Signs Your Trading System Is Failing

No system works forever. The skill is not in denial — it's in spotting the failure early enough to act.

Why systems decay

Markets evolve. Edges that existed because of behavioral bias, market structure, or technology arbitrage can disappear when:

• Other traders exploit the same edge until it's competed away
• Market structure changes (e.g., decimalization in equities, MiFID II in Europe)
• Volatility regime shifts (low-vol mean reversion fails in high-vol environments)
• New participants enter (high-frequency firms, algorithmic funds)
• Instruments change (forex pair liquidity shifts, futures contract specifications update)

A system that worked for 5 years can die in months. Recognizing failure early protects capital.

The baseline: define "normal" first

To detect failure, you must first know what "normal" looks like. Define these from your backtest and forward test:

• Expected win rate (with confidence band, e.g., ±5%)
• Expected expectancy per trade (with confidence band)
• Expected maximum drawdown in R
• Expected trade frequency per week
• Expected average slippage

Without baselines, "it's not working" is just a feeling.

Hard warning signs

1. Expectancy goes negative over 50+ trades

A single bad month is noise. Negative expectancy over 50+ trades, with the rules followed exactly, is a real signal.

2. Win rate drops 10+ percentage points below baseline

If baseline was 50% and you're at 35% over 50 trades, the setup has lost its edge.

3. Trade frequency collapses

If your system used to fire 10 times a week and now fires 2, the conditions it exploits are rarer — or your filters are too restrictive for the new regime.

4. Trade frequency spikes

Conversely, if a system that fired 10 times a week now fires 30, the market may be in a regime that triggers spurious signals — usually a sign the system is misfiring.

5. Drawdown exceeds 1.5× backtested MDD

A drawdown beyond 1.5× the backtested maximum is statistically improbable for a working system. Investigate.

6. Drawdown exceeds 2× backtested MDD

Stop trading. Either the system is failing or the regime has fundamentally changed.

7. Win/loss distribution shifts

Average win shrinks or average loss grows beyond historical norms. Stops are getting hit more often or at worse prices — execution quality has degraded or volatility is shifting.

Soft warning signs

8. Slippage increasing steadily

Stop fills are 2 ticks worse than trigger, vs. 0.5 ticks in the backtest. Liquidity is thinner or your broker's execution has degraded.

9. Spread widening

Average spread on your instrument is 30% wider than during backtest. This alone can flip expectancy from positive to negative.

10. Loss clustering

Losses come in groups of 5–7, instead of being scattered. The market regime no longer suits the system.

11. Out-of-sample performance breaks down

In walk-forward analysis, recent OOS windows show declining performance. The system is decaying.

12. New max drawdowns set every few months

Each drawdown is a new record. The trend in max drawdown is upward.

Diagnostic steps when signs appear

1. Confirm rules were followed: review every trade — manual errors cause false alarms
2. Check execution quality: pull broker reports, compare slippage to baseline
3. Inspect regime: has volatility shifted? Is liquidity different? Has the instrument's character changed?
4. Re-run backtest on most recent 6 months: is the edge gone in the recent past?
5. Compare to peers: are other systems on the same instrument also failing? Likely regime-driven.
6. Check for instrument changes: contract rolls, ticker changes, broker updates

What to do if failure is confirmed

Reduce size immediately

Cut risk per trade to 50% of normal. This slows drawdown while you investigate.

Pause trading

If signals confirm failure, stop entirely. Capital preservation matters more than FOMO on missed trades.

Investigate root cause

Determine whether the failure is:

• Regime shift (temporary — system may recover)
• Structural change (permanent — system is dead)
• Execution decay (broker or instrument issue)
• Parameter drift (filters no longer relevant)

Decide: revive, retire, or re-specify

• Revive: if regime shift is temporary, wait it out with reduced size
• Retire: if structural change is permanent, stop using the system
• Re-specify: if parameters have drifted, re-test with new parameters (with full out-of-sample validation)

Common mistakes in failure detection

• Holding on too long because of sunk cost (effort spent building the system)
• Trading through drawdowns hoping for "the rebound"
• Adding funds to "make it back faster"
• Ignoring soft signs until hard signs become undeniable
• Comparing to backtest best-case instead of median performance

The abandonment rule

Define in advance: "If expectancy is negative over 50 trades AND drawdown exceeds 1.5× baseline AND execution quality is unchanged, I stop trading this system for 30 days and re-evaluate."

A pre-defined abandonment rule removes the emotional debate. Without one, traders rationalize holding losers far too long.

The meta-skill

Building a system is one skill. Knowing when to retire one is rarer and arguably more important. Many professional traders are profitable not because their systems are better, but because they kill failing systems faster.

---

Next: rather than rely on a single system, build a portfolio of systems with low correlation.