ai trading / algorithmic trading / backtesting / day trading / fact check
I fact-checked a day trading strategy with 423,000 views. It loses to doing nothing.
Erick Jablonski's four-timeframe trend strategy has 423,900 views and 41,000 likes, and he says it has paid him out of three prop firms. So I gave the video to my agent and had it check.
The agent watched the video rather than reading a transcript, pulled his eight rules out of it, turned them into a real strategy, and backtested it on a full year of minute by minute data. His rules, coded exactly as he describes them, return -0.25% over the three months to 10 September 2026. Buying and holding the S&P over the same three months returns +1.02%.
Then I let a computer tune it. A sweep trained on nine months of data and picked its winner on a separate validation window, and on the year it was fitted to that winner returns +15.90% against buying and holding's +16.27%. Roughly a tie. On the three months the sweep had never seen, it returns -2.40% while the index gains 1.02%. A strategy that ties on the data it was fitted to and loses on data it has not seen is the definition of a curve fit.
Every assumption in the test is in his favour. I ran it on SPY, built real candles on all four timeframes he describes, and tested both directions rather than only the easy one. Commission, spread and slippage are all literal zero, verified as totalFees: 0 on every run, and the engine does not model taxes at all. The tuned version realises 1,491 short-term round trips in a year against buy-and-hold's zero, so tax would only widen the gap.
What he genuinely wins on is drawdown: 5.90% worst drawdown over the full year against buying and holding's 9.61%. Sitting in cash is why he trails and also why he draws down less. His median deployment is 0% against the index's 97.34%, which is the mechanism behind the whole result.
The portfolio is public and forkable, and every figure above names the backtest it came from.
Transcript
0:00Erick Jablonski: "And that's literally it. Just let the trade play out. This is a variation of a strategy I learned from Ray Trades, which has got me three payouts from prop firms."
0:06We about to fact check this [expletive].
0:08Erick Jablonski, giving the rules: "On the 15-minute chart, if the chart is going up, we want to buy. This is called an uptrend. If the chart is going down, we want to sell. This is called a downtrend. Now do the same thing on the one-hour chart, daily chart, and weekly chart. Do not take a trade until these four charts all have the same trend. So now we just wait for at least one red candle to form and for a candle to break above that previous red candle. Once that happens, you can enter long. The stop loss will go underneath the most recent low. And for TP, just aim for a 1 to 2.5 risk-reward ratio."
0:24What the hell is he talking about?
0:26If this strategy made money, every freshman taking intro to CS could stop worrying about the job market and start creating bots in their sleep.
0:33So I gave the video to my agent. The agent watched the video, it pulled out his rules, and it turned them into a real strategy. Then it backtested that strategy across a full year of minute by minute data.
0:44And his strategy actually works, sort of. I gave it every advantage I could think of. It still lost to a guy who bought the S and P and went to sleep.
0:53So I let a computer tune his strategy on nine months of data, and keep the best version it could find. On the stretch it was tuned on, it just about tied with buying and holding. Then I showed it three months it had never seen. It lost two and a half percent, while the index went up.
1:09Now let me be fair to him. I ran it on SPY, built real candles on all four charts he describes, and tested both directions instead of just the easy one. And I charged him nothing at all. He paid zero commission and zero spread, every fill was perfect, and nobody ever taxed his gains. None of that is how real trading works. Every one of those assumptions helps him, and he still lost.
1:30If you think I got this wrong, the whole thing is public and you can fork it and prove me wrong.
1:35If you want me to fact check someone else then tag me in their video, and let us hold these guys accountable.
Join the conversation
Loading conversation…