TradingView MACD ReLoaded Strategy: Best Settings, Backtest

The short version. This is a TradingView MACD ReLoaded Strategy backtest. KivancOzbilgic's script (7,672 boosts) is a regular MACD where all three averages can be one of 11 moving-average types, VAR by default. I ported it to Python and ran it on Exness tick data with real spread, commission and, for Bitcoin, Exness swap.

• Bitcoin H4 is one of the better results in this series. PF 1.30 and +76,316 USD per BTC from January 2024 to November 2026 after spread, commission and swap, 223 trades, 24,685 drawdown. Buy and hold made +18,063 after swap, with a drawdown of about 67,000.
• And it held up after the script was published. On Dukascopy Bitcoin, H4 made PF 1.29 before July 2021 and 1.26 after. The two earlier ones I checked this way (ANN Strategy and Flawless Victory) didn't.
• But it's fading. By year: 2024 PF 1.64, 2025 1.23, 2026 so far 1.09.
• Gold H1 looked good (PF 1.21) until I split it by year. 2026 alone made 75% of the profit, and from 2004 to 2019 the same setup broke even.
• Forex lost. EURUSD, GBPUSD and NZDUSD: PF 0.69 to 1.07 from M5 to H4.
• The author's tip "use SMA or WMA on daily charts" didn't hold. SMA was the best daily setting on Bitcoin before publishing (PF 2.31) and dropped to 1.16 after. The default VAR went from 1.38 to 1.50.

My verdict: on Bitcoin H4 with the default settings, this is the first one in the series I'd keep watching. It's not proven: one market, a shrinking edge, and a statistics test it didn't clearly pass.

What is the MACD ReLoaded strategy on TradingView?

It's the strategy version of KivancOzbilgic's MACD ReLoaded indicator, first published on 4 July 2021 (current Pine v5 code from January 2023). The chart on the page is Binance BTCUSDT, daily.

Gerald Appel's MACD uses exponential averages. This one lets you pick: SMA, EMA, WMA, DEMA, TMA, VAR, WWMA, ZLEMA, TSF, HULL or TILL. The default is VAR, also called VIDYA: an EMA whose speed scales with a 9-bar momentum reading (CMO), so it nearly stops moving when price chops and speeds up when it trends.

The trading part is short:

MA12 = getMA(close, 12)
MA26 = getMA1(close, 26)
src2 = MA12 - MA26              // MACD line
MATR = getMA2(src2, 9)          // signal line
hist = src2 - MATR

if ta.crossover(hist, 0)
    strategy.entry('MacdLong', strategy.long)
if ta.crossunder(hist, 0)
    strategy.entry('MacdShort', strategy.short)

So it goes long when the MACD line crosses above its signal line and reverses to short when it crosses below. It's always in the market. No stop loss, no take profit. There are From/To date inputs, but the window() function that would use them is never called, so they do nothing.

That's the typical picture on Bitcoin H4. One long rode the rally from 110,000 to 125,000, one short caught most of the drop back to 105,000, and a two-candle flip in the middle cost a little.

Does MACD ReLoaded repaint?

No. Every input is the bar's close and the averages only look back. Signals come on the bar close and I fill them at the next bar's open, same as TradingView's tester with default settings.

One small thing: in Pine, VAR starts at zero on the first bar of the chart and takes a while to settle. TradingView loads enough history that you won't notice. My data starts in 2020, so I start VAR at the first price and skip the first 200 bars. On H1 and H4 the two starting rules gave the same PF within 0.02.

How I tested it

• Port: Python, signals on the close, fills at the next open, stop-and-reverse like strategy.entry. Default settings: VAR, 12/26/9, close.
• Data: Exness Raw Spread ticks for XAUUSD, EURUSD, GBPUSD, USDJPY and NZDUSD from January 2020 to September 2026, BTCUSD from 1 January 2024 to 3 November 2026.
• Costs: buys fill at the real ask on the bar's first tick, sells at the bid, plus 3.5 USD per lot per side (forex, gold) or 1 USD per BTC per side.
• Bitcoin swap: from the Exness symbol spec (September 2026): longs pay 16.39 USD per BTC per night, triple on Friday, nothing for Saturday and Sunday; shorts pay nothing. Roughly 7.6% a year at 78,000.
• Longer history: Dukascopy Bitcoin M5 from May 2017 and Dukascopy gold M1 from 2004.
• Size: fixed, no compounding. Gold in USD per 0.01 lot, forex in pips, Bitcoin in USD per 1 BTC.

The results

Profit factor after spread and commission, default settings:

Small timeframes lose everywhere. On M5 the strategy takes around 19,000 trades in six and a half years and the spread eats it (forex PF 0.69 to 0.83, gold 0.95). From H1 up it gets interesting on three markets: gold, USDJPY and Bitcoin.

Is MACD ReLoaded profitable on Bitcoin?

On H4 so far, yes, including swap. This is the part that surprised me.

BTCUSD, 1 Jan 2024 – 3 Nov 2026, long + short unless noted:

USD per 1 BTC. Swap took 8,095 from the H4 strategy (only the long half pays it) and 16,992 from buy and hold.

Both sides made money on H4: longs +51,477 after swap, shorts +24,839. That matters because a Bitcoin strategy that only profits from longs is mostly just holding Bitcoin. This one made money on the way down in early 2026 when buy and hold gave back most of its gain.

Then I changed the settings. Across 57 combinations of fast (6 to 24), slow (20 to 52) and signal (5 to 14) lengths, all 57 had PF above 1 before swap, and 93% still did after swap (median 1.20). It's not one lucky setting.

Now the worrying part:

The edge has shrunk every year. Long only is already losing in 2026.

Did it work on Bitcoin after it was published?

Yes, and this is the strongest evidence in its favour. Exness only has Bitcoin from 2024, so I went back to May 2017 with Dukascopy data and a 0.06% round-trip cost per trade, and split it at the publish date:

The second column is data the author couldn't have seen. H4 with the default settings made +223% compounded after publishing while Bitcoin itself rose 75%. Before publishing it made +201% while Bitcoin rose 2,127%, so it didn't beat the big bull run, but it held through the bear market of 2022 better than holding did.

What are the best MACD ReLoaded settings?

The author's description gives two tips: on shorter timeframes TILL, WWMA and VAR avoid whipsaws because they give fewer signals, and on daily charts WMA and SMA "are more accurate according to backtest results." I tested all 11 types on six markets and four timeframes.

Tip one holds. VAR, WWMA and TILL produce far fewer trades (on H1, a median of about 1,600 to 2,100 per market, versus about 6,900 for HULL and ZLEMA). On M15 and H1, VAR and WWMA ranked best on average, with TILL close behind. Fewer trades means less spread paid. They still lost on M15, just less.

Tip two doesn't. Here's Bitcoin, sorted by what looked best before publishing:

SMA was the clear winner on daily Bitcoin before July 2021, which is probably what the author's own backtests showed. After publishing it fell to 1.16. WMA fell from 1.53 to 1.09. VAR, the default, was mid-table before and the second best after. Across the six Exness markets, VAR and WWMA also ranked best on D1 on average, with SMA and WMA in the middle.

So my answer for "best settings" is boring: leave the default. VAR 12/26/9 on H4.

Is the gold result real?

Partly. Gold H1 made PF 1.21 and +3,025 USD per 0.01 lot from 2020 to 2026, 1,540 trades, 887 drawdown. Buy and hold made +2,588 with a 1,600 drawdown. Sounds good. Here's the timeline:

Five years of almost nothing, then 2026. On older gold (Dukascopy, 2004 to 2019, swap on longs scaled to price) the same setup made PF 0.99 on H1 and 1.09 on H4, while gold rose 266%. Swap on 2020–26 gold also trims H1 from 1.21 to 1.18.

The random timing test is interesting here. I slid the strategy's own long/short sequence to a random point in time 200 times, keeping every trade's length. On 2004–2019 gold the real signal beat 97% to 99% of the random copies on both H1 and H4. So the timing is better than chance, just not by enough to pay the spread. I saw the same thing with the ANN Strategy v2.

USDJPY D1 shows PF 2.05, but on 69 trades, with most of the profit in the 2022–24 yen slide and 2026 at PF 0.39. I wouldn't read much into it.

Risk structure

• No stop loss. A trade ends only when the histogram crosses zero again. On Bitcoin H4 that means holding a wrong-way trade for days.
• The worst losses on Bitcoin were shorts during breakouts. Ten worst H4 trades after swap, USD per BTC:

Eight of ten are shorts. The worst one sold just before Bitcoin broke its old high in March 2024 and held for 132 hours. At 0.01 lot that's −77.56 USD, so size it for at least 250 USD of drawdown per 0.01 lot.

• Win rate is low: 42% on Bitcoin H4, 35% on gold H1. Profits come from a few long trends.
• What isn't a red flag: no martingale, no grid, no averaging down.

Is the Bitcoin result statistically real?

Not proven. With the random timing test, Bitcoin H4 beat 97% of random copies (p = 0.03). That sounds good, but I ran the same test on 36 combinations (six markets, three timeframes, long + short and long only), and with that many tries a 0.03 can show up by luck. None of the 36 passed once I adjusted for that.

What makes me take it more seriously than the number alone: it held past the publish date, it held across all 57 parameter combinations, and it held on two different data sources. What makes me careful: almost three years of Exness data is only 223 trades, and the yearly PF keeps dropping.

Who should use it?

If you trade Bitcoin and want a mechanical trend system, MACD ReLoaded with the default settings on H4 is a reasonable thing to paper trade. It's simple, it doesn't repaint, and it kept its result after publication, which the earlier ones I checked didn't. Expect single losses of up to about 7,800 USD per BTC and long losing stretches.

I wouldn't run it on forex, and I wouldn't trust the gold H1 result to continue after the 2026 rally ends.

What I couldn't verify

• TradingView's own report. My numbers come from my port, not from TradingView's strategy tester.
• Binance BTCUSDT. The author's chart is Binance spot; mine are Exness CFD ticks and Dukascopy. Spot has no swap but has different fees.
• The exact swap on Exness Raw. The swap I used comes from the Exness symbol spec on a different account type (September 2026). Swap rates change, and the Raw account may differ.
• Forex swap. Not included. It wouldn't rescue PF 0.69–1.07.
• Out-of-sample on Exness Bitcoin. Only 2024–2026 exists. I'll rerun it as new data comes in.

Bottom line

The TradingView MACD ReLoaded strategy doesn't repaint, loses on forex, and only beat gold buy and hold thanks to 2026. On Bitcoin H4 with the default VAR settings it made PF 1.30 after Exness swap and kept its edge after publication, but that edge has shrunk every year since 2024.

This is part of a series where I port the most-boosted open-source TradingView strategies and test them on broker ticks with real costs. For another KivancOzbilgic script on the same data, see the TOTT backtest; for one where a backtest setting creates fake profits, the UT Bot backtest.

I'm not affiliated with TradingView or KivancOzbilgic. Past results, including backtests, don't guarantee future results.