Trading / Research diagnostics
20260909T034008.157793Z-98aa0646
Local artifact · no external services
What survives
the strategy search?
Backtest overfitting diagnostic: compare the selected Sharpe ratio with what repeated searching can produce under the declared assumptions.
Synthetic demonstration
Selected training Sharpe
1.71
Annualized · selected in sample
Selected holdout Sharpe
1.24
Annualized · informational only
Deflated Sharpe diagnostic
50.1%
Declared screen ≥ 95.0%
Expected maximum Sharpe
1.71
Annualized · selection-adjusted benchmark
The search changes the benchmark
Synthetic zero-drift prices · educational demonstration
Below the declared diagnostic threshold
This screen applies to the reported training DSR. Passing is not a strategy-validation decision.
Selection contract
Highest training per-period sample Sharpe; lexical name tie break. Holdout never selects the winner.
- Selected trial
- random_00729
- Declared trials
- 1,000
- Evaluated trials
- 1,000
- Effective independent trials
- 1,000.00
- Sharpe dispersion · per period
- 0.03315
- Periods per year
- 252
- Training observations
- 756
- Holdout observations
- 252
- Training interval
- 2000-01-02T00:00:00+00:00 → 2002-01-26T00:00:00+00:00
- Holdout interval
- 2002-01-27T00:00:00+00:00 → 2002-10-05T00:00:00+00:00
Follow the selected path
Compounded excess-return indices, each starting at 1; these are not account wealth. The holdout is reported after selection and is not used to rank trials.
Training · in sample
Holdout · out of sample within this run
Out of sample describes this run's split. Prior use of imported data cannot be verified here. Display paths are sampled to approximately 1,200 points when longer; full returns remain in the run exports.
Selected return profile
Sample moments expose assumptions a single Sharpe ratio hides.
| Metric | Training | Holdout |
|---|---|---|
| Sharpe · per period | 0.1080 | 0.0783 |
| Mean return · per period | 0.001080 | 0.000788 |
| Return standard deviation | 0.010001 | 0.010062 |
| Skewness | 0.032 | 0.231 |
| Kurtosis · non-excess | 2.972 | 2.719 |
| Lag-1 autocorrelation | 0.022 | 0.022 |
| PSR · zero-Sharpe benchmark | 99.8% | — |
How much searching?
Sensitivity to assumed independent trial counts; the same Sharpe dispersion and selected return moments are held fixed.
| Assumed trials | Expected max SR · ann. | DSR |
|---|---|---|
| 1.00 | 0.00 | 99.8% |
| 2.00 | 0.27 | 99.4% |
| 10.00 | 0.83 | 93.7% |
| 100.00 | 1.33 | 74.6% |
| 1,000.00 | 1.71 | 50.1% |
These scenarios do not estimate independence from strategy correlation or recover unlogged experiments.
Simple comparators
Annualized Sharpes. An undefined Sharpe, including a constant-return cash series, is shown as —.
| Baseline | Training | Holdout |
|---|---|---|
| First predeclared candidate: random_00001 | 0.01 | 1.16 |
| Buy and hold | 0.28 | 0.40 |
| Cash: zero excess return (zero variance; Sharpe undefined) | — | — |
Inspect the full search
Training ranks determine the selection. Use the column buttons to inspect results; changing the display does not change the selected trial.
| random_00729Selected on training data | 1.71 | 1.24 | 99.8% | 50.1% |
| random_00657 | 1.52 | 0.58 | 99.6% | 36.6% |
| random_00047 | 1.48 | -0.38 | 99.5% | 34.3% |
| random_00061 | 1.46 | -0.49 | 99.4% | 33.3% |
| random_00866 | 1.45 | 0.65 | 99.4% | 32.3% |
| random_00730 | 1.44 | 0.33 | 99.4% | 32.1% |
| random_00939 | 1.34 | -2.50 | 98.9% | 25.9% |
| random_00081 | 1.33 | 0.38 | 98.9% | 25.5% |
| random_00278 | 1.32 | 0.25 | 98.9% | 24.6% |
| random_00107 | 1.31 | -0.35 | 98.8% | 24.5% |
| random_00585 | 1.31 | -1.11 | 98.8% | 24.4% |
| random_00774 | 1.30 | -0.23 | 98.8% | 24.0% |
| random_00031 | 1.30 | -2.58 | 98.8% | 23.9% |
| random_00743 | 1.27 | -0.61 | 98.7% | 22.1% |
| random_00754 | 1.27 | -1.88 | 98.6% | 22.1% |
| random_00170 | 1.24 | 0.06 | 98.4% | 20.5% |
| random_00017 | 1.23 | -1.18 | 98.3% | 20.0% |
| random_00696 | 1.22 | -0.24 | 98.2% | 19.7% |
| random_00772 | 1.22 | 0.84 | 98.3% | 19.5% |
| random_00157 | 1.18 | -1.05 | 97.9% | 17.8% |
| random_00997 | 1.15 | -0.53 | 97.7% | 16.7% |
| random_00145 | 1.15 | -0.82 | 97.7% | 16.4% |
| random_00638 | 1.14 | -0.44 | 97.6% | 16.4% |
| random_00595 | 1.14 | -0.04 | 97.6% | 15.9% |
| random_00241 | 1.12 | 0.30 | 97.4% | 15.4% |
| random_00960 | 1.10 | -0.03 | 97.2% | 14.6% |
| random_00280 | 1.10 | 0.41 | 97.1% | 14.5% |
| random_00886 | 1.08 | 0.65 | 96.9% | 13.5% |
| random_00074 | 1.07 | -0.97 | 96.8% | 13.4% |
| random_00817 | 1.07 | -1.81 | 96.8% | 13.3% |
| random_00264 | 1.06 | -0.26 | 96.7% | 12.9% |
| random_00750 | 1.06 | 0.72 | 96.6% | 12.9% |
| random_00683 | 1.06 | 0.94 | 96.6% | 12.8% |
| random_00224 | 1.04 | 1.13 | 96.3% | 12.1% |
| random_00132 | 1.02 | -0.56 | 96.2% | 11.5% |
| random_00674 | 0.97 | -0.68 | 95.3% | 10.1% |
| random_00890 | 0.97 | 0.18 | 95.4% | 9.8% |
| random_00773 | 0.96 | -0.37 | 95.2% | 9.7% |
| random_00513 | 0.96 | 1.91 | 95.2% | 9.8% |
| random_00622 | 0.96 | -1.50 | 95.2% | 9.6% |
| random_00409 | 0.95 | -1.21 | 94.9% | 9.3% |
| random_00957 | 0.95 | -1.48 | 94.9% | 9.3% |
| random_00335 | 0.95 | -0.06 | 95.0% | 9.2% |
| random_00410 | 0.94 | -0.60 | 94.8% | 9.1% |
| random_00586 | 0.94 | 0.94 | 94.8% | 9.1% |
| random_00558 | 0.93 | 1.82 | 94.5% | 8.7% |
| random_00146 | 0.93 | 0.87 | 94.4% | 8.8% |
| random_00460 | 0.91 | 0.63 | 94.2% | 8.4% |
| random_00397 | 0.91 | -0.14 | 94.4% | 8.2% |
| random_00417 | 0.91 | 1.61 | 94.3% | 8.3% |
| random_00882 | 0.91 | 0.04 | 94.3% | 8.3% |
| random_00189 | 0.91 | -0.10 | 94.1% | 8.3% |
| random_00982 | 0.90 | 0.39 | 94.2% | 7.9% |
| random_00121 | 0.90 | 0.33 | 94.0% | 7.9% |
| random_00641 | 0.90 | 0.53 | 93.9% | 7.9% |
| random_00072 | 0.89 | 2.59 | 93.9% | 7.8% |
| random_00317 | 0.88 | 0.04 | 93.6% | 7.7% |
| random_00392 | 0.88 | 0.21 | 93.7% | 7.5% |
| random_00879 | 0.88 | 0.23 | 93.7% | 7.5% |
| random_00196 | 0.88 | -1.34 | 93.5% | 7.4% |
| random_00637 | 0.87 | 1.57 | 93.4% | 7.4% |
| random_00085 | 0.86 | 1.01 | 93.2% | 7.0% |
| random_00125 | 0.86 | 1.60 | 93.2% | 6.9% |
| random_00379 | 0.85 | -0.79 | 93.1% | 6.8% |
| random_00805 | 0.85 | -0.20 | 93.0% | 6.8% |
| random_00016 | 0.85 | -0.14 | 92.9% | 6.6% |
| random_00450 | 0.85 | -0.48 | 92.8% | 6.7% |
| random_00672 | 0.84 | -0.94 | 92.6% | 6.5% |
| random_00130 | 0.83 | 0.84 | 92.5% | 6.4% |
| random_00212 | 0.83 | 0.89 | 92.4% | 6.3% |
| random_00499 | 0.82 | -1.00 | 92.2% | 6.2% |
| random_00480 | 0.82 | 2.03 | 92.2% | 6.1% |
| random_00534 | 0.82 | 0.06 | 92.1% | 6.0% |
| random_00120 | 0.81 | -2.53 | 92.0% | 6.0% |
| random_00722 | 0.81 | 0.69 | 92.0% | 5.9% |
| random_00118 | 0.81 | -1.04 | 92.0% | 5.8% |
| random_00056 | 0.81 | -0.82 | 91.8% | 5.9% |
| random_00168 | 0.80 | -0.41 | 91.7% | 5.7% |
| random_00603 | 0.80 | 0.63 | 91.6% | 5.7% |
| random_00561 | 0.80 | -0.35 | 91.6% | 5.7% |
| random_00483 | 0.79 | 0.30 | 91.5% | 5.5% |
| random_00627 | 0.77 | -0.83 | 91.0% | 5.2% |
| random_00414 | 0.77 | 1.48 | 91.0% | 5.2% |
| random_00350 | 0.77 | -0.83 | 90.9% | 5.1% |
| random_00782 | 0.77 | 0.07 | 90.8% | 5.0% |
| random_00169 | 0.76 | 0.44 | 90.4% | 4.9% |
| random_00834 | 0.75 | 0.44 | 90.3% | 4.9% |
| random_00665 | 0.75 | 0.36 | 90.3% | 4.7% |
| random_00154 | 0.75 | -0.91 | 90.1% | 4.7% |
| random_00273 | 0.74 | -0.48 | 90.1% | 4.7% |
| random_00268 | 0.74 | -1.67 | 90.1% | 4.7% |
| random_00183 | 0.74 | -1.43 | 90.1% | 4.7% |
| random_00728 | 0.74 | 0.84 | 89.9% | 4.6% |
| random_00162 | 0.74 | -0.88 | 89.9% | 4.6% |
| random_00756 | 0.73 | 0.86 | 89.8% | 4.4% |
| random_00215 | 0.73 | -0.86 | 89.6% | 4.5% |
| random_00445 | 0.73 | -0.55 | 89.5% | 4.4% |
| random_00487 | 0.72 | -0.04 | 89.4% | 4.3% |
| random_00052 | 0.72 | 0.56 | 89.3% | 4.3% |
| random_00112 | 0.71 | -0.55 | 89.1% | 4.2% |
Showing first 100 of 1000 trials
Sharpe columns are annualized. PSR and DSR use training per-period returns. Trial-by-trial diagnostics are descriptive; inspecting them introduces further selection choices.
Interpretation and limits
- DSR is an assumption-dependent statistical diagnostic, not a posterior probability that a strategy is real, a probability of backtest overfitting (PBO), or a forecast of live returns.
- Selection uses training data only. Holdout results are informational; imported data cannot establish that this window was previously untouched. Reusing this report to revise a strategy consumes the holdout.
- Annualized Sharpe uses illustrative square-root-of-periods scaling, which assumes independent, identically distributed returns. Dependence can invalidate the scaling and the diagnostic calibration.
- The declared trial count must include prior searches and failed or missing variants. Effective independent trials and the cross-trial Sharpe dispersion are assumptions; correlated strategies can make a simple trial count misleading.
- A diagnostic screen does not establish an economic edge. Leakage, selection and survivorship bias, multiple testing, nonstationarity, and unmodeled trading costs require separate review.
- DSR is an asymptotic, assumption-dependent selection diagnostic, not the posterior probability that a strategy is real, not PBO, and not an expected live return.
- DSR and square-root annualization assume independent observations and sufficiently stable finite moments; autocorrelation is descriptive only and is not corrected.
- The effective independent trial count is declared, not estimated. Correlated, adaptive or unrecorded searches may invalidate the benchmark. A full-count scenario is not a calibrated dependence correction.
- A single holdout split is descriptive evidence; imports or previously inspected data cannot be certified untouched. No walk-forward, regime, liquidity or capacity validation is established.
- Cross-trial Sharpe dispersion and return skew/kurtosis are estimated with uncertainty. An expected maximum is not a family-wise significance threshold; a DSR screen is not a multiple-testing guarantee.
Run-specific warnings
- Synthetic prices and dates are fabricated. The 252 periods/year default is a display convention; no actual trading calendar is simulated.
- Price-independent unit long/short schedules use 0.0 bps per unit one-way turnover, charged on segment entry, changes and exit.
- Long/short price-return arithmetic is hypothetical. Borrow, funding, market impact, spread beyond declared fee, liquidity, latency and executable fills are not established.