Video: Why 90% of Traders Keep Changing Strategies (And Still Lose)
Creator: Jason Graystone
Original Video: Watch the original video
Difficulty: Intermediate
The Big Idea
Before replacing a strategy after disappointing live results, establish what actually changed. Jason Graystone proposes comparing a rules-based reference return with the trader’s actual return over the same opportunities. The difference can help identify missed setups, altered exits, invalid trades, and execution problems.
The comparison is useful only when the reference is credible. A reconstructed “perfect” return is a hypothetical counterfactual, not automatically money the trader could have earned. It must use rules defined before the outcome, executable fills, realistic costs, and the same opportunity set as the live record.
The video’s title uses a 90% figure without supplying a population study in the transcript. Its numerical examples are demonstrations, not proof that most traders already have a profitable strategy. The episode also promotes the creator’s Trader Performance OS product. This guide explains the audit method independently and does not verify that software’s calculations or claims.
Freeze the Strategy Before Comparing Results
A strategy includes more than an entry pattern. Record the instrument, timeframe, permitted session, setup definition, trigger, initial stop, position sizing, targets, management changes, invalidation, and conditions that prevent an entry. Give that configuration a version identifier.
If you widen stops, move targets, exclude Mondays, or add a confirmation indicator, you have changed the model. Historical results for version one do not establish the performance of version two. Keep legitimate research changes separate from unplanned deviations during live execution.
Write the execution assumptions too. Does a close-based signal enter at that close or the next available price? Can two positions overlap? How are unfilled limits, session cutoffs, data gaps, and bars touching both stop and target handled? Without those details, two equity curves can disagree even when the trader followed every visible chart rule.
Use the Same Live Period and the Same Opportunities
Do not compare three disappointing live months with an unrelated profitable historical period. Run the frozen reference model over the live dates, on the relevant data, with documented assumptions. Then match every valid opportunity to its actual action.
A journal containing only executed trades cannot, by itself, establish the outcomes of every missed valid entry. To measure those, keep an independent signal log or conduct a complete replay of the period using the unchanged rules. Record missed winners and missed losers, not just the attractive trades visible afterward.
Discretionary setups require additional care. If the rules depend on judgments such as “clean structure” or “strong momentum,” annotate evidence before seeing the final result, ideally through blinded replay or time-stamped observations. Otherwise, the reference may quietly admit winners and reject losers with hindsight.
Measure the Execution Gap in Consistent Risk Units
Let the reference outcome for opportunity i be Sᵢ and the actual outcome be Aᵢ. Define signed discrepancy = Aᵢ − Sᵢ. Negative values mean the actual result fell below the reference; positive values mean it exceeded the reference. Use the same normalization for both.
For this audit, 1R can be the planned monetary risk of the frozen reference trade. If that opportunity risks $100, divide both the reference and actual dollar results by $100. If the trader doubles size, do not divide the actual result by the new $200 risk and conceal the size deviation. Separately record actual initial risk and the size ratio.
The video’s +46R reference and +7R actual example gives a signed difference of −39R. Its +20R and +3R illustration gives −17R. These describe arithmetic gaps in hypothetical examples; they do not prove that all of the difference is recoverable through discipline.
Original Example: Why Net Discrepancy Can Hide Mistakes
| Opportunity or action | Reference R | Actual R | Actual minus reference |
|---|---|---|---|
| Valid winner closed early | +2.0 | +0.8 | −1.2 |
| Valid setup skipped; it later loses | −1.0 | 0 | +1.0 |
| Valid setup skipped; it later wins | +2.0 | 0 | −2.0 |
| Trade outside the strategy wins | 0 | +1.5 | +1.5 |
| Valid losing trade follows the plan | −1.0 | −1.0 | 0 |
| Total | +2.0 | +1.3 | −0.7 |
This illustrative sample has 3.2R of negative discrepancies and 2.5R of positive discrepancies. They offset to a net −0.7R, yet the sum of absolute differences is 5.7R. Small net underperformance therefore does not establish consistent execution. Skipping a loser or taking an invalid winner can mask other deviations.
For an invalid trade with no reference risk, choose and document a fixed audit unit, such as that day’s standard trade budget. Do not silently mix differently defined R values. The table assumes a common $100 unit solely to make the comparison transparent.
The transcript refers to gross and net discrepancy without fully specifying its formulas. The signed total and absolute total above are this guide’s proposed definitions, not verified definitions of the creator’s product. Absolute discrepancy measures divergence, not an amount the trader “lost.” A favorable deviation also is not automatically luck: classify the rule violation separately from the outcome.
Separate Behavior From the Execution Environment
A reference-to-live gap can have several causes. Distinguish them before assigning blame:
- Missed or extra opportunities: valid entries omitted or trades added outside the plan.
- Management deviations: unauthorized stop moves, early exits, or changes in target.
- Sizing differences: monetary results changed because exposure differed.
- Implementation differences: latency, spreads, slippage, commissions, liquidity, and broker restrictions.
- Reference-model errors: hindsight selection, impossible fills, future-data leakage, or ambiguous rules.
These categories can interact. An early entry changes both fill price and stop distance; a larger position may also receive worse execution. Define an attribution order or report a combined effect instead of adding overlapping estimates and counting the same loss twice.
Use executable reference fills rather than the best price inside a candle. TradingView’s strategy documentation (paid link) explains that simulated results require scrutiny for lookahead, selection bias, overfitting, and execution assumptions. A reference containing future information can make disciplined live trading appear deficient.
Plot Two Curves Without Changing Their Meaning
For a process audit, cumulative R curves can show the reference and actual outcomes using a shared opportunity timeline. Include zero actual outcomes for skipped signals and zero reference outcomes for invalid trades. Agree on whether results are booked by opportunity completion, exit time, or daily valuation; using different clocks can create apparent temporary gaps.
A cumulative R comparison is not necessarily an executable dollar-equity curve. Missed entries can free capital for other trades; altered exits can change whether later signals are available. If the system permits only one position at a time, simulate the full reference portfolio with its own position constraints rather than assuming every missed winner could simply be added to the actual account.
When comparing dollar curves, keep starting capital, sizing, compounding, currency conversion, and withdrawals consistent. When comparing process adherence, report the common-unit R audit alongside those cash results.
Segment the Gap, Then Test the Explanation
Graystone suggests reviewing results by instrument, system, session, day, timeframe, and entry type. That can replace a vague complaint with an identifiable pattern, such as early exits during New York after two consecutive losses.
Start with a small number of categories chosen for a reason. Display opportunity count, actual trade count, reference net result, actual net result, discrepancy, and adherence rate for each. A subgroup with three trades should not receive the same confidence as one with 100.
Searching dozens of combinations can produce an apparently terrible subgroup by chance. Do not automatically exclude an instrument because it is the worst recent slice. Separate discovery from validation: investigate the explanation on one period, then test the proposed correction on later untouched data. Save the original reference version so the audit remains interpretable.
Choose the Fix That Matches the Diagnosis
| Observed pattern | Question to resolve | Possible next step |
|---|---|---|
| Reference remains positive; actual repeatedly exits early | Were those exits deviations, or was the plan unclear? | Clarify the rule and test a predefined management workflow. |
| Reference and actual both deteriorate with strong adherence | Is this expected variance, a regime change, or a weak model? | Review the strategy and reduce or pause exposure under preset criteria. |
| Gap appears mostly in fast or illiquid conditions | Does the reference underestimate execution costs? | Update the fill assumptions and evaluate implementation. |
| Actual exceeds reference through invalid winners | Is favorable P&L reinforcing violations? | Review adherence separately; research any new setup as a new version. |
Positive backtest expectancy over 200 trades is evidence to examine, not proof of a permanent edge. Sample independence, market coverage, parameter selection, costs, and uncertainty all matter. Conversely, a short losing period does not prove the strategy failed. Set review and risk-reduction thresholds before the drawdown so the response is consistent.
The transcript’s later example mentions 3.3R expectancy while discussing totals of 27R and 4R over 100 trades. Those figures do not provide a coherent shared benchmark as stated: 3.3R per trade would imply 330R of expected arithmetic return over 100 opportunities under fixed assumptions. Do not silently reinterpret it as 0.33R or use it as measured performance.
A Practical Weekly Audit
- Export fills, fees, screenshots, and the complete opportunity log for the frozen strategy version.
- Reconstruct the reference using information and orders that could have existed at the time.
- Match opportunities, including skipped losses, skipped winners, and invalid trades.
- Calculate signed and absolute discrepancies using consistent units and realistic costs.
- Classify causes without double-counting interacting effects.
- Choose one measurable correction and assess it on a later sample while retaining the baseline.
A useful correction is specific: “For the next review period, follow the documented exit rule on valid setups and record any exception before changing the order.” “Be more disciplined” offers no observable way to assess whether the problem improved.
Final Takeaway
Audit the difference between a credible reference and actual execution before replacing a strategy. Include all opportunities, normalize risk consistently, account for executable fills, and distinguish signed underperformance from total deviation. The goal is to identify whether the next change belongs in the strategy, the trading workflow, or the model used to judge them.
Risk Disclosure
This article is educational and is not financial advice or a recommendation to trade or buy trading software. Trading can cause substantial losses. Reference returns, reconstructed opportunities, and backtests are hypothetical and depend on rules, data, timing, costs, and fill assumptions. They cannot guarantee achievable live results. Slippage, spreads, commissions, liquidity, gaps, market regimes, portfolio constraints, and discretionary judgments may materially change performance. Historical results do not guarantee future outcomes. Validate the reference and risk controls before using an audit to increase exposure.
CandleChasers independently interprets this source and is not affiliated with or endorsed by Jason Graystone or Trader Performance OS. Product and performance claims belong to the creator; the audit definitions, examples, and decision framework here are CandleChasers’ own educational work.