Why 0x2ALL does not predict the price

Seven of the most used methods, each one described in its own terms, then in terms of what it measures and where it breaks. None of them is called a fraud here: they are descriptions of the past, computed with a parameter chosen after the past had already been seen.

Candlestick patterns

Promise: a reversal or continuation you can read at a glance — a hammer, an engulfing, a doji. What it measures: how the price moved inside a fixed number of candles, drawn and named after the fact. Where it breaks: the same shape appears thousands of times with opposite outcomes, and the name is applied only after the result is known. The pattern does not predict the next candle; it describes the candles that came before, with a story attached to them.

Elliott waves

Promise: the market moves in repeating five-wave impulses and three-wave corrections. What it measures: a count that an analyst assigns to a price series that has already happened. Where it breaks: several counts fit the same chart, the count is revised when the market disagrees, and the correction to a wrong count is presented as a new wave. Nothing in the method can be falsified in advance, which is also why it can never be shown wrong.

Trend lines and support/resistance

Promise: the price reacts at lines and levels that the chart reveals. What it measures: places where the price happened to stop before. Where it breaks: any two points make a line, so a line can always be drawn where one is needed; a broken level becomes the next level, and the explanation follows the price instead of anticipating it. The support that held three times was chosen with the benefit of having seen those three times.

Moving average crossover

Promise: a golden cross marks the start of an uptrend, a death cross its end. What it measures: the arithmetic mean of past closes, and the moment two means of different lengths cross. Where it breaks: by construction the signal arrives after the move began, and in a market without a trend the same crossover produces a sequence of small losses. The periods — 20, 50, 200 — are chosen by looking at where they would have worked.

MACD

Promise: momentum shifts before the price turns. What it measures: the difference between two exponential averages of closes, smoothed once more, with a signal line drawn on top of that. Where it breaks: it is the same price data as the averages, delayed twice and smoothed twice, so it turns later than the price it tries to anticipate. Its three settings are free parameters, adjusted until the indicator looks correct on the chart being studied.

RSI

Promise: above 70 the asset is overbought and a fall is due; below 30 it is oversold and a rise is due. What it measures: the ratio between average gains and average losses over a lookback window. Where it breaks: in a strong trend the reading stays above 70 for weeks while the price keeps rising, and the oversold reading appears at the start of a fall rather than at its end. The levels 70 and 30 are conventions of the indicator's author, not properties of the market.

Bollinger Bands

Promise: the price returns to its mean, and a squeeze announces a breakout. What it measures: the standard deviation of past closes around a moving average, drawn at a fixed multiple of it. Where it breaks: the band measures how much the price moved, not where it is going; touching it says nothing about direction, and a squeeze is followed as often by a breakout as by nothing at all. The width is one more parameter chosen after the fact.

The common point: all seven are descriptions of the past, computed with a parameter chosen after the past was seen. That freedom — the period, the level, the count, the line — is exactly what makes any of them look excellent on one particular chart. On the next chart the parameter is different, and it is chosen again.

Fundamental analysis does not fail for the same reasons as technical analysis, and the criticism of it has to be more careful. The problem is not that the fundamental case is false; it is that it answers a different question from the one a leveraged position asks.

1. Horizon

A fundamental case is informative over years. A perpetual position at 10x to 50x lives minutes to days, and in that window the result is dominated by flow, funding and liquidation. A correct thesis does not survive a 1% stop — not because it is wrong, but because the position does not live long enough for it.

2. Reflexivity

Keynes described the market as a beauty contest in which the prize goes not to the face you find prettiest but to the face you believe the average opinion will find prettiest — and, one step further, to what the average opinion expects the average opinion to expect. Price does not follow fundamentals in a straight line: the two determine each other. The fundamental case can be right while the market is busy with another narrative, and that is how the market normally works, not a failure of the market.

Keynes, The General Theory, chapter 12 · Reflexivity: price and fundamentals determining each other

3. Public information is already priced

In the semi-strong form of market efficiency, public and widely known information does not produce systematic return: the advantage would have to come from private information or from processing speed, and neither is available to retail. A quarterly report read after publication is public information.

Market efficiency — the semi-strong form

4. The instrument you hold is not the asset you analyzed

The venue lists tokenized equities, indices and commodities as well as crypto, and the distinction matters. In a tokenized equity or index the underlying asset has a real fundamental case — audited cash flow, quarterly results, material disclosures. But the position is not the equity: it is a perpetual financed by funding, marked against the venue's own reference price, with no dividend, no voting right, no ownership and no expiry date — what it has instead is funding paid periodically and the risk of forced liquidation at any moment. A gold or index perpetual on a venue that trades 24 hours can move while the spot market for the asset is closed, and the venue's reference price can drift away from spot in that window. A funding spike or a basis dislocation can liquidate a correct fundamental case with the underlying intact. For most crypto tokens the argument arrives from the other side: often there is no audited cash flow at all, and what is called a fundamental is an issuance schedule plus narrative. At both ends, what the position prices is not what the analysis describes.

5. You need the market to agree while the position is open

Even being right is not enough: the market has to agree while the position is still open. A perpetual has no expiry date — but it has funding to pay and it can be liquidated at any moment. The thesis has neither of those constraints.

There is a concession here, and it is not rhetorical: the value premium, the quality premium and the drift after earnings announcements are documented anomalies, and this page does not deny them. What it says is that those effects are realized over years, require surviving long drawdowns, and are not compatible with a leveraged position of short duration.

Value premium · Quality investing · Drift after earnings announcements

You do not have to trust this page. Three steps, ten minutes, no data subscription.

1. Generate a random price series: a random walk, or a spreadsheet column of random numbers presented as closes.
2. Run your own indicator on it — RSI, MACD, moving averages, whatever you use — with your usual settings.
3. Read the signals. The random series will produce buy and sell signals, oversold readings and crossovers: the same shapes you see on a real chart, describing a market that never existed.

This tests the tool, not our result. An indicator always finds structure, because it is built to describe movement and any movement can be described. What the random series cannot do is tell you the next move — and that is the only thing a trade needs.

The base rate: 74–89% of retail accounts trading leveraged derivatives lose money. That range is measured across retail accounts, the overwhelming majority of which are operated without any systematic rule for position size or exit. It is the starting point this project is designed against — not a prediction of what happens to an account that follows these rules. No page on this site claims to know how any particular account will end.

No indicator changes that rate, because in most cases the problem is not the entry. It is the size of the position and the moment of the exit. An entry can be right and the account can still end at zero: because the position was too large for the move, or because the loss was held open waiting for the entry to be vindicated.

Issuance has become cheap and keeps getting cheaper. Tokenization turns into a tradable instrument everything that already exists in the real world — equities, bonds, funds, real estate, commodities, private credit — and automation tools make creating new products and vehicles cheaper. The result is not a forecast: it is an observable asymmetry. The count of available instruments grows with the cost of issuance, which falls. The coverage and the history of each instrument grow with the cost of analysis, which does not fall at the same pace.

Consequently, the fraction of instruments with an analyst, with a trustworthy history and with relevant public information falls. A newly issued instrument has a short and non-stationary history, and a rule fitted to it is a rule fitted to noise. The number of testable hypotheses grows faster than the evidence base available to test them, and that is what makes multiple testing the central problem of selection.

Multiple testing and overfitting, as background

When the universe stops fitting anyone's attention, the binding constraint stops being the quality of the analysis and becomes what is executable. Liquidity, book depth, the existence of an oracle and venue support are the only properties verifiable across every instrument, including the ones nobody analyzed. In the tail, a wide spread, slippage and an oracle gap are not execution details: they are the cause of the loss.

This page makes no forecast about how many instruments will exist. It states an asymmetry that is already observable: issuance is cheap and getting cheaper, while coverage and history are not.

The strategy does not replace the indicator with a better indicator. It replaces the question: instead of asking where the price is going, it asks what happens to the account when the next move is unknown.

• It chooses the asset by liquidity: a uniform draw among the 20 highest-volume assets on the exchange, never by signal. Volume is where the order fills; it says nothing about where the price is going.
• It sizes each position at the whole wallet balance, using the asset's maximum leverage up to a 50x cap, so the margin per position stays small; each position carries risk of about 1% of the balance.
• It opens with the stop defined before the entry, and the stop lives on the exchange, not in the app: if the phone goes offline, the protection stays.
• It sets open targets in progression — +30%, +90%, +210% — realizing one third of what remains at each target and raising the stop to two thirds of the way.
• It moves the stop to break-even when the position is +3% in favor, and locks in +3% once it is +6%.
• It withdraws half of the realized profit per event from US$ 22 net, leaving the other half as collateral, where it keeps working.
• It never averages down and never closes on opinion: a position ends at its stop or at a target.

Selection by volume is not a quality filter. It deliberately includes whatever is being traded most right now — an index perpetual, a tokenized commodity, a large-cap crypto or a memecoin — with no regard for sector, fundamentals or the reputation of the asset. That is precisely why the risk rules, not the selection, carry the strategy.

The object being priced is the contract, its risk is mechanical, and the only lever that genuinely belongs to the operator is how that risk is sized and exited. Fundamental analysis is not incompatible with the market. It is incompatible with risk management: one is a statement about direction over a horizon, the other is a statement about survival at every instant. They do not compete, and the second does not require the first to be right.

The backtest was built on 1-minute crypto candles — assets correlated with each other. The strategy now trades whatever has the most volume on the venue, which includes tokenized commodities, indices and memecoins that were not in the simulation. The published distribution therefore describes the strategy applied to correlated crypto, not the experience of an operator who ends up positioned in an index or a commodity.

What remains, plainly: partial losses are expected and frequent, the left tail of the distribution is real, execution can slip, the exchange can fail, and a compromised device can lose the keys. A stop removes none of that — it only defines the size of each loss in advance.

The published backtests carry both sides, and the caveat above applies to every figure: the simulation covers correlated crypto only. In the second one, 1,000 simulated users with 10 wallets each: a median of +30.7% in one year, 31.3% of users finishing at a loss and P05 of −45.4% — one in twenty lost 45.4% or more. Those are past results of a simulation on crypto, and they guarantee nothing about the future.

See the whole distribution, both sides