Why 92% of New Traders Fail Without AI —
And How to Join the 8%.

The number is roughly consistent across every regulator that forces brokers to publish it: between 70% and 90% of retail CFD and forex traders lose money inside twelve months. In the toughest regulatory samples the number touches 94%. Rounded, generalised, a bit uncomfortable: 92% fail.

If you take that number seriously, you have to then ask the follow-up: why?

Spoiler — it's not because the market is rigged. It's because of a very specific stack of human errors that compound. Automated trading, done honestly, removes about two-thirds of them. Here's the breakdown.

Failure mode 1 — Position sizing done emotionally.

A new trader sees a "sure thing" and sizes 10% of the account into it. A better trader sizes 1%. A great trader sizes less than 1% at first and scales only on confirmation.

No amount of market knowledge survives routine 10% positions. The math wins against enthusiasm every time.

What AI does. An agent can enforce a maximum position size as a hard rule. SilvaBit AI never sizes above a configured ceiling, no matter how "good" the setup looks. Humans override themselves. Agents don't.

Failure mode 2 — Averaging into losers.

The classic retail story. A trade goes against you. You tell yourself the setup is still valid and double down. It goes further against you. You triple down. The chart recovers to half of your original entry, you're out at a horrifying loss, and you tell yourself the market is "just irrational."

It's not. You violated your own position sizing rules.

What AI does. The agent is forbidden from averaging into a losing position unless the strategy explicitly supports it with pre-trade sizing. There is no "just one more entry" mode.

Failure mode 3 — Revenge trading after a loss.

A losing trade stings. The feeling demands a response. The response is almost always another trade — this time with a worse setup, bigger size, and a shorter time horizon. Compounded over a bad week, this behaviour alone will consume a significant portion of a retail account.

What AI does. Agents don't feel anger. They don't take losses personally. SilvaBit AI follows the same evaluation on the next trade as it did on the previous — regardless of outcome.

Failure mode 4 — Not knowing which setup actually works.

Most retail traders don't have a strategy. They have a loose collection of setups, each tried a handful of times, with no systematic record of outcomes. Ask them "what's your win rate on breakout trades on 15m charts?" and the honest answer is "I don't know."

What AI does. Every trade and non-trade is logged, with reasoning. Over months you can tell which setups actually earn money and which just feel good. That log is probably the single most valuable artefact of using an agent.

Failure mode 5 — Fees that quietly eat the account.

Spread, commission, swap, financing, withdrawal fees. For an active retail trader, these can quietly erase 10–20% of a year's gross profit without anyone noticing. Many brokers rely on this.

What AI does. A clean fee architecture (SilvaBit AI runs flat 8% on withdrawn profit, shared with the broker — no spread markup) is a feature. It is dull. It is unsexy. It is also a huge part of why some accounts compound and others don't.

Failure mode 6 — Trading in windows that don't suit them.

A new trader with a day job trading at 11pm because that's when they have free time will have a terrible result. They're tired, the markets they're trading often have low volume, and they're chasing moves. Their losses are structural, not strategic.

What AI does. The agent trades when the market structure, volume and volatility support it — not when you have a free hour. This alone is a surprising amount of edge.

So why 8% and not 50%?

You might reasonably ask: if AI removes most of these failure modes, why isn't the success rate much higher? Three reasons.

1. AI doesn't fix the market. Even a perfectly disciplined account faces drawdowns. Some users close the account during a drawdown and never return.
2. AI doesn't fix capitalisation. Accounts under €200 are too small to absorb normal volatility. Losses hit faster than compounding.
3. AI doesn't fix impatience. Compounding is boring. Many users want "10x in a quarter" and quit when it's "8% in a quarter" — even though 8% in a quarter is historically excellent.

How to join the 8%.

Three practical steps:

1. Open an account and run SilvaBit AI in observation mode for 10 trading days before you fund it. Just watch what the agent would do.
2. Fund with capital you can lose entirely without affecting your life. If €1,000 would, don't fund €1,000.
3. Set loss bands at defaults. Don't touch them for 90 days. Review on day 91.

None of this is a guarantee. The 92% statistic exists because trading is genuinely hard, not because every losing trader is foolish. But the gap between the 92% and the 8% is more about discipline than cleverness — and discipline is exactly where a well-built agent can carry you.

Read our risk thresholds piece next, or open the platform overview to see the mechanics live.