Trend following and breakout strategy:
Write your view as verifiable rules.

1. First clarify: trend‑following, breakout and momentum are not synonyms

1.1 Trend following reacts, it does not prophesy

Trend following typically increases exposure only after price has shown direction. It accepts delayed signals in exchange for rules that are easy to describe and test and that don’t require predicting turning points in advance. The most common costs are whipsaws during consolidation, lagging signals, and late exits at the end of trends. This payoff shape — many small losses and a few large winners — differs from high‑win‑rate mean‑reversion strategies.

Academic research typically places cross‑asset time‑series momentum — which judges direction using each asset's own past returns — in the trend‑following context. Moskowitz, Ooi and Pedersen observed return continuation over one to twelve months across 58 liquid instruments (equity indices, FX, commodities and bond futures); the study also notes partial reversal over longer horizons. AQR's long‑historical work traces time‑series momentum back to 1880 to test whether trend‑following's recent success was accidental. These are historical results and do not guarantee any personal strategy will be profitable going forward.

1.2 Breakouts are a signal form; the strategy still needs the surrounding context

A breakout (breakout) usually compares the current price against highs, lows, channels or structures over a lookback period. Price crossing the upper boundary can generate a long candidate signal, crossing the lower boundary can generate a short candidate signal; but “crossing” itself does not tell you whether to use intraday price or close, which bar to execute on, where to put the stop, how to handle gaps, or when to stop trading. Thus a breakout is an entry to the rule set, not a complete strategy.

1.3 First write the return profile of the strategy

For this return shape you should also compute the break‑even win rate: |average loss| ÷ (average gain + |average loss|). If a strategy relies on a few large winners, the break‑even win rate is only a theoretical threshold — actual trading costs, slippage, consecutive losses and max drawdown still need scrutiny.

2. First determine market environment: the same rule won’t work the same in all markets

2.1 Distinguish trend vs. consolidation using observable conditions

“There’s a trend now” cannot be only a subjective feeling. Define the environment using relative positions and slopes of moving averages, channel width, volatility regime, or recent high/low structure. For education, an environment condition might be: SMA50 above SMA200 and SMA200 not declining recently; this is not a universally optimal parameter set, but a falsifiable assumption that can be backtested and sensitivity‑analysed. If parameters only work at 49/203, suspect overfitting.

2.2 Market selection should put execution costs front and center

Daily volume is only a starting point. Also record spread, depth, trading hours, event‑period behaviour, executability when stops trigger, shorting restrictions, contract multiplier and quote currency. Small stocks, obscure futures or crypto markets with thin weekend liquidity can hide slippage risk under calm historical curves. Group markets — don't mix different product types into a single average.

2.3 Trend following suits multi-market observation, but diversification is not risk-free

Rules across equity indices, rates, FX and commodities can reduce reliance on a single market failure, but correlations may rise during liquidity squeezes or macro shocks. Different markets also have different trading hours, gaps, margin and roll rules. Proper diversification is listing each market's risk units, common factors and worst‑case scenarios — not just increasing the number of instruments.

Before execution you can usePre‑trade rules checkerConfirm environment, events, costs, positions and non‑trading conditions are documented in writing; this completeness only indicates process completeness, not signal validity.

3. Establish breakout rules: decompose "crossing" into observation, confirmation and execution

3.1 Reproducible definition of the Donchian channel

Using an N‑period Donchian channel of K bars as an example: to avoid leaking current‑period data into thresholds, first compute the upper and lower bounds using the previous N completed K bars:

If N = 20, `Upper20(t)` is the highest high of the 20 completed candles prior to the current close, excluding the candle currently forming. This detail decides whether a backtest used future information. You can add volatility filters or long MA conditions, but each filter must be fixed in the data contract and separately checked that gains are not just parameter mining.

3.2 Close confirmation and intraday cross are two different systems

If the rule says “current close > highest high of previous 20 bars”, the signal is only known at close; if you use next‑bar open fills you must record at least one‑bar information delay. If the rule says “intraday price crosses the upper channel and place stop‑buy”, execution may happen earlier, but price‑gaps, nonfills and slippage models differ entirely. You cannot confirm a signal at close and then assume perfect fills at intraday low or close.

Investor.gov explains market orders normally guarantee execution but not execution price, limit orders set price boundaries but may not execute, and stop orders convert to market orders when the stop price is reached.[3] FINRA also warns that delayed quotes, execution time and volatility will cause actual execution prices to differ from quoted prices.[4] Therefore, “signal is correct” and “can execute at ideal price” must be considered separately for breakout trades.

3.3 Breakout buffer is not a free improvement

To reduce the chance of immediate reversals upon touching the channel, you can raise the threshold to `Upper_N + buffer`, or require the close to be a certain proportion above the channel. A larger buffer may reduce noise but can delay entries, forfeit part of a trend, and increase chase distance. Buffer settings should be presented in a parameter sensitivity table rather than selecting a single best‑looking result.

4. Orders & execution: strategy assumptions are most easily self‑deceived at execution

4.1 Signal, order submission, execution must have three timestamps

A trading journal should at minimum store signal formation time, order submission time and execution time. If a signal forms at 10:00:00, the order is sent at 10:00:02 and filled at 10:00:03, those three timestamps let you identify strategy latency, platform latency and market slippage respectively. A single "entry price" field cannot distinguish rule failure from execution deviation.

4.2 Slippage and costs should be modeled by direction

Long breakouts are usually bought on advances; adverse entry slippage means you buy at a higher price. Long stop exits are sold on declines, i.e., you sell at a lower price. Short breakouts open by selling cheaper; covering is bought back at a higher price. Fees should be calculated on the actual executed amount; fixed fees, taxes, stock borrow costs and rollover costs must be listed separately and cannot be replaced by a vague "estimated cost" that lumps all frictions together.

Can start withTrading expectancy calculatorRun cost sensitivity, then paste the R result from the actual executed trade.Trading journal analyzerThe former tests plan assumptions; the latter analyzes realized outcomes — they are not interchangeable.

4.3 No‑trade rules are part of execution design

Breakout strategies can predefine non‑trading conditions: spread above the cap, breakout occurring during low‑liquidity periods, stop distance causing position to fall below executable minimum, no acceptable gap handling before/after major events, daily loss limit reached, or signal incompatible with exchange/broker rules. These conditions are not for cherry‑picking losing trades afterwards but must be fixed before order placement.

5. Volatility and position sizing: let risk budgeting come before "how bullish I am"

5.1 ATR is a scale, not a predictor

True Range 可以寫成當期最高低、跳空與前收之間最不利的距離，常見定義為：

ATR reflects recent price activity range and is useful for standardizing stop distances or per‑position risk; it is not a directional predictor and does not guarantee the next period will stay within the ATR range. In high volatility, the monetary risk of the same position size increases; if positions are not adjusted, a strategy's risk exposure will expand with market conditions.

5.2 Recomputable position example

If you further assume variable cost per unit is NT$0.20, the simplified unit risk becomes 8.20, the max integer position becomes floor(1,000÷8.20) = 121 units, estimated risk NT$992.20. This shows that “compute price distance first then ignore costs” leads to a slightly oversized position versus the true risk budget. UsePractical trading plan and position checkerIn any case, you should still verify the actual contract multiplier, fees, taxes, minimum tick and gap risk yourself.

5.3 Multipliers, minimum trade size and total exposure

Futures, CFDs, options and crypto derivatives should not be modeled directly with a spot per‑share model. Point multipliers, margin, expiries, forced liquidations, securities lending, currency and minimum order sizes all change worst‑case losses. Even if each trade uses only 1% account risk, multiple same‑direction markets can still be simultaneously triggered by the same macro event; position checks must extend to the portfolio level.

CME Group’s educational materials place position and risk management in the pre‑trade process, emphasizing that you should consider allowable loss and appropriate position size before deciding how much to buy.[5] The risk budget here is a research model, not a clinical assessment of an individual’s loss tolerance.

6. Exits, stops and position management: don't cut off the tail of winning trades.

6.1 Trade‑offs between fixed targets and trend exits

Fixed targets are easy to calculate but can prematurely end long trends; exiting at N‑day lows, an ATR trailing stop, or on a moving‑average breach can preserve more of the tail of gains but will give back some unrealized profits. Neither approach is inherently better — they produce different reward shapes. If a strategy originally researched as trend‑following later adds very close fixed targets, it may effectively become a different strategy; rule changes should not be hidden by keeping the original name.

6.2 Stop‑loss must define failure, not act as an emotional switch

Stops can be based on price structure, channel reversal breakouts, ATR distance or time conditions, but must be written before entry. Moving the stop farther changes the original risk budget; moving the stop to breakeven can increase short‑term noise. Every stop move must be predefined in the historical rules and marked in the trade log as “rule‑based” or “temporary modification.”

6.3 Adding to positions and partial exits require recalculating total risk

Adding to winners after a trend improves is not free. New entries, stops, sizes and existing positions should be combined to calculate the worst case; partial exits change realized average price and remaining exposure. If you only record the final average price, backtesting cannot show how each tranche of risk accumulated. You can split each fill into an event and check total risk and max drawdown at the portfolio level.

6.4 Trading and risk shutdown conditions

• If consecutive losses reach a pre‑set research warning threshold, stop and check execution and data instead of immediately ramping parameters up or down.
• If the equity curve drawdown exceeds the strategy’s historically acceptable range, trigger manual review and environment checks.
• Suspend applying old backtests when market liquidity, trading hours, contract specs or broker rules change.
• If an event or a gap makes execution assumptions unprovable, mark that trade as out‑of‑model instead of pretending precise fills.

7. False breakouts and failure modes: decompose "not making money" into fixable causes

7.1 Structure of false breakouts

A false breakout is not simply “breakout then drop.” You must define an observation period first: a breakout is a failure if within M bars the channel is reclaimed or the originally defined structural stop is hit. Different choices of M, close confirmations and exit methods yield different failure rates. If you label any pullback after a breakout as a false breakout, your research will be contaminated by look‑ahead information.

7.2 A failed breakout is not an immediate reversal

An original long breakout failing only means the long hypothesis wasn’t supported; it does not automatically establish a short strategy. To trade the opposite you must write separate short signals, fills, position sizing, stops and cost agreements. Otherwise binding long exits to short entries unintentionally increases trade frequency and slippage.

7.3 Research journal should record 'rule error' vs 'execution error'

A loss can be the natural cost a strategy pays in a ranging market, or it can stem from miscalculated signals, late order placement, oversized position or failure to execute stops as per rules. A trade journal should include at least strategy tag, environment tag, planned vs actual fills, slippage, rule compliance, result R, holding period and exit reason. UseTrading journal analyzerAfter reviewing aggregate statistics, group by environment and rule‑compliance levels.

8. Backtest validation: establish the data contract first, then look at performance curves.

8.1 Minimum backtest contract

• Data ordered old→new; tag exchange, timezone, adjustment method, missing values and corporate action.
• Upper／Lower／ATR／moving averages use only data known prior to signal formation.
• Explicitly state the time lag between signal and execution; do not treat a close signal as executed on the same close.
• Include fees, spread, directional slippage, taxes, no‑fill, partial fill and gap handling.
• Split periods by time into in‑sample (research), validation, and out‑of‑sample; parameters must not use later‑period data.
• Simultaneously report trade count, win rate, average P/L, expectancy, Profit Factor, MDD, losing streaks and time‑under‑water.
• Redo tests across different markets, volatility regimes and neighboring parameters to avoid a single equity curve driving conclusions.

8.2 Trend strategies should examine the distribution, not just CAGR

If all cumulative returns come from two extreme big wins, the actual holding experience of the strategy can be unstable over long periods. Inspect monthly or trade‑level distributions, worst losing streaks, maximum single loss, tail‑gain share, drawdown duration and recovery speed. Removing top trades or doubling costs is a simple stress test to identify whether the strategy relies on fortuitous events.

8.3 Out‑of‑sample and walk‑forward

You can split data into a fixed research window and a subsequent test window: decide N, buffer, ATR multiplier or environment filters in the research window and execute the next test without changes; slide the window forward and repeat. If after each test you go back to retune, the next segment is no longer a clean out‑of‑sample. The most important backtest output is not the “best parameter” but the rules’ fragilities across time and regimes.

If you want to turn signals and features into an auditable process, further reading:Quant research workbench、Article on backtesting and overfitting與Article on boundaries of technical indicatorsThis article does not package any backtest results as live trading signals.

8.4 Equity curve and drawdown must be reconciled with the trade journal

At the trade level, the R series, equity curve and asset price curve are three distinct layers. Price appreciation does not imply a strategy is profitable, and a profitable strategy does not mean drawdowns are tolerable. You can useRisk‑return and maximum drawdown analysis toolPaste the equity or return series for analysis, then review the log to identify which trades, market environments and execution deviations caused the maximum drawdown.

9. Ready‑to‑copy trend‑following and breakout Playbook table

The table below is not a recommendation of parameters but a research template you must fill. If any field can only be answered "watch the market" or "depends", the rule is not yet at a verifiable level.