Building a Crypto Trading Journal That Actually Gets Used
A useful trading journal records the setup, the planned invalidation, size in R, the outcome and — most importantly — the reason the trade closed. Exit reason is the field that turns a log into an analysable dataset, because it lets you separate strategy problems from execution problems.
Most trading journals fail the same way: they record what happened, which you can already see in your exchange history, and not why, which is the only part that changes future behaviour. This article covers the specific fields that make a journal analysable, the ones that are noise, and how to review it without confirming what you already believe.
- Exit reason is the highest-value field and the one most journals omit entirely.
- Record the plan before the outcome — after the fact, memory rewrites the reasoning.
- Log in R, not dollars, or trades across different account sizes are not comparable.
- A good trade that lost and a bad trade that won are different events; only a journal can tell them apart.
- Review on a schedule with a specific question. Browsing your journal is not review.
Why exchange history is not a journal
Your exchange gives you entry, exit, size and P&L. What it cannot give you is why you entered, what would have made you wrong, whether the exit was planned, and what you were thinking at the time.
That missing half is the entire value. Without it you can compute returns and you cannot answer the question that changes anything: was this a good decision that lost, or a bad decision that happened to work? Those are opposite situations that look identical in a P&L column, and confusing them is how traders reinforce the habits that are costing them money.
The fields that carry weight
Recorded before the trade
| Field | Why it matters |
|---|---|
| Setup / strategy name | Lets you compute expectancy per strategy rather than in aggregate |
| Entry, stop, targets | The plan, as it existed before the outcome was known |
| Risk in R | Makes trades comparable across account sizes |
| Invalidation | What would make you wrong — the field that prevents thesis drift |
| Market context | Trending or ranging; needed for regime analysis later |
| Confidence | Lets you test whether your conviction is calibrated. Usually it is not. |
Recorded after the trade
| Field | Why it matters |
|---|---|
| Exit reason | The single most valuable field. See below. |
| Result in R | The comparable unit |
| Fees and funding paid | Gross expectancy is a fiction; you need net |
| Max favourable excursion | How far it went your way — reveals whether targets are wrong |
| Followed the plan? | Separates strategy performance from execution performance |
| Emotional note | One honest line. Patterns emerge across dozens of entries. |
Why exit reason is the load-bearing field
A journal that records only P&L can tell you a strategy lost money. A journal that records exit reason can tell you which failure is costing you, and those have completely different fixes:
- Stopped out — the thesis was wrong, or the stop was too tight for the instrument's noise.
- Target hit — the plan worked.
- Manually closed early in profit — you cut a winner. If frequent, your targets are unrealistic or your conviction is weak.
- Manually closed early in loss — you cut before invalidation. Different problem, different fix.
- Time exit — the setup expired without resolving.
- Liquidated — a risk-control failure, categorically not a strategy failure.
Consider two traders with identical P&L. One is mostly "stopped out" — a strategy or stop-placement problem. The other is mostly "closed early in profit" — a behavioural problem, and their strategy may be fine. Without exit reason those look the same, and both would go looking for a new strategy when only one of them needs it.
What to leave out
Journals fail from friction more than from missing fields. If logging a trade takes ten minutes, you will stop within a fortnight, and a journal you abandoned is worth nothing.
Skip: long prose narratives (unreadable at volume and unanalysable), screenshots of every chart (rarely reviewed, expensive to produce), indicator values you can recompute, and any field you cannot imagine querying later. If you would never filter or group by it, it is decoration.
Reviewing without confirming what you already think
Browsing your journal is not review. Review means arriving with a specific question and letting the data answer it.
Weekly — execution
What fraction of trades followed the plan? Where did I deviate, and did deviation help or hurt? Any trade I would not take again knowing only what I knew at the time?
That last qualifier is the whole discipline. Judging past decisions with information you did not have then teaches you nothing except hindsight.
Monthly — strategy
Expectancy per setup, net of fees and funding. Distribution of exit reasons. Performance split by market regime. Which setup carried the month, and which quietly bled?
The regime split matters more than most people realise: a strategy that is strongly positive in trends and mildly negative in ranges shows up as mediocre in aggregate. That is a good strategy missing a filter, not a bad strategy — and the aggregate number hides exactly that.
Quarterly — the honest questions
Is my confidence calibrated — do high-confidence trades actually perform better? Is my risk per trade consistent, or does it drift up after wins and after losses? What percentage of losses came from a small number of situations? Is there a time of day, or a state, where I trade measurably worse?
Analyses that repay the effort
Expectancy by setup. Almost always reveals that one setup carries the account while another consumes the gains. Removing the losing one is the highest-leverage change most traders can make, and it requires nothing but the data.
Time-since-last-loss. Group trades by how soon after a loss they were opened. If the ones opened shortly after losses underperform badly, that is a measurable revenge-trading pattern with a structural fix. The data on that window.
Excursion analysis. If winners routinely run far past your target before you exit, targets are too tight. If they routinely reach a level and reverse, they are too far. This is not a judgement call once you have the data.
Cost analysis. Fees and funding as a fraction of gross P&L. Frequently larger than expected, and it decides whether tight-stop strategies are viable at all — round-trip cost in R scales inversely with stop distance.
Plan adherence versus outcome. Compare expectancy of planned trades against deviations. This settles the "my instincts are good" question with evidence rather than argument.
Journaling across Binance, Bybit and Hyperliquid
Traders who run more than one venue usually end up with more than one journal, and the split quietly destroys the analysis. Per-exchange exports are not comparable out of the box, for three specific reasons:
- Fee structures differ, so gross P&L from two venues is not the same quantity. Net is the only comparable figure.
- Funding is charged on different schedules, which changes the cost of an identical position held for an identical time.
- Exit reasons are not exported by any exchange. You get fills; you do not get why the position closed, which is the field that carries most of the value.
Two consequences worth acting on. First, normalise to R before comparing anything across venues — dollar P&L on Binance and on Hyperliquid are different units once fees and funding differ. Second, tag every trade with its venue, because "my strategy stopped working" and "my strategy works everywhere except the venue where my fills are worse" look identical in an aggregate and have completely different fixes.
If you trade the same strategy on several venues, per-venue expectancy is one of the more revealing splits available to you: it separates a strategy problem from an execution-quality problem, and it is invisible in a combined number.
Journaling automated trades
Automation changes what needs recording. The execution details are captured for free; what you lose is the reasoning layer, because there is no moment where a human formed a thesis.
For automated trades, log the strategy version and configuration at the time of the trade (a change three weeks ago that you have forgotten is the usual explanation for a performance shift), the market regime, and any manual intervention — because manual overrides of an automated system are the highest-value entries in the whole journal. They are the exact moment a human decided they knew better, and the aggregate record of how those turned out is unusually persuasive.
Making it stick
The journal that gets used is the one attached to the trading surface rather than living in a separate spreadsheet. Log at entry while the reasoning is present — after the fact, memory reconstructs the plan to match the outcome, reliably and invisibly. Keep the required fields minimal, put optional depth behind a second click, and review on a schedule rather than when you feel like it.
TradeFloor records exit reason on every close from a fixed vocabulary, tracks fees and funding per trade, and computes expectancy per strategy and per regime from that data — so the analysis above is a query rather than a spreadsheet exercise.
Frequently asked questions
What should I record in a crypto trading journal?
Before the trade: setup name, entry, stop, targets, risk in R, what would invalidate the thesis, and market context. After: exit reason, result in R, fees and funding paid, how far it went in your favour, whether you followed the plan, and one honest line about your state. Exit reason is the highest-value field.
Why is exit reason the most important field?
Because it separates strategy problems from behavioural ones. Two traders with identical P&L can have opposite problems — one is mostly stopped out (a strategy or stop-placement issue), the other mostly closes early in profit (a discipline issue whose strategy may be fine). Without exit reason both look the same and both go looking for a new strategy.
Should I journal automated trades too?
Yes, but record different things. Execution details are captured automatically; what you must add is the strategy version and configuration at the time, the market regime, and any manual intervention. Manual overrides of an automated system are the single most valuable entries you will have.
How often should I review my trading journal?
Weekly for execution — did I follow the plan, where did I deviate. Monthly for strategy — expectancy per setup net of costs, exit-reason distribution, performance by regime. Quarterly for the harder questions about calibration and risk drift. Arrive with a specific question; browsing is not review.
Should I log trades in dollars or R?
R — multiples of the amount risked. A $50 loss on a small position and a $200 loss on a large one are both −1R, and treating them as the same event is what makes trades comparable across time and account sizes. Dollar amounts make a growing account look like an improving strategy.
How do I keep a trading journal across Binance, Bybit and Hyperliquid?
Normalise to R rather than dollars, because differing fees and funding schedules make dollar P&L from two venues incomparable, and tag every trade with its venue. Per-venue expectancy is one of the more revealing splits available: it separates a strategy that stopped working from one whose fills are simply worse on a particular exchange, and that distinction is invisible in a combined number.
What makes traders abandon their journal?
Friction. Long prose entries, screenshots of every chart, and fields nobody ever queries turn a two-minute task into a ten-minute one, and the habit dies within a fortnight. Keep required fields minimal, log at entry while the reasoning is still present, and put optional depth behind a second click.
Revenge Trading: The 40-Minute Window That Eats Returns
What the time-since-last-loss analysis reveals, and why willpower is the wrong fix.
// questions or corrections · [email protected] · more essays · /blog