Quick answer: In our backtest of Trading Rush's MACD + 200 EMA strategy on BTCUSDT 30-minute candles, 1 Oct 2023 to 30 Sep 2026, fees included, all four versions lost money. Each one was ahead before costs on a fixed $100 risk view, and costs turned every one negative. Unseen data and random entries were not run.
A MACD cross with a 200 EMA trend filter is one of the most shared strategies on YouTube. It's short, mechanical and easy to follow, which also makes it easy to test.
That's what Episode 2 of Creating Bots does. Watch it here: MACD + 200 EMA on Bitcoin: What We Built, What We'd Change. Each week we take a famous public strategy, turn it into rules with Pulse and test it in the open. This week's strategy comes from Trading Rush, in his video BEST Bitcoin Trading Strategy EVER with PROOF tested 100 TIMES.
What did Trading Rush claim?
Here is his result, in his own words:
"After 100 trades on the world's largest cryptocurrency, Bitcoin, the MACD strategy got an approximate win rate of 62 percent with a 1.5 to 1 reward risk ratio."
On his screen at about 7:35, his BackTester read: win rate 62%, 100 trades, 62 won and 38 lost, starting balance 10000, current balance 15500, current profit 5500 and highest profit 5550. Those are figures from his own test, shown with credit. They are not our results, and they are not a forecast.
Earlier in the same replay, at about 7:09, the screen read: win rate 63%, 92 trades, 58 won and 34 lost. It isn't a different test. His BackTester steps through the trades as he talks, so the figures change as trades are added.
Credit where it's due. Trading Rush tested in the open, and he said plainly where his strategy struggles:
"When the price movement is good like it is right now, the MACD win rate will be good because the MACD strategy needs a trend. But if the Bitcoin movement slows down like it did last time, and the price starts to stay in a range for longer periods of time, the win rate will be obviously lower because the MACD strategy is not a range market strategy. If the price itself is not moving, you can't magically make money."
He describes his own stretch as one where "most of the current market structure is making good moves like before" (about 6:06), and he mentions a sideways spell: "the price was staying in the sideways direction for a longer period of time" (about 5:58). His video shows no timeframe, date range or fees, so we say nothing about them.
How does the MACD + 200 EMA strategy work?
These are his rules, from the Bitcoin video and his first MACD video, "I risked MACD Trading Strategy 100 TIMES Here's What Happened... - Forex Day Trading":
Buy when the MACD line crosses above the signal line, below the zero line, only if the candle is above the 200-period moving average.
Sell is the mirror: the MACD line crosses below the signal line, above the zero line, with the candle below the 200-period moving average.
Put the stop beyond the pullback of the trend.
Aim for a profit target of 1.5 times the stop, which he calls "the best reward to risk ratio we have found after taking 300 trades".
His first video also makes a point that suits this series well: "just because i showed you a hand picked buy setup that had 3 winners doesn't prove anything".
Which gaps did we have to fill?
We described the strategy to Pulse in plain English. A bot can't guess, so four parts of the published rules needed a decision.
Gap 1: how far below zero?
His first video says "buy signals that are way below the zero line". Way below isn't a number, so our rule is simply below zero for longs, and above zero for shorts.
Gap 2: where does the stop go?
His wording is "the stop loss is below the pullback of the trend". Two traders would draw that pullback in two different places. Our stop is the lowest low of the last 10 candles, including the entry candle, for a long, and the highest high of the last 10 candles for a short.
Gap 3: which exit?
His first video describes taking 25 percent off at 1R, moving the stop to breakeven, then aiming for 2R. Both of his 100-trade tests used 1.5R, though, so ours is a fixed 1.5R target.
Gap 4: the trades he skipped
In his first video he left out 3 trades "that had confusing entry. the entry signals were not clear." A test can't skip trades on a feeling. Our version takes one position at a time and ignores any signal that arrives during a trade.
Two smaller choices complete the rules. The 200-period average is exponential, because his first video says so. His Bitcoin video doesn't name a timeframe, so we chose 30-minute candles. Change any of these and you have a different strategy.
How did we test it?
Market: BTCUSDT, 30-minute candles
Period: 1 October 2023 to 30 September 2026, three full years, trends and ranges included
Headline view: a compounding account, $10,000 start, 1% risk per trade, fees included
Cost view: fixed $100 risk per trade, no compounding, with fees and slippage shown separately
MACD settings: 12, 26, 9
The cost-view settings were fees of 0.06% and slippage of 0.02%, each in and out. Those are the settings of this run, not a statement about any exchange, so check your own account's rates.
What did the four versions return?
Version 1 is Trading Rush's rules as published, with the gaps filled as above. Everything else is ours.
ADX filter (Version 2, OUR ADDITION): only take a trade when ADX(14) is at or above 25 on the entry candle, for both long and short entries. The ADX indicator measures how strong a trend is, not which way it points. He says the strategy is not a range market strategy, so this rule tries to keep it out of ranges. We fixed 25 before running anything, ran it once and tried no other value.
200 EMA slope (Version 2 alt, OUR ADDITION): only take a long when the 200 EMA is higher than it was 20 candles earlier, and only take a short when it is lower, judged on the entry candle.
20 MA exit (Version 3, OUR ADDITION): entries exactly as Version 1. On top of the stop and the 1.5R target, a long closes on the first close below the 20-period simple moving average, and a short on the first close above it.
Backtest results, compounding account ($10,000 start, 1% risk per trade, fees included):
As published (Version 1): 580 trades, return -50.7%, max drawdown -52.4%, profit factor 0.84
ADX filter (Version 2, OUR ADDITION): 203 trades, return -22.3%, max drawdown -29.5%, profit factor 0.83
200 EMA slope (Version 2 alt, OUR ADDITION): 550 trades, return -55.2%, max drawdown -55.8%, profit factor 0.81
20 MA exit (Version 3, OUR ADDITION): 407 trades, return -37.5%, max drawdown -38.0%, profit factor 0.84
Our test: BTCUSDT, 30-minute, 1 Oct 2023 to 30 Sep 2026, fees included. Backtest.
A profit factor below 1 means the losing trades cost more than the winning trades made. Our ADX rule lost less, mostly because it took far fewer trades, but its profit factor didn't improve: 0.83 against 0.84. None of the four versions cleared a profit factor of 1.
Where did the money go?
Here the second view matters. On a fixed $100 risk per trade, with no compounding, every version was ahead before costs and behind after them.
Backtest results, fixed $100 risk per trade (fees 0.06% and slippage 0.02%, each in and out):
As published (Version 1): 580 trades, +$6,000 before costs, fees $9,469, slippage $3,156, total costs $12,626, net -$6,626
ADX filter (Version 2, OUR ADDITION): 203 trades, +$1,700 before costs, fees $3,045, slippage $1,015, total costs $4,060, net -$2,360
200 EMA slope (Version 2 alt, OUR ADDITION): 550 trades, +$4,750 before costs, fees $9,268, slippage $3,089, total costs $12,357, net -$7,607
20 MA exit (Version 3, OUR ADDITION): 407 trades, +$4,050 before costs, fees $6,316, slippage $2,105, total costs $8,421, net -$4,371
Our test: BTCUSDT, 30-minute, 1 Oct 2023 to 30 Sep 2026, fixed $100 risk per trade, fees and slippage included. Backtest. Figures rounded to the dollar.
For Version 1, that works out at $21.77 of costs per trade, or 22% of the $100 risked. A win at the 1.5R target is $150, so costs ate a large slice of every winner and added to every loser. The cost settings were the same for every version, and the differences in total costs come mostly from how many trades each one took.
On the compounding account, Version 1 paid $9,968 in costs ($7,476 in fees and $2,492 in slippage) on about $12.5M of volume, with an average position of about $10,700. Backtest.
Version 1, year by year
Backtest results, same fixed $100 risk cost view:
2023 (Oct to Dec): 48 trades, -$50 before costs, -$981 net
2024: 193 trades, +$3,700 before costs, +$130 net
2025: 198 trades, +$2,200 before costs, -$2,827 net
2026 (Jan to Sep): 141 trades, +$150 before costs, -$2,949 net
Figures rounded to the dollar, so the yearly nets add to $1 more than the -$6,626 total.
Only 2024 finished ahead after costs, and only just. Late 2023 was behind even before costs. In 2025 and 2026 the strategy was ahead before costs, and costs took all of it and more.
100 trades vs three years: what does this mean?
None of this says Trading Rush did anything wrong. A hundred trades on one stretch and several hundred over three years are different tests, so neither cancels the other. His stretch was one he describes as "making good moves", and he warned that ranges would hurt. Our three years included plenty of both.
It also doesn't prove the idea is dead. Our version used our choices for the zero line, the stop, the exit, the timeframe and the costs. Unseen data and random entries were not run, so nothing here is graded, and nothing here says the rules work or fail in general.
The fixed line we use in every episode: A backtest is the past. A simulation is practice. Neither predicts future performance.
What would we change next?
These are ideas to test, not results:
Lower costs: for example, limit orders, at your own account's rates.
A higher timeframe: wider stops mean smaller positions for the same risk, so each trade pays less in costs.
The exit from his first video: 25 percent off at 1R, stop to breakeven, then aim for 2R.
Unseen data and random entries first: before any version earns more time.
Test your own MACD version with Pulse
Trade a MACD cross yourself? Describe your version to Pulse in plain English and it will turn it into rules you can see and edit. Backtest them in Strategy Lab, then let Evidence Lab check for luck, unseen data and fragility. Live trading involves risk and can lose money.
FAQ
Does the MACD + 200 EMA strategy work on Bitcoin?
In our backtest, no version made money after costs. Over three years of BTCUSDT 30-minute candles with fees included, Version 1 returned -50.7% on a compounding account. Other rule choices could behave differently, so test your own version before you trust it.
Why did our results differ from Trading Rush's 62% win rate?
They're different tests. His covered 100 trades on a stretch he describes as making good moves, while ours covered three years, ranges included, with fees. He also said himself that the strategy is not a range market strategy.
Did fees cause the losses?
Mostly, in this backtest. On the fixed $100 risk view, all four versions were ahead before costs and behind after them. For Version 1, costs were $21.77 a trade, which is 22% of the $100 risked.
Did adding an ADX filter help?
It lost less, -22.3% against -50.7%, largely by taking 203 trades instead of 580. Its profit factor was 0.83 against 0.84, so the filter didn't make the trades themselves any better.
Educational only. Not advice. Crypto is high risk.
