risk-mgmt

Validate any strategy in 4 stages with Quant Lab

Sweep → Walk-Forward → Monte Carlo → Multi-Symbol = the gold standard

45 min · intermediate

What you'll have when finished

  • Have one strategy validated across all 4 stages on a single window
  • Know what "passing" looks like at each stage
  • Be honest about whether the strategy is worth paper trading

Before you start

  • A strategy that fails ANY of the 4 stages is not ready for live capital, regardless of how good the others look
  • Walk-Forward is the highest-value test — most "backtested strategies" fail it
  • Monte Carlo confidence interval is your realistic worst case, not the median

Walkthrough

  1. Pick a strategy template + a single exchange + pair

    Open Quant Lab. Pick a template (e.g., "RSI Mean Reversion" or "EMA Crossover"). Pick Coinbase + BTC/USDT + 1h timeframe + last 180 days. We will use this same setup across all 4 stages so caching makes each test fast.

    Success criteria: Configuration filled in for Parameter Sweep

  2. Stage 1 — Parameter Sweep (find true optimum)

    Parameter Sweep tab. Pick one parameter (e.g., RSI threshold). Sweep range = 20 to 40 in steps of 2 (= 11 backtests). Metric = Sharpe ratio. Click Run. Read the chart: is there a clear peak (real edge) or a flat noisy line (overfit / no edge)? Note the optimal value.

    Success criteria: Chart shows a clear peak at one specific parameter value (NOT flat noise)

  3. Stage 2 — Walk-Forward at the optimal parameter

    Walk-Forward tab. Use the optimal value from Stage 1. Pick a split (e.g., 70% in-sample / 30% out-of-sample). Run. Read: in-sample Sharpe vs out-of-sample Sharpe. PASSING = out-of-sample holds ≥60% of in-sample Sharpe. FAILING = out-of-sample drops to <30% = strategy was over-fit.

    Success criteria: Out-of-sample Sharpe ≥ 60% of in-sample Sharpe

  4. Stage 3 — Monte Carlo (confidence interval)

    Monte Carlo tab. Use the same parameter + same window. Run 1000 trade-order shuffles. Read the 5th-percentile equity curve — your realistic worst case. PASSING = 5th-percentile final equity still positive. FAILING = 5th-percentile bankrupts the account = strategy is fragile to win/loss ordering.

    Success criteria: 5th-percentile final equity is positive (above starting capital)

  5. Stage 4 — Multi-Symbol (cross-asset robustness)

    Multi-Symbol tab. Same strategy at the optimal parameter. Run across BTC + ETH + SOL + AVAX + LINK on the same window. Read returns per symbol. PASSING = ≥4 of 5 are positive (even if smaller than BTC). FAILING = only BTC is positive = asset-specific overfit, will not generalize.

    Success criteria: At least 4 of 5 symbols show positive return

  6. Read the verdict

    GOLD STANDARD = all 4 stages passed. The strategy has a real edge at a specific parameter, holds out-of-sample, has positive worst-case Monte Carlo, and generalizes across assets. PARTIAL = 1-2 failures = needs more work (rebuild, retry). FAIL = 3+ failures = scrap the strategy and rethink.

    Success criteria: You can articulate the pass/fail status of each of the 4 stages

  7. Save the experiment + write notes

    Click Save Experiment after the final run. Name it descriptively (e.g., "RSI MR Coinbase BTC/USDT 1h — 4-stage validated, optimal RSI=30, Sharpe 1.2 / OOS Sharpe 0.9 / MC 5th +5% / 4-of-5 cross-asset positive"). Future-you searches this list before re-running identical experiments.

    Success criteria: Saved experiment visible in the experiments list with descriptive notes

What's next

If all 4 stages passed: paper-trade the validated strategy for 14-30 days on the validated parameter. If 1-2 stages failed: investigate which (Sweep peak? Walk-Forward gap? MC 5%? Multi-symbol concentration?) and rebuild. If 3+ failed: scrap and rethink — no amount of parameter tuning can save a fundamentally bad strategy.