📊 TradingView Backtest Deep Validation

Period 2019-12-15 – 2026-08-22 · 923 trades · 2,441 days (6.7 yrs)

Leverage 1× Commission/side 0.04% (RT 0.08%) Slippage 0% Profit share 20% (per trade) Initial capital 10,000 USDT Risk-free rate 2%

Total trades

923

2019-12-15 – 2026-08-22 · 2,441 days

Cumulative return

8,950,853.6%

after fees

Annualized return

450.8%

138 trades/yr

Sharpe

2.92

1.5+ good / 2.0+ excellent

MDD (closed trades)

-24.61%

incl. intratrade lows -25.23%

Profit factor

2.05

1.5+ good

Win rate

43.2%

payoff 3.89

Equity Curve & Drawdown

Strategy equity (month-end, log)

Longest drawdown

116 days

51 trades · 2023-07-14 ~ 2023-11-08

Days since last high

106 days

Drawdown episodes

105

peak → full recovery

Avg drawdown (trough)

-4.32%

median -3.44%

Avg underwater depth

-5.40%

avg depth while in drawdown

Avg recovery time

17 days

median 13 days

The chart aggregates month-end values only. Max drawdown is -24.61% on a per-trade basis; intra-month troughs look milder on a month-end curve. Drawdown duration measures psychological pain — the longest stretch without a new equity high was 116 days.

Top-5 Longest Drawdowns (peak → new peak)

Start	New peak	Duration	Trades	Max depth
2023-07-14	2023-11-08	116 days	51	-8.6%
2023-04-14	2023-07-14	91 days	47	-15.4%
2020-05-06	2020-07-26	80 days	42	-14.2%
2023-12-11	2024-01-26	46 days	29	-8.9%
2025-12-14	2026-01-29	45 days	22	-10.9%

Performance Metrics

Metric	Value	Benchmark
Sharpe Ratio	2.92	1.5+ good · 2.0+ excellent
Sortino Ratio	15.97	2.0+ good
Calmar Ratio	18.31	2.0+ good · 3.0+ excellent
Profit Factor	2.05	gross profit 3,631,551,095 / gross loss 1,769,562,261 USDT
Recovery Factor	5.02	net profit 1,861,988,833 / MDD 370,559,835 USDT
Kelly fraction (full)	28.64%	half Kelly 14.32%
Expectancy per trade	1.377%	138 USDT/trade (on initial capital)
Avg win / avg loss	4.81% / 1.24%	payoff 3.89
MDD (closed trades)	-24.61%	within −25% is safe
MDD (incl. intratrade lows)	-25.23%	TradingView method
Risk of Ruin	0.0000%	Z=0.3374 · 40 losses to ruin

Risk & Distribution

Metric	Value	Description
Skewness	3.102	positive = long profit tail
Excess kurtosis	10.878	higher = fatter tails
Bootstrap 95% CI	[1.020%, 1.750%]	1,000 resamples
Monte Carlo 95% / 99% worst MDD	-31.97% / -37.52%	bootstrap
Shuffle 95% / 99% worst MDD	-30.82% / -36.63%	reorder only
Monte Carlo profitable scenarios	100%	median 8,473,864% · range [865,429%, 107,127,486%]
Longest losing / winning streak	9 / 8	theoretical max streak 12.1
Capital hit at worst streak	−10.6%	loss unit 1.24%/trade

Losing-Streak Distribution

X = streak length · Y = occurrences

MDD Distribution — Simulation Methods Compared

Shuffle (reorder only)Bootstrap (resampling)

Merely reordering the trades made 28% of scenarios suffer a deeper MDD than the actual -24.6%. Design margin and psychological stop limits around the 99% value (-36.6%), not the realized MDD. There were 229 losing streaks in total, averaging 2.3 losses.

Random Missed-Trade Simulation (400 runs each) — alert→API fill risk

Miss rate	Median multiple	Worst 5%	Profit probability
5%	51,634.1×	27,294.9×	100%
10%	29,060.0×	12,880.1×	100%
20%	8,598.0×	3,586.9×	100%

Statistical Tests

Test	Result	Verdict
T-test (vs random entry)	T = 7.630 · p = 0.0000	✅ significant at 1%
Runs Test	actual 458 · expected 454.0 · Z=0.00 · p=0.9998	✅ random
Deflated Sharpe Ratio	100.0%	✅ significant
PBO (overfitting probability)	0.0% · 0 of 20 splits lost OOS	✅ low
IS/OOS (70/30)	IS 3.21 → OOS 2.27 (retention 71%)	✅ good
Execution-delay sensitivity	1-trade delay → 2.96 (1.2% change)	✅ robust
Look-ahead diagnostics	same-bar 0 · min hold 20 min · MFE capture 69%	✅ no red flags

Deflated Sharpe Ratio #32

Probability that the Sharpe exceeds the chance maximum after correcting for multiple-parameter search bias. Higher is better; 95%+ is statistically significant.

PBO — Overfitting Probability #33

Share of splits where the in-sample winner loses out-of-sample. Lower is better; below 25% means low overfitting risk.

Random Entry Test (t-distribution) #25

Sharpe by Execution Delay #24

Performance by Market Regime

Segment	Trades	Mean return	Win rate	Sharpe
Early (1/3)	307	2.14%	43.6%	3.72
Middle (2/3)	307	0.93%	42.7%	2.47
Late (3/3)	309	1.07%	43.4%	2.53

Mean Return by Segment

Historical Crisis Windows

Crisis window	Exposed	Entries in window	Strategy return	Window MDD	Win rate	Market over window	Verdict
COVID crash (2020-02–04)	9	9	+70.1%	-5.4%	56%	-42.3%	✅ defended
May-2021 crash (2021-05–07)	19	18	+19.7%	-10.8%	32%	-34.8%	✅ defended
LUNA collapse (2022-05–06)	19	19	+33.9%	-10.6%	47%	-57.2%	✅ defended
FTX bankruptcy (2022-11–12)	14	13	+23.7%	-5.6%	57%	-25.4%	✅ defended
Yen-carry unwind (2024-08)	19	18	+34.7%	-6.8%	47%	-33.2%	✅ defended
Tariff shock (2025-02–04)	26	25	+50.5%	-6.1%	50%	-44.1%	✅ defended
Oct-2025 mass liquidation (10-10–11)	1	0	+17.4%	0.0%	100%	-22.0%	✅ defended

Of the 7 crisis windows with open exposure, 7 were defended successfully. An exposed trade is one whose holding period (entry–exit) overlapped the crisis by at least a day; counting by exit time alone would mix in post-crisis trades, so overlap is used instead.

Halving-Cycle Breakdown

Cycle	Trades	Compound return	Sharpe	MDD	Win rate
Pre-3rd halving (–2020-05)	39	233.9%	5.20	-6.0%	51%
3rd cycle (2020-05 – 2024-04)	526	94,082.7%	3.05	-15.4%	43%
4th cycle (2024-04 –)	358	2,746.5%	2.47	-24.6%	42%

Volatility-Regime Breakdown

🔥 High-volatility regime

709,501.3%

451 trades · Sharpe 3.76 · WR 46%

🧊 Low-volatility regime

1,161.4%

472 trades · Sharpe 1.96 · WR 41%

High/low volatility is split at the median of the 20-trade rolling standard deviation of price log-returns.

Yearly & Monthly Returns

Compound Return by Year

Monthly Return Heatmap (%)

Time Patterns

Profit Sum by Exit Hour (simple %)

Profit Sum by Weekday (simple %)

Monthly Trade Frequency

The strongest exit hour is 5:00 (+150%) and the weakest 3:00 (-8%), with the top-3 hours concentrating 27% of profit. By weekday, Sun is strongest and Wed weakest; the weekend share is 43%. Last-12-month average 13.2 trades/month vs overall 11.4 (+16%).

Period Win Rates & Volatility

Profitable days

49%

676 trading days

Profitable weeks

59%

291 weeks

Profitable months

90%

81 months

Monthly volatility (σ)

17.0%

annualized 58.8%

Anti-Overfitting — Walk-Forward & K-Fold

WFE · rolling

83%

60%+ robust

WFE · anchored

74%

expanding window

Profitable OOS steps

6/6

Embargo buffer

18 trades

isolates train↔test

Step	Test window	IS Sharpe	OOS Sharpe	Retention	OOS return
1	2021-03-06 ~ 2022-03-09	3.16	3.81	121%	778.3%
2	2022-04-10 ~ 2023-03-20	4.54	3.59	79%	461.7%
3	2023-05-07 ~ 2023-12-22	3.48	0.95	27%	18.8%
4	2024-01-21 ~ 2024-12-10	0.70	3.36	482%	441.9%
5	2025-01-12 ~ 2025-11-28	3.11	3.18	103%	426.0%
6	2026-01-01 ~ 2026-08-08	3.27	0.27	8%	3.0%

Train vs Test Sharpe by Step

Purged & Embargoed 6-Fold Cross-Validation

Fold	Test window	Train Sharpe	Test Sharpe	Test return	Purged
1	2019-12-15 ~ 2021-03-13	2.79	3.88	2,666.4%	18
2	2021-03-15 ~ 2022-06-28	2.59	3.55	1,171.6%	36
3	2022-07-04 ~ 2023-07-22	3.02	2.79	350.5%	36
4	2023-07-23 ~ 2024-08-19	3.07	2.16	193.1%	36
5	2024-08-21 ~ 2025-09-01	2.84	3.12	754.2%	36
6	2025-09-01 ~ 2026-08-22	3.15	1.81	125.6%	18

Train vs Test Sharpe by Fold

6 of 6 folds were profitable in their test window. Walk-forward validates in time order while K-fold rotates every segment through validation — similar results from both are cross-evidence of low overfitting risk.

Leverage Analysis

Kelly-optimal leverage

11.5×

maximizes compounding

Half Kelly (recommended)

5.8×

half the volatility

1%-ruin safe cap

9.6×

ruin = 50% capital loss

Continuous Kelly (μ−rf)/σ²

4.5×

Markowitz tangency solution

Ruin at current 1×

0.00%

Sharpe 2.92

Long-Run Compound Growth by Leverage (Kelly)

Leverage	Sharpe	MDD	Ruin prob.	CAGR (%/yr)	Cumulative return
0.5×	2.89	13.0%	0.00%	145.7%	40,600.8%
1.0× 📍current	2.92	24.6%	0.00%	450.8%	8,950,853.6%
2.0×	2.94	44.1%	0.00%	2,108.4%	96,063,059,758.4%
3.0×	2.94	59.3%	0.00%	6,845.3%	2.033e+14%
5.0×	2.95	79.4%	0.02%	39,201.4%	2.181e+19%
9.6× 🛡️safe cap	2.95	96.6%	1.02%	280,239.2%	1.099e+25%
10.0×	2.95	97.2%	1.23%	298,719.3%	1.685e+25%
11.5× 🎯Kelly	2.95	98.6%	2.18%	331,605.8%	3.385e+25%

The discrete scan puts Kelly at 11.5× while the closed-form continuous solution gives 4.5×. The wider the gap, the further the return distribution sits from a lognormal approximation — a common signature of a stop-loss that artificially truncates the left tail. The textbook approach is to start live at the lower of half Kelly (5.8×) and the safe cap (9.6×).

CAGR and cumulative return in the high-leverage range are theoretical. They assume full profit reinvestment, no margin limits, no funding fees, and fixed execution slippage — so they are not achievable targets. Weight the ruin-probability and MDD columns far more heavily.

🦢 Black-Swan Injection Stress (−30% at 1×, injected 3×)

Metric	Before	After
Kelly-optimal leverage	11.5×	11.5×
Half Kelly (recommended)	5.8×	5.8×
1%-ruin safe cap	9.6×	7.5×
95% worst MDD (current 1×)	-30.8%	-54.4%

History only contains the crises that actually happened. Injecting an unseen −30% event roughly once every two years lowers the Kelly optimum from 11.5× to 11.5×. The more honest live ceiling is the post-injection figure — the lower of half Kelly 5.8× and the safe cap 7.5×.

Costs, Slippage & Funding

Scenario	Final equity multiple	Cumulative return	Sharpe	MDD
① Zero cost (ideal)	186,541.23×	18,654,023.4%	3.12	-20.8%
② Maker RT 0.04%	129,227.16×	12,922,616.2%	3.04	-22.7%
③ Taker RT 0.10%	74,490.90×	7,448,990.3%	2.91	-25.5%
④ Taker+slippage RT 0.16%	42,924.94×	4,292,394.2%	2.78	-28.3%
⑤ Current setting RT 0.08%	89,509.54×	8,950,853.6%	2.95	-24.6%

Slippage Sensitivity

Slippage (per side)	Round-trip	Final equity multiple	Cumulative return	Sharpe
0bp	0bp	89,509.54×	8,950,853.6%	2.95
1bp	2bp	74,490.90×	7,448,990.3%	2.91
2bp	4bp	61,989.95×	6,198,895.1%	2.87
5bp	10bp	35,717.42×	3,571,642.0%	2.74
10bp	20bp	14,239.45×	1,423,844.6%	2.52

Funding-Fee Approximation (perpetuals, 0.01%/8h)

Avg holding time

28.1 h

avg 337 bars × 5 min

Total funding cost

32.45%

vs capital

Equity multiple after funding

65,838.32×

before 89,509.54×

Sharpe after funding

2.89

-0.06

An average hold of 28 hours means about 3.5 funding events per trade, and funding alone removes 26% of the final equity. At the current setting (0.04% per side), each trade costs roughly 0.0800% of capital, compounded 923 times.

Profit Structure Decomposition

Top-Trade Removal — Luck Dependence

Test	Top-trade profit share	Multiple after removal	Sharpe after removal	Verdict
Remove top 5	7.8%	24,193.01×	2.83	✅ still profitable
Remove top 10	15.6%	6,538.98×	2.71	✅ still profitable
Remove top 20	27.5%	842.59×	2.42	✅ still profitable

Long / Short Breakdown

Side	Trades	Win rate	Compound return	Sharpe	MDD
🟢 Long	635	42.2%	104,702.2%	2.61	-20.7%
🔴 Short	288	45.5%	8,440.8%	3.75	-16.7%

Long–short monthly corr ρ

-0.30

more negative = internal hedge

Monthly vol (long / short)

14.6% / 10.2%

Combined monthly vol

15.1%

if uncorrelated 17.8%

Long–short monthly correlation ρ = -0.30. Negative — the short side fills in the long side's losing months, an internal hedge. Combined volatility 15.1% vs 17.8% if uncorrelated.

MAE / MFE Analysis (stop-loss & take-profit placement)

Metric	Value	Description
Avg MAE — winners	0.64%	avg adverse move after entry
Avg MAE — losers	1.52%	2.4× that of winners
Separation (Cohen's d)	1.13	0.8+ = clear separation → stop line is valid
Suggested stop (winners 90th pct)	1.49%	cuts 47% of losers earlier
MFE capture (winners)	69%	share of peak unrealized profit actually banked
Holding–return corr ρ	0.63	positive = trend-following behavior

Winners move only 0.64% against entry on average, while losers go 1.52% (2.4× the difference). Placing the stop near 1.5% would preserve 90% of winners while cutting 47% of losers earlier than now.

💰 Profit Share — 20%, Paid Per Trade

Total payout (fixed principal)

38.4K USDT

399 payouts

Total payout (compounded)

18.76M USDT

when payouts leave the principal

Effective rate vs net profit

30.2%

vs 20% nominal

Largest single payout

598 USDT

My net profit after payout

88.7K USDT

before payout 127.1K

Equity & Cumulative Payout (month-end, log scale)

Equity before payoutEquity after payoutCumulative payout

Metrics — Before vs After Payout (compounded equity)

Scenario	Cumulative return	Annualized	Sharpe	Profit factor	MDD	Final equity
① Before payout	8,950,853.6%	450.8%	2.92	1.89	-24.61%	895.10M USDT
② After per-trade payout	296,059.7%	230.7%	2.49	1.65	-27.65%	29.62M USDT

Payouts by Year (fixed principal, USDT)

Year	Net profit (pre-payout)	Per-trade payout	Kept
2019	273 USDT	82 USDT	191 USDT
2020	29.3K USDT	7,771 USDT	21.5K USDT
2021	25.8K USDT	7,092 USDT	18.7K USDT
2022	21.9K USDT	6,028 USDT	15.9K USDT
2023	5,761 USDT	3,054 USDT	2,707 USDT
2024	19.9K USDT	5,931 USDT	14.0K USDT
2025	20.7K USDT	6,256 USDT	14.5K USDT
2026	3,444 USDT	2,149 USDT	1,296 USDT

The nominal rate is 20%, but the effective rate on net profit is 30.2%. Per-trade settlement does not net losses — you share the wins and eat the losses alone — so the effective burden exceeds the nominal rate. Under compounding, payouts also take away principal that would have kept compounding, so the final-equity gap widens far beyond the rate (895.10M → 29.62M USDT).

📋 Validation Checklist (21 items)

✅ Passed

⚠️ Warning

🚨 Caution

ℹ️ Info

Items validated

computable items only

📁 1. Data Quality

1. Data Integrity Check^#1✅ Passed

Exhaustively checks data-quality issues such as missing values (NaN), infinities (Inf), duplicate rows, and out-of-order dates. Many missing values distort every metric — Sharpe, MDD, and more — making the entire validation untrustworthy.

📏 Benchmark 0 missing → perfect ✅ | 1–5 → inspect and fix ⚠️ | 5+ → strategy needs re-review 🚨

Result 0 missing / 1846 rows × 17 cols

💡 Interpretation 0 missing values — the data is complete, so every statistic below was computed from the raw records. Results are trustworthy without preprocessing distortion.

📁 2. Model & Parameter Risk

2. Walk-Forward Analysis^#3✅ Passed

Splits the full period into rolling (fixed-window) and anchored (expanding-window) segments and repeatedly checks whether performance in each in-sample (IS) segment carries into the next out-of-sample (OOS) segment. WFE (Walk-Forward Efficiency) = mean OOS Sharpe ÷ mean IS Sharpe. Embargo buffer trades between segments prevent adjacent-regime information leakage. ⚠️ Trade-history CSVs contain no parameters, so this version measures performance consistency only, without re-optimization.

📏 Benchmark WFE ≥ 60% → robust ✅ | 40–60% → moderate ⚠️ | below 40% or many OOS losses → suspect overfitting / strategy decay 🚨

Result WFE rolling 83% / anchored 74% | profitable OOS 6/6 | embargo 18 trades

💡 Interpretation Rolling through 6 steps, 6/6 OOS segments were profitable, with WFE of 83% rolling / 74% anchored. Above 60% means the training-period edge largely reappears in the next segment — good temporal robustness. An embargo of 18 trades between train and test blocks adjacent-regime leakage.

3. Out-of-Sample Test (70/30)^#4✅ Passed

Final validation on future data never used in strategy development. Going live on in-sample performance alone, without OOS validation, is a common recipe for disappointment. This tool automatically splits the trade history chronologically 70% (IS) / 30% (OOS) — see the 🧪 Deep Validation tab for detailed curves.

📏 Benchmark OOS return ≥ 60% of IS → good ✅ | 30–60% → caution ⚠️ | below 30% or negative → unfit for live trading 🚨

Result IS Sharpe 3.21 → OOS 2.27 (retention 71%) | IS 936,699.5% / OOS 855.5%

💡 Interpretation Sharpe in the untouched last 30% is 2.27, i.e. 71% of the first 70% (3.21). Retention above 60% means the edge survives market change — overfitting is unlikely.

4. PBO (Probability of Backtest Overfitting)^#33✅ Passed

CSCV (Combinatorially Symmetric Cross-Validation): splits the data into 6 blocks, forms C(6,3)=20 IS/OOS combinations, and measures how often the in-sample winner loses out-of-sample. 50% is coin-flip level — pure luck. (Accuracy degrades with few trades.)

📏 Benchmark PBO < 10% → no overfitting ✅ | 10–25% → low risk ✅ | 25–50% → caution ⚠️ | over 50% → overfitting confirmed 🚨

Result PBO 0.0% | 0 of 20 splits lost OOS | purge/embargo 18 trades/boundary

💡 Interpretation In 0% of the 20 train/test combinations, the in-sample winner lost out-of-sample. Low — a structural edge that holds however the period is sliced.

📁 3. Stress Tests

5. Monte Carlo Path Risk^#5✅ Passed

Simulates thousands of equity paths by randomly reshuffling trade order. If the actual result sits near the top of the distribution, luck likely played a role. The 95% worst-case MDD shows the realistic loss boundary.

📏 Benchmark 95% worst MDD within −25% → safe ✅ | −25% to −40% → caution ⚠️ | beyond −40% → high risk 🚨

Result 95% worst MDD -31.97% | 95% final return [865,429%, 107,127,486%]

💡 Interpretation Reshuffling the same trades 1000 times gives a 95% worst MDD of -32.0%, which is deeper than the -24.6% actually experienced — the real path was a lucky ordering. Set margin and psychological stop limits to the 95% value (-32.0%), not the realized MDD.

6. Historical Stress Test^#6✅ Passed

Tests resilience during extreme drawdown periods such as the 2020 COVID crash (−50%) and the 2022 LUNA collapse. Strategies that look fine in normal times often break in crises. This tool auto-validates against the actual date ranges of COVID, the May-2021 crash, LUNA, FTX, the 2024 yen-carry unwind, the 2025 tariff shock, and the Oct-2025 mass liquidation — details in the 🧪 Deep Validation tab.

📏 Benchmark Crisis MDD ≤ 2× normal-period MDD → robust ✅ | 2–3× → caution ⚠️ | over 3× → crisis-fragile 🚨

💡 Interpretation Judged on 'exposed trades' whose holding period overlapped a crisis. Of 7 exposed windows, 7 were defended successfully.

7. Transaction Cost Impact^#7✅ Passed

Checks whether profits survive after deducting commissions, slippage, and spread. Scalping and high-frequency strategies are extremely cost-sensitive — a 0.1% fee difference can flip annual returns.

📏 Benchmark Profit after costs > 0 → pass ✅ | profit shrinks 50%+ → consider revising ⚠️ | turns negative → unfit for live trading 🚨

Result commission 0.04%/side + slippage 0.00% (923 trades × 1×) | before 18,654,023.4% → after 8,950,853.6%

💡 Interpretation At 0.04% per side (round-trip 0.08%) × 1× leverage, the cumulative return is 8,950,853.6% — still profitable: a live-worthy edge that beats cost friction. Costs consumed 9,703,169.8pp.

📁 4. Performance Metrics

8. Kelly Criterion Allocation^#9✅ Passed

Kelly formula: f* = win rate − (loss rate ÷ payoff ratio). The theoretically optimal bet fraction that maximizes long-run geometric growth. Full Kelly is extremely volatile, so half Kelly (50%) or quarter Kelly (25%) is the practical standard.

📏 Benchmark Kelly > 0 → positive expectancy ✅ | half Kelly 5–15% → stable ✅ | 15–30% → aggressive ⚠️ | Kelly ≤ 0 → losing strategy 🚨

Result Full Kelly 28.64% | Half Kelly 14.32%

💡 Interpretation Risking 28.6% of capital per trade mathematically maximizes long-run compound growth. Full Kelly's drawdowns are extreme, though — start live at half Kelly (14.3%) and adjust.

9. Win Rate & Win/Loss Ratio^#10ℹ️ Info

Win rate and payoff ratio complement each other: a low win rate can still profit with a high payoff (trend-following), and a low payoff can profit with a high win rate (market-making). What matters is expectancy = (win rate × avg win) − (loss rate × avg loss) > 0.

📏 Benchmark Expectancy > 0 required 🚨 | WR ≥ 50% + payoff ≥ 1.3 → good ✅ | WR 40–50% + payoff ≥ 2.0 → good trend-follower ✅ | expectancy ≤ 0 → not tradable

Result Win rate 43.2% | payoff 3.89 | W 399 / L 524 / total 923

💡 Interpretation Win rate 43.2% × payoff 3.89 → expectancy +1.38% per trade. Classic trend-following profile — loses often but wins big. Losing streaks are inevitable, so money management to survive them is key.

10. Sharpe Ratio^#11✅ Passed

Sharpe = (mean return − risk-free rate) ÷ return std dev × √(trades per year). The standard performance yardstick for hedge funds and institutions. High volatility can drag it down even with high returns; crypto's volatility means different thresholds than equities.

📏 Benchmark < 1.0 → unfit 🚨 | 1.0–1.5 → moderate ⚠️ | 1.5–2.0 → good ✅ | 2.0–3.0 → excellent ✅ | above 3.0 → outstanding (re-check for overfitting)

Result Annualized Sharpe 2.92 (1.5+ good / 2.0+ excellent)

💡 Interpretation The strategy earned 2.92 units of excess return per unit of risk (volatility). Excellent — meets institutional standards.

11. Sortino Ratio^#12✅ Passed

Sortino = (mean return − risk-free rate) ÷ downside std dev × √(trades per year). Unlike Sharpe, upside volatility isn't treated as risk. Better suited to trend-following and long-biased strategies; Sortino > Sharpe indicates a favorably asymmetric profit structure.

📏 Benchmark < 1.5 → unfit 🚨 | 1.5–2.0 → moderate ⚠️ | 2.0–3.0 → good ✅ | above 3.0 → excellent ✅ | Sortino > Sharpe → positive asymmetry ✅

Result Annualized Sortino 15.97

💡 Interpretation Sortino 15.97 clearly exceeds Sharpe (2.92) — most volatility comes from the profit side. Losses are small and uniform while gains spike: a desirable asymmetry for risk management.

12. Calmar Ratio^#13✅ Passed

Calmar = annualized return ÷ |max drawdown|. Shows how much you earn per year relative to the worst loss period you had to sit through. A key screening criterion for long-horizon CTA funds; a larger MDD lowers Calmar, cleanly reflecting return-per-risk efficiency.

📏 Benchmark < 0.5 → unfit 🚨 | 0.5–1.0 → marginal ⚠️ | 1.0–2.0 → moderate | 2.0–3.0 → good ✅ | above 3.0 → excellent ✅

Result Calmar 18.31 (annual return 450.8% / MDD -24.61%)

💡 Interpretation Earning 450.8%/yr against a worst drawdown of -24.6% — one year's return covers the MDD about 18.3 times. The worst drawdown is quickly repaid by a year's returns.

13. Profit Factor^#14✅ Passed

Profit Factor = gross profit ÷ gross loss. The most intuitive profitability metric: 1.0 is break-even, 1.5 means $1.50 earned per $1 lost. Best checked both before and after trading costs.

📏 Benchmark < 1.0 → losing strategy 🚨 | 1.0–1.3 → marginal ⚠️ | 1.3–1.5 → moderate | 1.5–2.0 → good ✅ | above 2.0 → excellent ✅

Result Profit factor 2.05 (gross profit 3,631,551,095 / gross loss 1,769,562,261 USDT)

💡 Interpretation Earned 2.05 USDT per 1 USDT lost. A strong structure earning 2×+ per unit lost.

14. Recovery Factor^#15✅ Passed

Recovery Factor = net cumulative profit ÷ max drawdown amount. Shows how decisively the strategy overcomes its worst loss. Below 1.0 means the MDD hasn't been recovered yet; higher is better.

📏 Benchmark < 1.0 → MDD not recovered 🚨 | 1.0–2.0 → marginal ⚠️ | 2.0–3.0 → moderate | 3.0–5.0 → good ✅ | above 5.0 → excellent ✅

Result Recovery factor 5.02 (net profit 1,861,988,833 / MDD 370,559,835 USDT)

💡 Interpretation Net profit is 5.0× the max-drawdown amount. Even repeated drawdowns leave overwhelming net profit.

15. Risk of Ruin^#16✅ Passed

Formula: Z = (loss rate × loss size) ÷ (win rate × avg win). Z < 1 means positive expectancy. RoR = Z^N, where N = ruin-threshold loss ÷ per-trade loss. The closer Z is to 1 and the smaller N is, the higher the ruin probability. Choose the loss-unit basis (average / max / custom) in the sidebar.

📏 Benchmark RoR < 0.1% → negligible ✅ | 0.1–1% → low risk ✅ | 1–10% → caution, review leverage ⚠️ | over 10% → rethink position size now 🚨

Result Estimated ruin 0.0000% [avg loss 1.24%/trade / ruin threshold 50%]

💡 Interpretation Under current conditions (win rate 43%, avg loss 1.24%/trade), losing 50% of capital requires roughly 40 losses' worth of drawdown, with probability 0.0000%. Effectively zero — the current position size is safe from a ruin standpoint.

16. Deflated Sharpe Ratio^#32✅ Passed

Bailey & López de Prado (2014): DSR = Φ[(√(N−1)·(SR − E[SR*])) / √(1 − γ₃·SR + (γ₄−1)/4·SR²)]. The probability that the observed Sharpe exceeds the expected maximum Sharpe (E[SR*]) implied by your parameter-search count. Enter the true number of backtest attempts (n_trials) in the sidebar for meaningful results.

📏 Benchmark DSR ≥ 95% → statistically significant ✅ | 80–95% → moderate ⚠️ | below 80% → suspect multiple-testing bias 🚨 | larger n_trials tightens the bar

Result DSR 100.0% | E[SR*] 0.0000 | attempts 1

💡 Interpretation Attempts is set to 1, so no multiple-testing correction is applied (DSR 100%). If you backtested many parameter variants, entering the true count including discarded ones corrects strategy survivorship bias — reporting only the surviving winner.

📁 5. Market Regimes & External Factors

17. Market Regime Test^#17✅ Passed

Splits the backtest into thirds (early / middle / late) and compares win rate, mean return, and Sharpe across them. If only the early segment is good and performance decays later, the strategy is falling behind the market or its pattern has been arbitraged away.

📏 Benchmark All 3 segments profitable + Sharpe spread ≤ 30% → excellent ✅ | 1 losing segment → caution ⚠️ | 2+ losing → possible strategy decay 🚨

Result Early (1/3): 2.14% (WR 44%, SR 3.72) | Middle (2/3): 0.93% (WR 43%, SR 2.47) | Late (3/3): 1.07% (WR 43%, SR 2.53)

💡 Interpretation All three segments profitable — the edge worked across the whole period rather than clustering in one era.

📁 6. Statistical Tests

18. Bootstrap Confidence Interval^#20✅ Passed

Bootstraps 1,000 resamples (with replacement) from the actual trades to compute a 95% confidence interval for the mean return. If the lower bound exceeds 0, profitability is statistically significant beyond sampling error. Fewer trades widen the interval.

📏 Benchmark 95% CI lower bound > 0% → statistically significant profit ✅ | lower bound ≤ 0% → possibly luck ⚠️ | interval wider than the return itself → insufficient sample 🚨

Result 95% CI [1.020%, 1.750%] (1,000 resamples)

💡 Interpretation The 'true mean return' lies within [1.020%, 1.750%] with 95% confidence. The whole interval sits above 0% — statistically rejecting the possibility that the profits are a lucky sample arrangement.

19. Trade-Sequence Randomness (Runs Test)^#23✅ Passed

Runs test: checks whether win/loss streaks in the trade sequence are statistically more clustered than chance would produce. p > 0.05 means the sequence is consistent with randomness — no streak clustering. Pronounced streaks can make live results diverge sharply from the report.

📏 Benchmark p > 0.10 → clearly random ✅ | 0.05–0.10 → borderline ⚠️ | p ≤ 0.05 → streak pattern present, review structural bias 🚨

Result Actual runs 458 | expected 454.0 | Z=0.00 | p=0.9998

💡 Interpretation p = 1.000 — the win/loss sequence is indistinguishable from random. No hidden streak clustering, so the Monte Carlo reshuffling results (#5) can be trusted as-is.

20. Time Delay Test^#24✅ Passed

Measures how Sharpe changes when entries are delayed by 1, 2, 3, or 5 trades after the signal. Delay sensitivity means the strategy depends on razor-thin entry timing that's hard to execute live; the bigger the slippage and order-book latency, the bigger the real-world degradation.

📏 Benchmark Sharpe drop ≤ 10% after 1-trade delay → robust ✅ | 10–30% → delay-sensitive ⚠️ | over 30% or negative → high live-execution risk 🚨

💡 Interpretation Entering one trade late moves Sharpe 2.92 → 2.96 (1% change). Performance holds even when fills lag a beat — a robust signal that tolerates alert→API delays and retries.

21. Random Entry Test^#25✅ Passed

One-sided t-test: H₀ = mean return = 0. The lower the p-value, the more significantly the strategy beats random entry. A t-statistic above 3 is a strong signal, and more trades mean more confidence. Below 30 trades, statistical power drops sharply.

📏 Benchmark T-stat ≥ 3.0 + p < 0.01 → highly significant ✅ | T-stat ≥ 2.0 + p < 0.05 → significant ✅ | p ≥ 0.05 → not significant 🚨 | ≥ 100 trades recommended

Result T-stat 7.630 | one-sided p 0.0000

💡 Interpretation T=7.63: the probability that a zero-mean strategy produces this performance by chance is 0.0000. Passes even the 1% significance level on 923 trades — this edge is very unlikely to be statistical noise.