Train a neural network on five years of 5-minute Nasdaq bars and ask it to call the direction of the next bar. With enough layers and enough training, it will get there: 90%, 95%, sometimes higher, on the data it was trained on. Show it next month's bars and the number collapses toward a coin flip.
That gap is not a bug in the code or a shortage of GPU time. It is the predictable result of pointing a very flexible function at data that is mostly noise, and it is the most common way people waste a year on neural network trading.
A neural network is a function approximator. You give it inputs (past returns, volume, indicator values) and a target (the next bar's direction), and it adjusts millions of internal weights until its output matches the target on your training data.
Nothing in that process asks whether the pattern will exist tomorrow. The optimizer has one job: reduce the error on the examples in front of it. If the only way to cut that error further is to memorize the odd sequence of bars on one Tuesday in March 2021, it memorizes that Tuesday. It has no concept of "this was a coincidence."
A neural network does not learn why a market moved. It learns whatever shape best reproduces the answers you gave it. When the answers are mostly noise, the shape it learns is the noise, and noise does not repeat.
Neural networks work extremely well on images and language. Financial price series have three characteristics that image and language data do not.
The third point is the one people underestimate. Five years of 5-minute bars sounds like a lot: roughly 100,000 rows. But consecutive bars are highly correlated, they come from one instrument, and they share a handful of market regimes. The number of genuinely independent observations is orders of magnitude smaller than the row count, while the model has millions of parameters. That ratio is the definition of a setup that fits noise.
You do not have 100,000 examples. You have a few dozen market regimes, sampled 100,000 times.
Almost every impressive result you will be shown is measured on data the model has already seen, or on data contaminated by it. Three mechanisms do most of the damage.
| Mistake | What it looks like | What it does to the result |
|---|---|---|
| Random train/test split | Shuffling bars, then splitting 80/20 | Future bars leak into training. Accuracy becomes meaningless. |
| Normalizing on the whole set | Scaling features using the full history's mean | Information from the future enters every training row. |
| Tuning on the test set | Trying 200 architectures, keeping the best test score | The test set becomes a training set. The score is now in-sample. |
The third is the quiet killer, because it feels like diligence. Every architecture you discard on the basis of test performance spends a little of that data's independence. Run enough trials and you will find a configuration that looks excellent on any dataset, including pure random numbers. That is the core argument of the backtest-overfitting literature, and it hits neural networks hardest because they are so easy to run hundreds of times. The general version is in overfitting and curve fitting, and the partial defense is in in-sample versus out-of-sample.
The systems on this site have no learned weights and no retraining. The same rules, on the same window, over years of historical data, with every new signal streamed live.
Quantitative firms do use machine learning, so the honest position is not "it never works." It is that the requirements are steep and almost nobody at home meets them.
That last one ends more projects than any statistical issue. Slight predictive accuracy is common. Accuracy that survives commissions and slippage is rare. More in machine learning in trading.
When a product says it uses AI to forecast price, there are only a few possibilities, and most are bad.
The test that cuts through all four is the same one you would apply to any system: can you reproduce the result yourself, on data the model never saw, with realistic costs applied? If the answer involves trusting a screenshot, you are not looking at evidence. See what a backtest is for what a defensible one requires, and can AI predict the stock market for the broader claim.
Neither Algomyth system is a neural network. The Algomyth engine is a fixed set of rules for one moment in the day: the window that opens at 9:30 AM ET, when the US cash market opens and the day's volume arrives. The stop and the target go into the market at the same instant as the entry, the target is 3× the risk, and no position is held overnight.
That design is deliberately hard to overfit. There are few parameters, one trading window and one decision a day, so there is very little surface for curve fitting to grab. The backtest runs day by day from January 2022 to September 2026 on a mid-size account: 1,164 trades, a 1.46 profit factor, a worst drawdown of $5,273, and $87,359 net, with the commission and a tick of slippage already deducted from every trade.
Note the geometry rather than any hit rate. With the target at three times the stop, break-even arrives at a 25% win rate, and that is the point. The system does not predict the next move. It takes a small structural advantage and lets a 3 to 1 payoff do the work over hundreds of trades. A model that claims to call direction correctly most of the time is claiming something categorically different, and much less believable. The comparison is in AI versus rule-based trading, and the wider picture in the AI trading guide.
Hypothetical performance. The Algomyth engine figures above come from a backtest over historical data, not from a live account. Simulated results are prepared with hindsight, carry no financial risk, and cannot fully reflect real execution, slippage or liquidity. Past performance, real or simulated, does not guarantee future results.
It can find genuine statistical structure, but usually a small amount, and rarely enough to survive transaction costs on one instrument. Institutions extract value from machine learning by applying it across hundreds of markets with strict validation and very low costs, not by forecasting the next bar of one index. Treat any claim of reliable directional prediction as an in-sample result until it is shown on data the model never saw.
Because more rows are not more information when the rows are correlated and come from the same regimes. Ten years of one index gives more bars but not many more independent market environments, and the model still has millions of parameters to spend. Length also makes non-stationarity worse, since the oldest data came from conditions that no longer exist. Breadth across instruments helps far more than length on one.
For most retail applications, yes, and for a specific reason: capacity. A model with more parameters can fit more noise, so with a low signal-to-noise ratio it will. Simpler models with few parameters are more constrained, which makes their failures more visible and their successes more likely to hold up. That is the same reason a rule-based strategy with a handful of conditions is easier to trust than one with forty tuned thresholds.
Run it on a time period that was never touched during development, in strict chronological order, with realistic commissions and slippage applied. If performance drops sharply, the model learned the training data rather than the market. Do this once and honestly, without going back to adjust anything afterward, because the moment you tune on that period it stops being out-of-sample and becomes part of the training set.
Everything we write comes back to one idea: verify, then decide. The live terminal exists for exactly that.
Unlock Algomyth Live