Testing ICT concepts in the backtester and journal
Lesson 16 · 5 min read · By DT Terminal
How to measure whether one ICT concept works for you: a written rule, bar-by-bar replay with every setup logged, sample size, a control as the base rate, tags for each concept, and the Backtest vs Live edge card.
Unlock the full lesson in the app
Worked examples, the quiz and the interactive tools are in the app. Free account.
Write the rule so a stranger could mark it
An ICT concept is only a word until you write it as a rule. An operational definition says exactly what counts, so two people marking the same chart get the same list. Here is one for the order block from Order blocks:
| Part | Rule |
|---|---|
| Order block | The last down-close M15 candle before a candle whose body closes above the most recent short-term high |
| Zone | That candle's open to its low |
| Entry | A buy limit at its body's 50% |
| Stop | The zone's low minus 2 pips |
| Target | 2R |
| Window | Only 02:00 to 05:00 New York time |
The written rule
What counts, and what does not
- Bullish order block
- Bearish order block
- Earlier low or high
Bullish order block
last down-close candle; zone open to low
Bearish order block
last up-close candle; zone open to high
Counts
a body closes above the STH
Wick only
no close above: not counted
Counts, sweep first
same rule, tagged sweep: yes
Replay, and log every setup
Scrolling finished charts tests how well you find examples, as ICT and your psychology showed. Replay hides the future, so you mark each setup before seeing the result.
Log every setup the rule allows, skips included, or your mood picks the sample.
Worked example
Example 01: One setup, one row
The worked example, with every number walked through, is in the app.
Replay one rule and tag every setup
Go deeper
When you close a trade, add tags for the concept, the window and whether a sweep came first.
How many setups before a number means anything
Say your rule wins 40% at 2R. Breakeven at 2R is 33.3%, so it looks like an edge. The table gives the likely range (the 95% Wilson interval) for that 40% at four sample sizes.
| Setups | Likely win rate | Clearly above breakeven? |
|---|---|---|
| 10 | 16.8% to 68.7% | Not yet |
| 30 | 24.6% to 57.7% | Not yet |
| 100 | 30.9% to 49.8% | Not yet |
| 300 | 34.6% to 45.6% | Yes |
A fair first look needs about 30 setups, and a number worth trusting 100 or more. Why small samples swing is in Day 9, The math behind trading, and the same range is in Day 20, Backtesting your trades.
Compare it with a control
A win rate alone says little. The base rate is what the same entry, stop and target make without the concept: say, a limit at the previous candle's 50%, in the same window, on the same days. That is your control. If the concept cannot beat it, it is not adding anything for you.
Worked example
Example 02: The order block rule against its control
The worked example, with every number walked through, is in the app.
Tag every trade by concept
Tags turn one log into many tests. Give every trade, in replay and live, the same three: the concept (ob, fvg, breaker), the window (london, ny-am) and whether a sweep came first (sweep-yes, sweep-no).
Then change one thing at a time. Compare sweep-yes against sweep-no on the same rule before adding a second filter. Change two at once and you cannot tell which did the work.
Read each tag on its own
Does it survive live trading?
A rule that works in replay can still fail live: slippage, missed fills, nerves. The backtester's Analytics tab has a Backtest vs Live edge card for each instrument you both backtest and trade. It sets win rate, profit factor and expectancy side by side, and reads Edge holds when the live profit factor is at least 70% of the backtest's.
Trades synced from your broker are recorded as they happened and cannot be edited later, so that comparison runs against a record you cannot flatter. Before a challenge account, check its rules in Day 25, Prop firms and brokers. The whole cycle is in Day 26, The loop that makes you profitable, and reviewing trades in Day 24.
Check yourself
Quick review · 0 of 0
Five questions on testing a concept. If one trips you up, reread that section.