🛠️From Hypothesis to a Working Strategy: Closing the Learning Loop with a Trading Journal

A good trading strategy isn't born from intuition — it's built the same way science works: you formulate a hypothesis, test it, and then refine the theory based on the results. Most retail investors skip this process entirely — they try an idea on a live account, it fails once, and they abandon it without ever knowing whether the idea was flawed or the execution was flawed. This article walks through the full cycle: hypothesis → rules → backtest → paper trading → small live positions → journal → iteration — and shows how to practically close this loop.

Step 1: Hypothesis

Every strategy starts with a clearly formulated idea, ideally phrased as "if X, then I expect Y, because Z." A vague hypothesis ("ČEZ is cheap") can't be tested. A testable hypothesis looks like this:

"Stocks in a long-term uptrend (price above the 200-day moving average) that pull back to the 50-day SMA and bounce on rising volume have positive expected value over the following 10 trading days, because this represents a temporary pullback within a strong trend rather than a reversal."

This is a hypothesis that can be measured, rejected, or confirmed with data.

Step 2: Defining rules (entry / exit / risk)

The hypothesis must be translated into mechanical, unambiguous rules:

• Entry: price touches the 50-day SMA (±1%), the closing price is above it, and volume on the bounce day is at least 1.2× the 20-day average.
• Exit (profit): sell at +8% or on the 15th trading day, whichever comes first.
• Exit (loss): stop-loss at -4% from entry price or below the recent swing low.
• Risk per trade: max 1–2% of portfolio capital per trade (so-called R-risk).
• Stock universe: e.g., the S&P 100 plus major Prague Stock Exchange names (ČEZ, Komerční banka) — sufficiently liquid names.

Without precise numbers, a strategy can't be backtested or evaluated repeatedly — this is the key difference between a "feeling" and a "system."

Step 3: Backtesting with out-of-sample validation

The strategy first needs to be tested on historical data — but with one crucial condition: split the data into a training (in-sample) period and a validation (out-of-sample) period.

Example: rules are tuned on 2015–2021 data (in-sample), and only then run unchanged on 2022–2024 data (out-of-sample). If a strategy only works on the period it was "tuned" on, it's very likely overfit — a trap that catches most homegrown strategies built on a single chart.

A model example across four names (illustrative, not a recommendation):

Phase	Period	Number of trades	Win rate	Profit factor
In-sample	2015–2021	142	58%	1.9
Out-of-sample	2022–2024	61	51%	1.3

A decline in metrics between phases is normal and expected — the question is whether the strategy remains consistently profitable (profit factor still above 1), or whether it falls apart entirely.

Step 4: Paper trading

Backtesting can't capture everything — slippage, emotions, execution latency, real-time liquidity. That's why a paper trading phase follows: the rules are applied to the live market, but without real money, typically for 4–12 weeks or until at least 20–30 trades accumulate. The goal isn't to make money — it's to verify that the person can actually follow the rules with discipline in real time.

Step 5: Small live positions

Only after a successful paper trading period does real capital enter — but at a minimized size (e.g., 0.25–0.5% of the portfolio per trade instead of the target 1–2%). The psychology of real money differs from the psychology of paper trading, and no backtest, however good, can substitute for it. This phase tests one's own discipline under real pressure.

Step 6: Evaluation in the journal — key metrics

This is where the loop closes. Every trade (entry, exit, reasoning, emotions, chart screenshot) is logged in a journal, and after a sufficient sample (ideally 30+ trades), four key metrics are calculated:

1. Expectancy (expected value per trade)

Expectancy = (Win % × average win) − (Loss % × average loss)

Example: 45% win rate, average win of 1,200 CZK, average loss of 600 CZK:
Expectancy = (0.45 × 1200) − (0.55 × 600) = 540 − 330 = 210 CZK per trade.

Positive expectancy means a strategy with less than a 50% win rate can still be profitable over time — the key is that the average win multiples exceed the average loss.

2. Profit factor

Profit factor = gross profit / gross loss

A value below 1.0 means the strategy is losing money. Around 1.0–1.3 is borderline, sensitive to costs and slippage. Above 1.5 is generally considered solid, and above 2.0 very strong (rare for realistic strategies and should be verified on a sufficiently large sample).

3. Maximum drawdown

The largest peak-to-trough decline in the equity curve, in percent or currency. For example, a 100,000 CZK portfolio that fell to 88,500 CZK at its worst point has a maximum drawdown of 11.5%. This metric shows how psychologically "painful" a strategy is and how large a position can be safely carried — a strategy with +5% expectancy but a 40% drawdown is practically untradeable for most people because they can't psychologically endure it.

4. MFE / MAE (Maximum Favorable/Adverse Excursion)

• MFE – how much a trade "earned" at its best point before closing (even if it later gave that back).
• MAE – how deep a trade dropped into loss before potentially turning profitable.

These two metrics are the most valuable for tuning exits. If the journal shows that winning trades averaged an MFE of +9% but were actually closed at +5%, it means profits are being cut too early — the solution might be adjusting the take-profit level or using a trailing stop. Conversely, if MAE on losing trades regularly reaches -3% well before the -4% stop is hit, it may mean the stop is set too loosely, exposing capital to unnecessary risk.

Step 7: Iteration — closing the loop

Based on reading these four metrics, the hypothesis from Step 1 gets refined — e.g., the entry threshold shifts, the stop-loss tightens, or the stock universe changes (perhaps the strategy turns out to work only on large-cap names like MSFT and AAPL but not on more volatile smaller companies). The revised version is backtested again on out-of-sample data, goes through paper trading again, and is tested again at small live scale. Through this repeated cycle, the strategy gradually gets refined — the journal acts as the "sensor" that shows exactly where reality diverges from theory.

How to do this in QMA

• /strategy – Strategy Lab lets you formulate entry, exit, and risk-management rules and run them on historical data with separate in-sample and out-of-sample periods, helping verify whether results are simply overfit to the past.
• /journal – The trading journal automatically calculates the key metrics described above (expectancy, profit factor, max drawdown, MFE/MAE) based on logged trades, eliminating the need to compute them manually in a spreadsheet.
• /portfolio – Portfolio Health provides a view of aggregate exposure, correlations between positions, and drawdown at the whole-portfolio level, complementing the per-trade view from the journal with overall risk context.
• /stocks/[symbol] – A stock's detail page (e.g., /stocks/AAPL or /stocks/CEZ.PR) helps verify the current technical and fundamental context when searching for candidates that meet entry rules.
• /smart-money – The institutional flow overview can serve as an additional filter when defining the stock universe for a given strategy.
• /screener – Lets you find stocks that currently meet the mechanical entry criteria defined in Step 2.

Disclaimer

QMA is an analytical and educational tool, not investment advice. Past performance is not a guarantee of future results. All investment decisions are your own responsibility.