MACD Reversal vs. Continuation: What a 100-Trade Test Shows

Video: Traded the MACD indicator 100 TIMES (RESULTS REVEALED)
Creator: Data Trader
Original Video: Watch the original video
Difficulty: Intermediate

The Big Idea

Data Trader compares two MACD entry policies on a 15-minute EUR/USD chart: a reversal-oriented crossover and a continuation-oriented recross. Both exit at the next MACD signal-line crossover.

The reported 100-trade samples produced 42% and 23% win rates. That is useful descriptive information, but it is not the win rate of MACD itself. Results belong to the complete combination of market, period, settings, entry timing, exits, exposure, and costs.

This guide turns the comparison into a reproducible experiment and examines two important limitations: uncertain accounting and the difference between margin allocation and loss risk.

Understand the Indicator Before the Test

The common EMA-based MACD line is the 12-period EMA of close minus the 26-period EMA. Its signal line is a nine-period EMA of that MACD line. The histogram is MACD minus signal, and the baseline is zero.

TradingView’s MACD documentation (paid link) identifies those default lengths. The supplied transcript does not explicitly confirm its settings, so they are research defaults rather than verified source parameters.

Above zero means the faster price EMA is above the slower one. Below zero means the reverse. A signal-line cross describes a different relationship: MACD moving above or below its own smoothed reference. Neither relationship guarantees the next price direction.

Two Entry Policies, One Exit Policy

The normal method seeks a momentum turn from the opposite zero-line region. The new method seeks renewed momentum within an established region. Calling them reversal and continuation methods is a convenient classification, not proof of the subsequent price pattern.

The transcript also delays entry to the next candle’s close when an exit coincides with a new entry signal. That rule can materially change results and should not disappear in a coded comparison.

What the Source Reports

The presenter counts breakeven trades as losses. Preserve that convention when comparing win rates, but record breakevens separately too.

The reported pip and cash totals cannot be independently reconciled without the trade ledger, quantity formula, currency conversions, and costs. Variable exposure can make aggregate pips and cash diverge. The second result may also contain a transcription or spreadsheet issue; this guide does not silently replace it with an assumed value.

A $10.13 gain on $10,000 is approximately 0.1013%. Even modest omitted spreads, commissions, or adverse fills could change its sign. A brief peak profit is not the final result or a full drawdown measurement.

Win Rate Does Not Determine Expectancy

A lower-win-rate method can earn if its average winners sufficiently exceed average losers. A higher-win-rate method can lose if its losing trades are disproportionately large.

In an original illustration, 42 wins averaging +2R and 58 losses averaging −1R sum to +26R before costs. The same 42 wins averaging +0.5R sum with those losses to −37R. The win rate is identical.

Record mean and median net outcome, payoff ratio, profit factor, drawdown, holding time, and tail losses. The source’s conclusion that the continuation method cannot improve with different management is not established by its win rate alone. A new exit policy requires a new test.

One Hundred Trades Provide a Sample

Under a simple independent-trial model, approximate 95% Wilson intervals are 32.8–51.8% for 42 wins and 15.8–32.2% for 23 wins. These calculations express uncertainty under that model, not guarantees for future markets.

Real trades can cluster by session, volatility, or trend state, weakening the independence assumption. Missing test dates also make it difficult to judge whether the samples represent comparable conditions.

Use consecutive opportunities and compare both methods on the same calendar period. Their 100th trades may occur on different dates because their frequencies differ. Report both common-period performance and any separately counted 100-trade results.

One Percent Margin Is Not One Percent Risk

The source describes allocating 1% of capital with 150× leverage and calls this 1% risk. Without a stop limiting losses, those are different quantities.

For an original $10,000 account example, 1% collateral is $100. At 150×, that can support $15,000 of notional exposure. A 1% adverse price move represents about $150 of price loss, or 1.5% of account equity, before costs. The exact outcome depends on contract and margin rules.

The CFTC’s forex advisory explains margin-related risks. Leverage availability and obligations depend on the product, provider, and jurisdiction; the video is not a universal account specification.

A crossover exit has an unknown future distance. Allocating collateral does not establish a maximum loss. Position exposure, a defined stop when used, liquidation rules, and execution risk must be evaluated separately.

Pine Script: Compare the Entry Rules

//@version=6
strategy("MACD reversal vs continuation - crossover exits", overlay=false,
     initial_capital=10000, pyramiding=0, margin_long=100, margin_short=100,
     commission_type=strategy.commission.percent, commission_value=0.005,
     slippage=1, process_orders_on_close=false, calc_on_order_fills=false,
     calc_on_every_tick=false)
method = input.string("Reversal", "Method", options=["Reversal", "Continuation"])
units = input.float(1000, "Fixed quantity; check symbol units", minval=1)
fast = input.int(12, "Fast EMA", minval=1)
slow = input.int(26, "Slow EMA", minval=2)
sig = input.int(9, "Signal EMA", minval=1)
if barstate.isfirst and (not chart.is_standard or fast >= slow)
    runtime.error("Use standard candles and fast length below slow.")
[m, s, h] = ta.macd(close, fast, slow, sig)
up = ta.crossover(m, s)
down = ta.crossunder(m, s)
above = m > 0 and s > 0
below = m < 0 and s < 0
var bool longArmed = false
var bool shortArmed = false
var int pendingSide = 0
var int pendingBar = na
if barstate.isconfirmed
    if not above
        longArmed := false
    if not below
        shortArmed := false
    if down and above
        longArmed := true
    if up and below
        shortArmed := true
    goLong = method == "Reversal" ? up and below : up and above and longArmed
    goShort = method == "Reversal" ? down and above : down and below and shortArmed
    candidate = goLong ? 1 : goShort ? -1 : 0
    if goLong
        longArmed := false
    if goShort
        shortArmed := false
    exitLong = strategy.position_size > 0 and down
    exitShort = strategy.position_size < 0 and up
    if exitLong or exitShort
        strategy.close_all(comment="Next crossover")
        pendingSide := candidate
        pendingBar := bar_index
    else if strategy.position_size == 0
        delayed = pendingSide != 0 and bar_index == pendingBar + 1
        side = delayed ? pendingSide : candidate
        pendingSide := 0
        if side == 1
            strategy.entry("L", strategy.long, qty=units)
        else if side == -1
            strategy.entry("S", strategy.short, qty=units)
plot(m, "MACD", color=color.blue)
plot(s, "Signal", color=color.orange)
plot(h, "Histogram", style=plot.style_columns, color=h >= 0 ? color.green : color.red)
hline(0, "Zero")

The following strategy uses the common 12/26/9 settings and selectable reversal or continuation entries. Continuation requires a preceding counter-cross in the same zero-line region; crossing out of that region cancels its armed condition.

It uses fixed quantity rather than the source’s unclear leveraged spreadsheet sizing. It deliberately retains crossover-only exits and therefore has no price-based protective stop. It is a research model, not a live-trading risk policy.

When an exit also qualifies as a new entry, the new direction is stored and submitted at the following candle’s close if flat. That stored signal is not revalidated against a new zero-line condition, a disclosed interpretation of the source’s delay.

Execution QA and Accounting

Use standard 15-minute EUR/USD candles and identical date ranges for both runs. Confirm how quantity maps to base-currency units on the selected feed. The default quantity is 1,000, with 100% margin assumptions; this does not reproduce 150× account leverage or compounding.

Signals are confirmed at candle close. By default, orders fill on the next bar’s open, which differs from the source’s close-to-close measurement. Enabling “Fill orders on bar close” in Strategy Properties provides a closer paper-accounting comparison but assumes those closing fills were obtainable. Report which mode is used.

Opposite crosses close positions without immediate automatic reversal. There are no higher-timeframe requests, retrospective swing labels, fill recalculations, or intrabar signals. The strategy exits only on a crossover and can hold through sessions and weekends.

Commission defaults to 0.005% per fill and slippage to one tick, adjustable in Strategy Properties. These are placeholders, not a verified EUR/USD cost model. Spread, financing, and dealer execution are not explicitly modeled. Inspect margin events because they can add exits outside the intended crossover policy.

The code has been reviewed for its stated logic but has not been compiled or backtested in TradingView. No reproduction of the reported totals is claimed. Export individual trades and retain the source’s convention that net breakeven counts as a non-win.

Backtest Lab

Any proposed stop or profit-target improvement should be defined before testing on unseen data. Choosing the exit that would have avoided the largest historical loss is hindsight, not proof of an improvement. Good money management limits exposure; it does not create positive expectancy from every indicator.

Final Takeaway

The video supplies a useful comparison to investigate, not a definitive verdict on MACD. Preserve its zero-line conditions, recross sequence, timing, and crossover exits; then reconcile the ledger and costs. Treat margin, loss risk, win rate, and net expectancy as separate measurements.

Risk Disclosure: This material is for educational and research purposes only and is not financial or investment advice. Trading can result in substantial losses. Backtests are hypothetical, depend heavily on their assumptions, and cannot reproduce every aspect of live execution. Commissions, spreads, slippage, liquidity, gaps, financing, order behavior, market regime changes, and discretionary decisions can materially affect results. Leverage amplifies exposure, and margin allocation does not cap loss. Crossover-only exits provide no fixed price-loss limit. Historical results do not guarantee future performance.

This article is an independent educational companion to the video and is not affiliated with, sponsored by, endorsed by, or authorized by Data Trader.