Findings · 2026-11-18
We tested 8 popular TradingView strategies. None beat the fees.
Supertrend, MACD, RSI, the Squeeze, the turtle breakout and three more, run exactly as published on ten large coins for six years. Then we put every trade through prop-firm challenge rules. Passing a challenge with them is a coin toss.
The claim. “Pick a popular TradingView strategy, run it with discipline, and pass a prop-firm challenge.”
What we found. Not one of the eight made money after fees on the years it had not seen. The prop-firm passes they produce look like luck.
Follow a signal channel? We score every call it made, after fees. About a minute.
What we tested
Eight of the strategies traders share most on TradingView, with the settings they are published with: Supertrend (10, 3) on 4-hour bars; the 9/21 EMA cross on 1-hour bars; the MACD (12, 26, 9) cross on 4-hour bars; RSI (14) buying below 30 and selling above 70 on 1-hour bars; LazyBear's Squeeze Momentum on 4-hour bars; the 55/20 Donchian "turtle" breakout on 4-hour bars; the UT Bot alert on 1-hour bars; and the 50/200-day golden cross.
Each one traded long and short on ten large coin futures (Bitcoin, Ether, BNB, XRP, Cardano, Dogecoin, Solana, Chainlink, Litecoin and Avalanche) from 2020 to October 2026: about 119,000 trades. Every trade paid exchange fees and funding, and was filled at the next bar's open price, not at the signal price.
The result
We wrote the rules and the pass mark down before downloading any data, and judged each strategy only on the trades from 20 September 2023 onwards. None of the eight passed. The two with the strongest results, the Squeeze and Supertrend, averaged +0.4% and +0.2% a trade, which is well inside what luck produces; take away their five best weeks and both lose money.
The fast strategies did worst. The UT Bot traded over 30,000 times in three years and lost 0.10% a trade: the fees came to more than the moves it caught. The 9/21 EMA cross won fewer than three trades in ten.
The win-rate trap
The RSI strategy won 64% of its trades. It still lost 0.25% a trade, in every year we tested. The wins are small, because it sells as soon as the price bounces back to the middle. The losses are large, because when a coin keeps falling the strategy keeps holding. A high win rate on a strategy page tells you very little until you see the average trade.
Trend following worked, then stopped
The trend strategies (Supertrend, the turtle breakout and the golden cross) made real money in 2020 and 2021, when crypto ran for months in one direction. Since then the results have faded: in 2025 all three lost money, and the two still trading in 2026 lost again. A backtest that starts in 2020 looks far better than what the same rules have done lately.
Then we took the prop-firm challenge
Crypto prop firms fund traders who pass a challenge, typically: make 10% before you lose 6% (some firms also want a second 5% stage), never lose more than 3 to 5% in a day, and trade a minimum number of days. We put every strategy's actual trades through two of these rulebooks, starting a new challenge every week of the unseen period: 155 attempts per strategy.
Under the looser rulebook between 8% and 31% of attempts passed, most often in the first few days. Under the stricter two-stage one, all but one passed fewer than one time in ten. Strategies that lost money over the whole period still passed 8% to 19% of the time under the looser rules, which is what luck looks like.
We also simulated a trader with no edge at all under the same rules. Under the stricter rulebook they pass about one challenge in ten, and what they collect afterwards roughly equals the challenge fee. To make paying for challenges worth it, a strategy needs a real edge first. None of these eight has one.
What this does and does not say
It says these eight strategies, with their published settings, on ten large coins, did not beat their costs on years they had not seen. It does not say no indicator can ever work. But a backtest on a strategy page is usually the settings that fitted the past best, without fees, from a start date that flatters it. Ours were fixed before we looked.
We hold our own systems to the same bar, and publish them on the Proving Ground win or lose.
How it was tested
- Binance USDT-M perpetual futures, 1-hour candles from data.binance.vision, January 2020 to October 2026; 4-hour and daily bars built from them.
- Signals on closed bars, filled at the next bar's open. Costs: 0.107% taker fees per round trip plus real funding.
- Rules, settings and the pass mark were locked on 17 November 2026 before any data was downloaded. Pass mark: a t-statistic of 3 on the trades from 20 September 2023 onwards, surviving an extra 0.2% cost per trade.
- Uncertainty measured across weeks, not across trades, because trades in the same week move together. Each strategy was also compared with random entries of the same number and length.
- Prop-firm rules modelled from the firms' own published rules in November 2026, checked hourly against open positions. The 155 weekly starts share one price history, so they are not independent tries.
Research on past data, not financial advice. Every test we run, failures included, is on the Proving Ground.