Risk metrics

Bagian ini menjelaskan risk metrics di tab Analytics section "Risk & statistics". Metric ini lebih advanced, kalau Anda baru mulai, fokus ke yang lain dulu. Tapi kalau Anda serius soal longevity, ini wajib dimengerti.

Kelly %

Formula: W - (1 - W) / R

Di mana W = win rate (decimal), R = avg win / avg loss.

Cara baca:

• Positif (misal +5%), strategy Anda punya edge. Ukuran bet optimal adalah Kelly % × modal.
• Negatif, strategy Anda secara matematika tidak punya edge. Turunin risk atau retire model.
• Sangat tinggi (misal > 25%), over-confidence. Pasti ada outlier yang nge-inflate win rate. Use ½ Kelly untuk safety.

[!WARNING] Jangan pernah trade full Kelly. ½ Kelly atau ¼ Kelly adalah rekomendasi umum. Full Kelly punya drawdown yang sangat dalam, dan kebanyakan trader tidak punya mental model untuk bertahan.

½ Kelly adalah rekomendasi konservatif. Ngasih Anda ~75% dari growth rate dengan ~50% dari drawdown volatility.

Risk of Ruin

Probabilitas Anda kehilangan 50% modal dengan risk tetap per trade.

Risk of Ruin dihitung dengan formula:

• edge = (p × avgWin - q × avgLoss) / avgLoss
• RoR = ((1 - edge) / (1 + edge))^N
• Di mana N = unit modal / risk per trade (untuk 2% risk dengan modal 100k, N = 50)

Cara baca:

• Hijau (< 5%), sustainable, edge kuat.
• Kuning (5-20%), watch list, pertimbangkan turunin position size.
• Merah (> 20%), danger zone. Turunin risk per trade atau pause trading.

Monte Carlo simulation

Kami run 1000 simulasi dengan 100+ trade acak dari dataset Anda, hitung distribusi P&L akhir.

Cara baca:

• P50 positif, secara statistik, strategy Anda punya edge.
• P10 positif, bahkan worst case Anda tidak ngehilangin modal. Bagus.
• P10 negatif, worst case Anda losing. Question: apakah Anda siap secara psikologis?

[!INFO] Monte Carlo bukan预言. Dia hanya bilang "kalau trade Anda acak dari distribusi yang sama, kemungkinannya X." Real trading tidak acak, ada streak, regime, dan efek psikologis. Tapi ini proxy yang bagus.

Streak analysis

• Current streak, win atau loss yang lagi Anda alami sekarang. Penting buat avoid revenge trading.
• Longest win streak, bias detection. Apakah Anda over-confident setelah streak?
• Longest loss streak, survival planning. Kalau Anda drawdown 8 losses berturut, bagaimana plan Anda?

[!TIP] Kalau longest loss streak Anda 6, berarti Anda HARUS siap secara mental untuk 6 losses berturut. Kalau Anda panik di 3, Anda bakal revenge trade dan kehilangan semua gain dari streak sebelumnya.

Trading frequency

• Avg trades per active day, apakah Anda over-trading? Misal > 10 trades/hari = quality turun.
• Longest break, apakah Anda burnout? Kalau break Anda > 7 hari, mungkin Anda butuh vacation.
• Weekend trades, apakah Anda trade di weekend? Crypto yes, futures/forex sebaiknya no.

Consistency

• Monthly win rate, % bulan yang profit. Trader yang konsisten biasanya > 60%.
• Best/worst day/week/month, see your extremes. Penting untuk sizing.

Action items

Kalau metric Anda keluar dari zona sehat, ini prioritas:

1. Risk of Ruin tinggi → turunin position size (1% instead of 2%) sampai RoR < 5%.
2. Kelly % negative → pause trading, review model, atau retire.
3. Monte Carlo P10 negatif → ukuran akun Anda terlalu kecil untuk strategy ini. Tambah modal atau turunin risk.
4. Longest loss streak > 5 → punya plan sebelum streak berikutnya. Plan harus specify: stop trading untuk X waktu, atau reduce size.
5. Monthly win rate < 50% → review month-by-month. Apakah ada pattern seasonality?

Worked examples

Example 1: Healthy setup (continue as is)

Trader profile:

• Modal: $50,000
• Risk per trade: 1% ($500)
• 50 trade di FlowJob
• Win rate: 55% (28W / 22L)
• Avg win: $1,200
• Avg loss: -$500
• Kelly: 0.55 - 0.45 / 2.4 = +0.36 → 36% (using full Kelly, ½ Kelly = 18%)
• Risk of Ruin @ 2%: ((1 - 0.21)/(1 + 0.21))^50 = 0.65^50 = 0.00004% → < 1% (excellent)
• Risk of Ruin @ 5%: ((1 - 0.21)/(1 + 0.21))^20 = 0.65^20 = 0.30% → < 1% (excellent)
• Monte Carlo P10/P50/P90: -$2,400 / +$3,800 / +$9,600 (100 trade forward)
• Sharpe: 1.4
• Sortino: 2.1
• Calmar: 3.2
• Max drawdown: 8% dari peak

Verdict: ✅ Sustainable. Anda punya edge yang jelas, risk management sehat, dan growth yang konsisten.

Action: Continue. Set goal next quarter: tambah 1 setup baru, improve avg R dari 1.4 ke 1.6+.

Example 2: Marginal setup (tighten risk)

Trader profile:

• Modal: $25,000
• Risk per trade: 2% ($500)
• 40 trade
• Win rate: 50% (20W / 20L)
• Avg win: $800
• Avg loss: -$500
• Kelly: 0.5 - 0.5 / 1.6 = 0.19 → 19% (full), ½ Kelly = 9.5%
• Risk of Ruin @ 2%: ((1 - 0.156)/(1 + 0.156))^50 = 0.73^50 = 0.0004% → < 1% (still OK)
• Risk of Ruin @ 5%: 0.73^20 = 0.45% → < 1% (still OK)
• Monte Carlo P10/P50/P90: -$3,200 / +$1,800 / +$6,200
• Sharpe: 0.7 (mediocre)
• Sortino: 0.9
• Calmar: 1.1
• Max drawdown: 18% dari peak

Verdict: ⚠️ Marginal. Edge tipis (50% WR, 1.6:1 reward-to-risk). P50 Monte Carlo positif tapi tipis. Drawdown 18% udah signifikan.

Action: Reduce risk per trade ke 1% (bukan 2%). Re-evaluate 30 trade lagi. Kalau masih marginal, consider pause + strategy review.

Example 3: Danger zone (stop trading)

Trader profile:

• Modal: $10,000
• Risk per trade: 5% ($500)
• 25 trade
• Win rate: 45% (11W / 14L)
• Avg win: $1,000
• Avg loss: -$500
• Kelly: 0.45 - 0.55 / 2.0 = 0.18 (NEGATIVE) → strategy losing long-term
• Risk of Ruin @ 2%: ((1 - 0.10)/(1 + 0.10))^50 = 0.82^50 = 5.6% → borderline
• Risk of Ruin @ 5%: 0.82^20 = 21% → DANGER ZONE
• Monte Carlo P10/P50/P90: -$5,200 / -$1,400 / +$3,800 (P50 NEGATIVE)
• Sharpe: -0.3 (negative)
• Sortino: -0.2
• Calmar: -0.5
• Max drawdown: 35% dari peak

Verdict: 🚨 STOP. Strategy Anda losing secara matematis. Kelly negative. P50 Monte Carlo negatif. Max drawdown 35%. Anda udah dalam bahaya.

Action:

1. STOP TRADING immediately
2. Reduce size ke 0.5% (bukan 5%)
3. Review 25 trade terakhir, identify WHY losing (mind game section)
4. Possible root cause: bad setup, wrong sizing, revenge trading, market regime change
5. Kalau setelah review, identifikasi fixable issue → Forward Test fix-nya 30 trade
6. Kalau after 30 FT masih negative → RETIRE strategy, move on

Example 4: Recovery scenario

Trader profile:

• Modal: $40,000
• Drawdown: 25% (modal jadi $30,000)
• Win rate recent 30 trade: 48%
• Avg R recent: 0.9
• Max consecutive loss: 7

Diagnosis: Drawdown 25% = serious. Recent performance below average. Emotional state likely Tilt / Revenge.

Action plan:

1. Stop trading 2-3 hari. Reset mental.
2. Reduce size ke 0.5% per trade (sebelumnya 1-2%). Ini otomatis reduce losses 50-75%.
3. Setup quality threshold naikin ke 8+ (was 6+). Filter hanya A+ setups.
4. Forward Test model yang lagi underperform (kalau ada). Validate 30 trade.
5. Plan: "I will resume full size setelah 10 profitable trades berturut-turut dengan size kecil."

Recovery math: dari $30K dengan 1R avg 0.9, expected return 9% per 10 trade. Setelah 10 profitable trades = $32,700. Butuh ~$10K gain untuk balik ke peak. Realistic: 30-50 trade (1-2 bulan) untuk recovery.

Don't: Increase size untuk "balik modal cepet." Itu revenge trading. Itu making things worse.

Decision tree: what to do with your metrics

Setelah Anda baca Risk Metrics section di Analytics, follow decision tree ini:

STEP 1: Cek Kelly %
├─ Kelly < 0%  → PAUSE. Strategy losing. Re-evaluate model.
├─ 0% < Kelly < 10%  → Marginal edge. Trade smaller than Kelly (use ½ Kelly).
├─ 10% < Kelly < 25%  → Decent edge. Use ½ Kelly (safer).
└─ Kelly > 25%  → Suspicious. Probably outliers inflating WR. Use ¼ Kelly.

STEP 2: Cek Risk of Ruin @ 2%
├─ RoR < 1%  → Excellent. Stay the course.
├─ 1% < RoR < 5%  → Good. Acceptable risk.
├─ 5% < RoR < 20%  → Caution. Reduce size ke 1%.
└─ RoR > 20%  → DANGER. Reduce size ATAU pause. Don't risk > 5% per trade.

STEP 3: Cek Monte Carlo P50
├─ P50 > 0  → Edge confirmed (expected positive).
├─ P50 < 0, P10 > 0  → Edge fragile. Worst case still positive. Trade with caution.
└─ P10 < 0  → Risk of significant loss. Reduce size until P10 > 0.

STEP 4: Cek Max Drawdown vs current drawdown
├─ Current < 30% of max  → Normal. Continue.
├─ Current 30-70% of max  → Caution. Reduce size. Review.
└─ Current > 70% of max  → STOP. Reduce size ke 0.5%. Review strategy. Re-evaluate.

STEP 5: Cek Longest Loss Streak
├─ Streak ≤ 5  → Normal. Continue.
├─ Streak 6-8  → Plan in place for next streak. Reduce size if needed.
└─ Streak ≥ 9  → Reduce size 50%. Streak suggests larger variance than modeled.

ACTION: Set targets for next 30 trade. Re-evaluate. Adjust if needed.

FAQ risk metrics

Q: Kapan metric reliable?

Butuh minimal 30 trade untuk basic confidence, 50+ untuk statistical significance, 100+ untuk robust pattern detection. Di bawah 30, treat sebagai "early signal" bukan "definitive truth".

Q: Kalau Kelly % berubah signifikan antara calculation methods (analitik vs spreadsheet), kenapa?

Kalau Anda compute Kelly dengan cara berbeda (raw % vs adjusted % vs fractional), hasilnya beda. Kami pakai W - (1-W) / R (formula standar). Kalau Anda pakai spreadsheet dengan different formula (e.g., W - (1-W) / (R+1)), hasilnya beda 5-10%. Tolerable. Tapi kalau beda 30%+, ada bug atau input error.

Q: Bagaimana cara improve Sharpe ratio?

Sharpe = avg return / std dev of returns. Improve dengan:

1. Increase avg R (better entries) → increase numerator
2. Reduce variance (consistent execution) → decrease denominator
3. Both (the holy grail)

Concrete actions:

• Filter by setup quality: trade hanya 8+
• Trade di hours/volatility yang match setup Anda
• Reduce size during drawdown (less variance)
• Cut losers faster, let winners run (asymmetric)

Q: Kapan compute Monte Carlo? Daily, weekly?

Trade baru → recompute on next page load. Real-time (kalau Anda refresh Analytics tab). Cost: ~5 detik untuk 1000 trade. Acceptable.

Q: Risk of Ruin gw 0%. Apakah strategy gw perfect?

No. RoR = 0% artinya math model bilang "given current edge, very unlikely to lose 50%." Tapi model punya asumsi:

• Distribusi return stabil (market regime tidak berubah)
• Anda follow risk management rules (tidak ngelanggar)
• Variance tidak meningkat

Realitanya, semua 3 bisa break. Treat RoR 0% sebagai "very safe under current conditions," bukan "invincible."

Q: Bisa leverage Kelly untuk compound faster?

Yes, dengan discipline. Use ½ Kelly (atau ¼ Kelly untuk safety). Compound dari gains. Jangan naikin size setelah drawdown (revenge). Jangan skip risk management rules (even temporarily).

Realistic compounding: 30-50% per year dengan ½ Kelly + risk management yang baik. Bukan 100%+.

Q: Kapan Monte Carlo P10 = P50 = P90 = 0 (no edge)?

Kalau WR × avg_win = WR × avg_loss (i.e., breakeven). Strategy tidak punya edge. Monte Carlo semua P = 0. Retire strategy atau re-evaluate fundamental.

Q: Metric mana yang paling penting?

Kalau hanya bisa lihat satu: Risk of Ruin @ 2%. Ini paling actionable. Kalau RoR > 5%, turunin size. Itu simple decision.

Tambahan: current drawdown vs max drawdown. Kalau current > 70% dari max, STOP. Drawdown berikutnya bisa > max (left tail risk).

Q: Anda punya risk management rules, di mana simpen?

Di FlowJob, Anda bisa:

• Set rules di daily journal pre-market section setiap pagi (reaffirm)
• Tulis di notes app / Notion
• Print + taruh di meja (physical reminder)
• Set calendar reminders untuk re-read rules (spaced repetition)

Best: kombinasi. Affirm di daily journal + physical reminder.