquidity you will be filled in. What it does to each family’s odds is a different claim.

Myth / hypothesis: “Quiet overnight hours reward fades and retests; the cash open rewards breakouts and continuation.” Lore It sounds like auction logic, and you may test it on your own journal. The nearest public tests on NQ did not settle it: levels showed no measurable difference between overnight and the cash session (−0.1 percentage points, 95% CI −3.1 to +3.0, n = 8,798), and lunch-hour breakouts did not fail measurably more often than morning ones (+1.5, CI −2.4 to +5.4, n = 2,512). Both are inconclusive, from one practitioner’s study. Treat a family that bleeds in one session as a hypothesis to log, not a regime law.

Match the family to your temperament (tính cách). These pairings are our heuristic for picking the family you will actually execute when it is uncomfortable House rule:

Why it is still worth knowing

None of the three matches is a measured edge. They are worth doing because a mismatch shows up directly in your journal: stops that come from the entry fighting the setup, or rules you skip because they feel wrong. A fit you can execute for 50 trades produces a sample you can score; a family you abandon at the first drawdown produces nothing.

Price the family before you pick it

Families differ in what you pay, and costs are measured in money, not in feel. Every family pays commissions and fees. Stop-order families (breakout, most continuation entries) also pay slippage, because the order becomes a market order once the trigger trades Definition. Limit-order families (rejection, retest) pay no slippage, but they pay in two quieter ways: the trades that run away without filling you, and the fills you do get inside a fast sweep, which tend to be followed by more movement against you PlausibleOther market. That last one is a well-understood mechanism in market microstructure, not yet measured on NQ by our lab.

How much costs matter: the clearest public breakout test on the Nasdaq-100, an opening-range breakout, made about +0.13R a trade before costs and about zero after them PlausibleNasdaq index (a working paper and one replication; NQ futures extrapolated). An edge that size disappears inside a family’s costs.

What the data says

Worked example: the mismatch, priced

Synthetic journal, numbers chosen to illustrate the arithmetic. A trader’s setups are all fades at the extremes of a range. Because breakouts “feel safer”, they enter with buy-stops and sell-stops through the level: stop 10 NQ points (1R = $200), target 15 points (1.5R), costs $15 round turn plus about 1 tick of slippage on the stop entry, 0.10R a trade in all.

• 40 breakout entries: 14 wins, 26 losses, a 35% win rate. Expectancy = 0.35 × 1.5 − 0.65 × 1 − 0.10 = −0.225R a trade, against a break-even of (1 + 0.10) ÷ 2.5 = 44%. The losses are mostly false breaks that returned inside: the entry was fighting the setup.
• They switch to rejection entries that match the fades: a limit entry back inside the range after the probe fails, same 10-point stop, target 20 points (2R), costs 0.075R. 40 trades: 17 wins, a 42.5% win rate. Expectancy = 0.425 × 2 − 0.575 × 1 − 0.075 = +0.20R, against a break-even of 35.8%. Six further signals ran away without filling them; those trades are not in the sample.

Now read the sample honestly. With 40 trades the 95% interval around 42.5% is about ±15 percentage points, roughly 27% to 58% Definition. That interval includes the 35.8% break-even, so the journal does not yet show a positive expectancy. What did change at once: the losses now have one explainable cause (initiative resuming through the fade), which is what makes the next 40 trades worth logging. The fit made the sample readable; it did not prove an edge.

The mirror and the failure

The mirror. A trader whose setups are breaks from balance enters with limit orders at the extreme of the range “to get a better price”. Every limit fill is against the very break the setup predicts, so the trades that work are the ones the setup said would not happen. The fix is the same: the breakout or pullback family the setup was written for.

The failure. A family that fits your setups and temperament can still lose, because the fit is not an edge. If after 80 to 100 logged trades the interval sits below your break-even, the problem is upstream: the setup itself, or costs too large for the target. Change the setup or the bracket, not the family, and do not switch families every time a streak goes against you.

On a prop evaluation

Evaluation accounts are simulated, and simulated fills are not live fills: a limit-order family can look cheaper on sim than it will be live. Log your fills as if they were live, and read the firm’s rule on simulated fills before you rely on them; M6.06 covers order handling and that rule.

Common mistake: Choosing the family with the best-looking backtest or hit rate instead of the one that matches your setups and that you can execute. A rejection family traded by someone who cannot sit through adverse spikes gets abandoned at the first drawdown; the family you cannot execute has no expectancy at all.

Common mistake: Treating a win rate from a small sample as an edge. Forty trades put about ±15 percentage points around the rate; compare the interval with the break-even win rate after costs before you call a family “working”.

Common mistake: Comparing families on win rate alone. A 60% family with a 1:0.7 bracket and stop-entry slippage can be worse than a 40% family with a 1:2 bracket. Compare expectancy after costs.

Drill: price your own families

Take your last 30 or more journaled trades (sim is fine).

1. Tag each trade with its family: breakout, pullback/retest, rejection or continuation. Trades you cannot tag are impulse trades; count them separately.
2. For each family, compute the win rate, the average win and the average loss in R, and your average cost per trade in R (commissions plus slippage).
3. Compute each family’s break-even win rate, (1 + c) ÷ (k + 1), and its expectancy.
4. Compute the 95% interval around each win rate: ±1.96 × √(p × (1 − p) ÷ n).
5. Pick the one family you will trade for the next 50 logged trades, and write one sentence on why it matches your setups.

Recap

• Trade one family on sim to at least 50 logged trades with the same written rule before adding a second; this is a house rule for getting a scorable sample, not a measured edge.
• Match the family to the setup (close to a definition), and treat session and temperament matches as hypotheses and heuristics you log, not as laws.
• Price each family: commissions, slippage on stop entries, missed and adversely selected fills on limit entries. Compare families on expectancy after costs, using the break-even win rate (1 + c) ÷ (k + 1).
• Read your sample with its interval: 40 trades carry about ±15 points, so a fit can make losses explainable long before it can show an edge.

Quiz

Further reading

• Order Flow, Measured: all claims: the public practitioner study behind the session tests in the myth box (claims TE-01, TE-06) and its break-even cost line. One practitioner’s pre-registered study; intervals not clustered by session; at most Plausible.
• Glosten and Milgrom (1985), Bid, ask and transaction prices in a specialist market with heterogeneously informed traders: the classic paper on why passive fills are adversely selected.
• Our lab: pre-market level families, 226 pre-registered tests on 721 sessions, were respected no more often than random prices at the same distance; and the opening-range breakout on the Nasdaq-100 is about +0.13R before costs and about zero after (working paper plus one replication; our own replication is pending).

Prerequisites: M6.01 Setup vs entry, M6.05 Continuation entries and M6.06 Order types for entries. Next: M6.08 Stops & targets, which turns the family you chose into a stop, a target and a break-even number after costs.

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