This means the validation is strong enough to progress, but the next smart step is a structured paper-trading phase instead of moving straight to live capital.
This example currently behaves as a selective trend-following breakout model rather than an all-weather system that should be expected to stay equally credible across all market conditions.
The strongest fit is in orderly higher-timeframe trend conditions where continuation logic stays clean. It is not framed as a strategy that must remain active in every market phase.
One shared profile can still be a reasonable starting point, but we highlight when separate market-specific profiles would be safer than pretending one setup fits every target market with equal credibility.
Keep the configuration fixed and move to a structured paper-trading phase before any live capital is considered. Validation strength does not remove the need for implementation discipline.
This sample starts on a BTC benchmark because liquid, decision-relevant markets produce a cleaner first read than random altcoin selection.
The audit numbers come from the shared Python validation engine. TradingView and Pine can support visual verification, but they are not treated as silent source-of-truth replacements.
The report combines walk-forward splits, nearby parameter checks, hostile-window testing, benchmark context, and implementation-aware judgment before confidence is raised.
A PASS still does not automatically become a live recommendation. The report deliberately separates strong evidence from the question of what should happen next.
| Field | Value |
|---|---|
| Report ID | TA-2026-0047 |
| Strategy ID | STRAT-8C31F0A2 |
| Strategy family | Trend Following Strategy |
| Submitted strategy key | trend_following_strategy_v1 |
| Primary scope | BTC Market Index - 1D |
| Benchmark route | Benchmark-first crypto |
| Data source | Python + shared engine feed |
| Dataset window | 2017-08-17 to 2026-04-04 |
| Walk-forward windows | 7 |
| Parameter variants | 18 |
| Methodology version | 2026.04A |
| Engine version | v3.0 Bundle A |
| Evidence layer | Status | Why it matters | Level |
|---|---|---|---|
| Primary backtest metrics | Available | Core return, drawdown, profit factor, win rate, and trade count are present for the main verdict layer. | Required |
| Cross-market comparison | Available | The report can compare how the same strategy family behaves across tested markets instead of assuming identical fit. | Required |
| Walk-forward validation | Available | Unseen-window evidence is included, so the report is not relying on one full-period backtest alone. | Required |
| Parameter robustness | Available | Nearby variants were tested to show whether the edge survives reasonable parameter drift. | Required |
| Hostile windows | Available | Named hostile periods were tested, which helps separate a clean bull-market story from actual survival evidence. | Required |
| Cross-timeframe check | Partial | Helpful robustness context exists, but it remains a supporting layer rather than the main decision anchor on 1D systems. | Supporting |
| Implementation layer | Available | The report still separates validation strength from live-readiness and explains why the next move is not automatic. | Required |
| Benchmark context | Available | Rank is grounded in a comparable benchmark pool rather than being shown as a floating number without context. | Required |
| Direct strategy description | Not supplied | Optional customer context that would improve claim-vs-evidence checks, especially for non-trend systems. | Optional |
BTC is used here as a benchmark-first starting point because it is one of the most liquid and decision-relevant crypto markets. The goal is not to showcase a random symbol. The goal is to show how we start where decision quality is strongest, then widen only if the strategy earns broader validation.
This sample shows the first validation step on a benchmark market before widening to more assets. The point is to establish whether the strategy has credible edge in a decision-relevant market before claiming broader portability.
This example is intentionally framed as paper-trade next because validation strength and live readiness are different questions. Signal handling, execution discipline, and forward confidence still matter after a strong historical result.
This sample uses a benchmark-first crypto market and a single-benchmark-first validation path before any broader expansion claim is made.
Return here is measured relative to starting capital, so it can exceed 100%. A 100% return means capital doubled. A 300% return means it became four times the starting capital.
5 of 7 splits pass (71.00%). The 2022-2023 split fails with PF 0.71, -44.30% drawdown, and negative Sharpe. The strategy has a genuine edge in trending conditions but lacks a mechanism to avoid sustained bearish market phases.
Both parameters are stable across +/-20% variation. No overfitting detected through parameter sensitivity.
Failed the 2022 full bear year: -44.30% drawdown, 5 consecutive losses. The trend-following signal generates bullish flips during relief rallies within macro downtrends. Add a 200-day EMA filter to block entries when macro trend is bearish.
With only 38 trades over the full period, the Monte Carlo p-value of 0.47 means 47% of randomly shuffled sequences match or beat the observed return. The strategy's edge is real (positive PF, consistent win rate) but the low trade count limits statistical confirmation. This is a known limitation of low-frequency trend strategies - not a disqualifying finding, but a risk to note.
The finding: 2022 bear-market exposure is still too damaging for deployment confidence.
The hypothesis: the current trend logic is still too willing to participate during sustained downtrends.
The proposed change: only allow longs when price is above the 200-day EMA.
Expected impact: max drawdown could compress materially while preserving the cleaner breakout phases.
Validation plan: re-audit the same strategy after one structural change only and require walk-forward and sensitivity to remain honest.
Do not change: keep the benchmark market, timeframe, and the rest of the baseline fixed so the next audit can still explain what actually improved.
Risk: a macro filter that is too strict can reduce trade count enough that the strategy becomes less useful.
The finding: drawdown depth is still too high relative to the otherwise credible edge.
The hypothesis: the strategy is taking too much exposure during high-volatility phases where the signal quality is not improving at the same pace as the risk.
The proposed change: scale position size inversely to ATR so volatility spikes automatically reduce exposure.
Expected impact: lower portfolio-level drawdown with less need to retune the entry logic itself.
Validation plan: compare the same baseline with only the sizing model changed and check whether drawdown improves without destroying PF or trade-count retention.
Do not change: do not retune entries, exits, and sizing at the same time. Keep one clean anchor version untouched.
Risk: gentler sizing can make the strategy look safer while also reducing its practical value if the return profile collapses too much.
The finding: the 2022 bear-year failure and -44.30% drawdown still breach the deployment threshold.
The hypothesis: implementation discipline alone will not solve a structurally weak hostile-window profile.
The proposed change: stop this exact version from progressing until one structural fix has been audited cleanly.
Expected impact: less wasted time on a version that has not yet earned paper-trading confidence.
Validation plan: only return to audit after the next version is clearly defined and scoped as a single controlled change.
Do not change: do not move the no-go line after the fact or let multiple tweaks blur the cause-and-effect read.
Risk: continuing to optimize a broken baseline can create false confidence from curve-fit improvements.
| Field | Previous | Current | Change |
|---|---|---|---|
| Strategy key | No earlier version submitted | trend_following_strategy_v1 | Compare unlocks on the next version |
| Historical verdict | No earlier version submitted | PASS | Compare unlocks on the next version |
| Current recommendation | No earlier version submitted | Paper-Trading Recommended Before Live Deployment | Compare unlocks on the next version |
| Primary return | No earlier version submitted | +5674% | Compare unlocks on the next version |
| Profit factor | No earlier version submitted | 3.24 | Compare unlocks on the next version |
| Max drawdown | No earlier version submitted | -18.88% | Compare unlocks on the next version |
| Walk-forward pass rate | No earlier version submitted | 71.4% | Compare unlocks on the next version |
| Sensitivity pass rate | No earlier version submitted | 77.8% | Compare unlocks on the next version |
| Trades | No earlier version submitted | 134 | Compare unlocks on the next version |
The next audit with the same Strategy ID will compare this baseline against the newer version across verdict, recommendation quality, return, drawdown, walk-forward stability, and sensitivity. That makes follow-up audits much more useful than isolated one-off reports.
| Field | Value |
|---|---|
| Benchmark family | Benchmark-first crypto |
| Comparable benchmark pool | 12 strategies |
| Visible benchmark rows in report | 5 |
| Customer rank | 2 / 12 |
| Benchmark confidence | High |
| Strategy | Score | Avg PF | Avg DD | Avg Return | Assets Pass |
|---|---|---|---|---|---|
| macro_filter_breakout_v3 | 91 | 3.61 | -16.20% | +6280% | 3 / 3 |
| trend_following_strategy_v1 | 82 | 3.24 | -18.88% | +5674% | 3 / 3 |
| breakout_filter_v2 | 78 | 2.87 | -21.40% | +4318% | 2 / 3 |
| btc_trend_channel_v4 | 73 | 2.52 | -24.10% | +3527% | 2 / 3 |
| adaptive_trend_core_v2 | 69 | 2.21 | -28.60% | +2874% | 2 / 3 |
This example strategy currently behaves more like a selective trend-following breakout model than a broad all-weather system. The edge appears when market structure is orderly enough for continuation logic to remain credible.
Medium confidence. This sample uses strategy metadata plus observed behavior evidence. In live customer audits, a short optional strategy description improves confidence, especially for non-trend and non-breakout systems.
The current label becomes much stronger when a direct strategy description is supplied. Without that context, the product should stay more provisional and avoid pretending every strategy can be classified tightly from metrics alone.
The same core logic can remain viable across several assets without behaving identically on all of them. We use this section to explain when one market looks cleaner than another and why future refinements should be tested market by market instead of assumed universal.
When the evidence suggests that one shared configuration is no longer the most honest answer across all target markets, the report can recommend separate market-specific profiles rather than forcing one setup to serve every asset with equal credibility.
No direct strategy description was supplied in this sample. Without a customer claim, the report can only infer behavior from metrics, validation evidence, and the submitted strategy family.
The evidence still reads as a selective trend-following breakout profile with directional exposure concentrated in cleaner continuation phases rather than in every market condition.
Because no explicit claim was supplied here, this section stays provisional. In a real audit, a short strategy description makes it easier to confirm whether the observed behavior matches what the customer thought they built.
Each test in this report answers a different question about your strategy. This guide explains what each test measures, what the numbers mean, and how to use the results to improve your strategy before deploying real capital.
What it measures: The overall performance of your strategy across the full historical dataset - return, drawdown, profit factor, win rate, trade count, and time in market.
Why it matters: This is the starting point. If the full-period numbers look strong but the other tests don't confirm them, the strategy is likely curve-fitted to history. If the full-period numbers are weak, no further validation is needed.
Profit Factor >= 1.5 - the minimum threshold for a tradeable edge. PF of 2.0+ is strong. PF of 3.0+ is excellent.
Max Drawdown <= 35% - the maximum loss from peak to trough. Above 35% means most traders would abandon the strategy mid-drawdown.
Total Trades >= 8 - fewer than 8 trades means the results aren't statistically meaningful.
Time in market - how long you're actually exposed. Lower is better if returns are similar - less time at risk.
Sharpe ratio >= 1.0 - risk-adjusted return. Below 0.5 means returns don't justify the volatility taken.
What it measures: Whether your strategy works on data it has never seen before. The full dataset is divided into 7 sequential time windows. The strategy is tested on each window independently - not on the full period at once.
Why it matters: A strategy that only looks profitable because it was optimised on historical data will fail walk-forward validation. It cannot fake results on unseen data. This is the closest simulation to what will happen in live trading.
5+ of 7 splits passing means the strategy has genuine, repeatable edge across different market conditions.
A failing split tells you exactly which market condition breaks the strategy - use that to identify what fix is needed.
All 7 passing is the gold standard. Combined with other tests, this is what makes a strategy PASS-ready.
Consecutive failing splits (e.g. 2022 and 2023 both fail) suggests market-condition dependency - the strategy only works in certain market types.
What it measures: What happens to performance when each parameter is varied by +/-20% from its default value - one at a time. For example, if your ATR period is 10, we test 8 and 12.
Why it matters: A strategy that only works at exactly one specific parameter value was probably curve-fitted to that value. A robust strategy maintains its edge when parameters are slightly adjusted - because the underlying logic is sound, not the exact numbers.
STABLE - the profit factor stays above 1.5 across all variants. The edge is robust to parameter choice.
REJECT - one or more variants drops below 1.5 or turns negative. The strategy is fragile at this parameter.
A "banana shape" PF curve (one peak, drops sharply either side) is the classic overfitting signature - avoid deploying.
If all parameters are STABLE, you have good evidence the strategy edge comes from the logic, not lucky parameter selection.
What it measures: How your strategy performs during the 7 most hostile named market events in crypto history - 2018 crash, COVID, LUNA collapse, FTX fraud, and more. Each period is tested in isolation.
Why it matters: A strategy that only shows strong results because of the 2020-2021 bull run is dangerous to deploy. Most bear markets and crashes are preceded by exactly the type of signal a trend strategy generates. This test determines if your strategy survives when conditions turn hostile.
AVOIDED - 0 trades. Best case. The strategy correctly identified a non-tradeable environment and stayed flat.
PROFITABLE - took trades and made money despite hostile conditions. Exceptional.
SURVIVED - took trades with drawdown under 35%. Acceptable.
FAILED - drawdown exceeded 35%, or multiple consecutive losses. This event is the primary risk to live deployment.
If FAILED: identify which condition triggered entries during the crash and add a filter to block it (typically a macro trend filter).
What it measures: Whether your strategy's results could have happened by chance. We run 1,000 simulations where your exact trades are shuffled into random order. If most random sequences produce similar returns, the edge may be luck rather than skill.
Why it matters: With only 10-20 trades, it's possible to look profitable purely by luck. A strategy with 100+ trades and a p-value below 0.05 has statistically proven edge. A strategy with 12 trades and p-value of 0.40 may simply have been lucky on a few large winners.
p-value < 0.05 - statistically significant. Less than 5% of random sequences beat your strategy. Strong confidence.
p-value 0.05-0.20 - borderline. Some confidence, but not conclusive. Monitor live carefully.
p-value > 0.20 - not statistically significant. The results may be luck. Needs more trades before deployment.
The fix: more trades = lower p-value. Either run on more assets, shorter timeframe, or extend the test period.
Note: low-frequency strategies (1D, swing) will almost always show borderline p-values due to low trade count. This is expected - weight this finding accordingly.
What it measures: Whether the same strategy logic produces positive results on other timeframes (4H, 1H, 15m). A genuinely robust edge should be visible on at least 2 timeframes.
Why it matters: If a strategy only works on exactly one timeframe (e.g. 1D) but fails on 4H and 1H, the edge may be an artefact of that specific data granularity rather than a real market phenomenon.
Passes 2+ timeframes - strong indicator of genuine, durable edge.
Only passes 1 timeframe - not disqualifying for daily strategies (1D is structurally different from intraday), but note the limitation.
Fails all other timeframes - treat as a warning. Deploy on the tested timeframe only, with extra caution.
Intraday strategies should pass at least 2 intraday TFs (e.g. 1H strategy should also work on 4H) to be considered robust.