Replay, the 100-trade sample and scaling up

The last lesson answers the three questions this course has deferred: how to practise without paying market costs, when a setup has been tested enough to judge, and when size can go up. The answers are a replay plan, a sample threshold, and a scaling rule that depends on weeks rather than on feeling ready. Together with the tilt protocol at the end, they are the difference between running this playbook and merely knowing it.

Replay practice

Bar replay in a charting platform, or tick replay through a broker's simulator, lets you trade a past session as if live, bar by bar, without knowing what happens next. It is the cheapest way to build the sample: no commissions, no slippage, no daily limit, and twenty sessions in the time the live market gives you five.

The plan:

Phase	Sessions	Focus	Done when
1. Day-type reads	20	Fill the 10:30 sheet only; no trades	70%+ accuracy
2. One setup	30	Trade one setup only, full rules, log everything	30 A-grade trades logged
3. Second setup	30	Add the second setup	30 A-grade trades in it
4. Full playbook mornings	40	9:30 to 11:00, all eligible setups, all rules	100 A-grade trades total

Rules that keep replay honest:

Do not peek. If the platform lets you see the day's outcome, it is not replay. Use sessions you do not remember; last month's are too recent.
Apply the real rules. Trade cap, consecutive-loss stop, 11am rule. Replay without limits builds the wrong habits.
Subtract cost. Add your product's cost per trade in R to every result. Replay results without cost are fiction.
Log identically. Same fields, same grade column. The replay log and the live log should be one spreadsheet with a "replay" flag.
Mix regimes. Ten sessions from a quiet month, ten from a volatile one. A playbook tested only on one regime will meet the other one live.

Replay overstates results in one known way: fills. Assume you get the worse side of every bar at entry and the stop fills a tick beyond the level. It understates in another: you cannot feel the pressure. Neither is a reason to skip it.

The 100-trade sample

No setup is judged, and no size is changed, before 100 A-grade trades in it, or 100 across the playbook if the question is the playbook as a whole. Not 20, not 50.

The reason is variance. Take a setup with a true expectancy of +0.25R, a 50% win rate and typical intraday outcomes. Over 20 trades, the observed expectancy will land anywhere from about −0.4R to +0.9R most of the time. Over 100, from about −0.05R to +0.55R. Over 300, from about +0.08R to +0.42R.

Trades	True +0.25R setup: usual range of observed expectancy	Chance a truly break-even setup looks like +0.25R or better
20	−0.4R to +0.9R	about 20%
50	−0.15R to +0.65R	about 10%
100	−0.05R to +0.55R	about 4%
300	+0.08R to +0.42R	under 1%

At 20 trades you cannot distinguish a good setup from a break-even one, and one in five break-even setups will look good. At 100 you mostly can. The Brazil study's persistent losers were, in effect, people who judged at 20 and kept going.

What the sample decides:

Expectancy above +0.15R after cost: keep the setup, continue, eligible for scaling.
Between 0 and +0.15R: keep at current size, look for a filter (day type, time window, delta) that separates the winners; re-test 100 more.
Below 0: remove the setup from the sheet. Not "trade it more carefully". Remove it.

Scaling up: the green-weeks rule

Size goes up on evidence, by a fixed step, and comes down on the same schedule.

Increase per-trade risk by 25% (for example, 0.33% to 0.42%) only when all of the following hold:

100 A-grade trades in the live log at the current size.
Net expectancy after cost above +0.15R over those trades.
Eight of the last ten weeks green on net R, with no week below −3R.
F-grade trades under 10% in the last four weeks.
No rule breach in the last four weeks.

Decrease per-trade risk by 25% immediately when either holds:

Three consecutive red weeks.
A single week below −5R, or any daily limit breach.

Return to the previous size only by meeting the increase conditions again.

  Scaling ladder, per-trade risk as % of account

  0.25%  ----  start (or after a decrease)
  0.33%  ----  after 100 trades, 8 of 10 green weeks    ^ each step: same conditions
  0.42%  ----  after the next 100 and 8 of 10                  again, no shortcuts
  0.50%  ----  ceiling for this playbook                       v drop a step on 3 red weeks

The ceiling matters. Beyond about 0.5% per trade with a three-trade cap, the daily limit exceeds 1.5% and the monthly drawdown arithmetic from Module 7 stops working. More money is made from here by a larger account, not a larger fraction.

The rule is slow on purpose. A trader starting at 0.25% who does everything right reaches 0.5% in roughly a year. The Brazil study's numbers say that year is the cheapest year of a day trading career; the ones who skipped it did not have a second.

Tilt protocol

Tilt is the state in which decision quality drops after losses or after a big win. Trading Psychology covers what it is and why. The intraday protocol is mechanical:

Trigger	Response
Two consecutive full stops	Flatten, close platform (Module 7 rule)
Any F-grade trade	Stand up, five minutes away from the screen, then one more trade maximum
Daily limit hit	Done; no chart until the review
"Get it back" answer is yes on the checklist	Half size or no trading
Three F-grade trades in a week	Half size for the following week
Any rule breach	Two days off, then half size for a week

None of these depend on noticing that you are tilted, which is the point. They fire on observable events.

The course in one page

Most day traders lose; you are not the exception until 100 trades say you might be.
One product, one 90-minute window, at most three setups.
Prep before 9:30, read the day at 10:30, trade what the day type allows.
Bracket every entry, size from the structure stop, never chase past the ceiling.
Daily limit, per-trade at a third of it, three-trade cap, two-loss stop, done at 11:00 if red.
Checklist before, 20-minute review after, replay in between.
Judge at 100 trades, scale on eight green weeks in ten, drop on three red.

Key idea: Practise in replay until the reads and setups are habits, judge nothing before 100 A-grade trades, and change size only on a weeks-based rule. The playbook is the easy part; this lesson is the part that makes it survive contact with the base rate.

Try it: Set up the replay log today, with the phase 1 sheet ready, and complete two day-type reads before the next session. Put a date in your calendar 100 A-grade trades from now, estimated at your planned pace, and do not compute the playbook's expectancy before it.

Recap

Replay in four phases: day-type reads, one setup, a second setup, full mornings; 100 A-grade trades before judging.
At 20 trades a break-even setup looks good one time in five; at 100, one in twenty-five.
Keep setups above +0.15R after cost, filter those between 0 and +0.15R, remove those below 0.
Increase per-trade risk by 25% only after 100 trades, eight of ten green weeks, under 10% F-grades and no breaches; decrease on three red weeks or a −5R week.
Tilt protocol fires on observable events, not on feeling tilted.