Binance and Bybit data both put retail loss rates above 90%. The reasons are more boring than people think — and every one of them is solvable by software.
Internal data published by Binance and Bybit over the last two years has converged on the same uncomfortable number: roughly 95% of retail derivatives traders lose money over a multi-month sample. The same pattern shows up across BitMEX, dYdX, and Deribit. It is not a "bad year" or a regional quirk — it is the steady state.
When you look at the trade-level data behind that statistic, the failure modes cluster into three buckets. Each one is mechanical, and each one is something a machine handles better than a human.
1. Emotion overrides plan. The single most common loss pattern is a trader who entered with a pre-defined stop, watched price approach the stop, and moved it. They tell themselves "just a little more room." Sometimes the trade reverses and they feel vindicated. Most times the next leg is faster than expected, and a planned 1.5% loss becomes 8%. The same trader, on the way up, will close at a 1R win because they're afraid of giving back gains. Small wins, large losses, mathematically guaranteed drawdown.
The fix is not "be more disciplined." Discipline is a finite resource that erodes under stress. The fix is to remove the decision from the moment of stress entirely. A bot's stop is a parameter set hours ago in a calm room. When price hits it, the order fires. There is no negotiation. The trade closes at the planned loss, capital is preserved, and the next setup is evaluated cleanly.
2. Limited analysis surface. Most retail traders use three indicators: RSI, MACD, and a moving average. Three numbers describing a market that has dozens of meaningful dimensions — orderbook depth, whale wallet flows, funding rates, on-chain net flow, sector rotation, regime state, ensemble ML output, manipulation risk, multi-timeframe alignment. Three indicators in a 20-dimensional space is functionally trading blind on most setups.
The fix is to widen the analysis surface to the point where false signals get filtered before they reach a decision. Steleum runs a 20-dimensional composite for every pair on every cycle: 8 classical price/volume signals, 7 V9 intelligence layers (regime, sentiment, ensemble, manipulation), and 5 V10 fundamentals (on-chain, alt data, options, attribution). A signal that wins on momentum but loses on regime + manipulation never makes it to the order book.
3. No real risk management. "Risk management" for most retail accounts is a stop-loss field on the order ticket. There is no portfolio-level cap, no exposure budget, no per-day loss limit, no consecutive-loss cooldown. One bad day with three correlated positions and the account is gone.
The fix is institutional risk plumbing applied retail-side. Steleum enforces total-exposure caps, per-symbol concentration limits, daily-loss soft and hard halts, consecutive-loss cooldowns, kill-switch triggers, and per-user position-size scaling that responds to recent equity drawdown. Most of these never matter in good conditions; they are the difference between "down 4% on a bad day" and "blown account."
The architecture that ties it together. Steleum doesn't just throw indicators at the problem. The 20 dimensions feed a four-agent debate — Technical Analyst, Sentiment Analyst, Risk Manager, Execution Agent — that has to argue toward consensus. A CIO Agent reads the debate transcripts and the dimensional scores together, and only fires a TRADE verdict when conviction crosses a hard threshold (default 70%). Below threshold, the system is patient by design — it waits.
Patience is something machines do well that humans do badly. The CIO will sit through hours of choppy market without entering a single position because the conviction simply isn't there. A human watching the same screens feels the pressure to act and makes the trade that wasn't there. The 95% loss rate is, more than anything, a statistic about action under uncertainty.
Steleum launches April 28, 2026. The thesis is straightforward: take the three things humans do worst at scale — emotional discipline, breadth of analysis, and risk plumbing — and hand them to software that doesn't get tired.
“The 95% loss rate is, more than anything, a statistic about action under uncertainty. Patience is something machines do well that humans do badly.”
— Steleum Research Team
More detailed content for this post will be available shortly. Follow us on Telegram or subscribe to our newsletter to be notified when the full article goes live.
