DCA modeled costs and candle quality: source evidence

Before you start

• Read the evidence type and method before interpreting any return.
• Use the downloadable files to inspect the complete recorded rows and limitations.

What this evidence covers

The source period is 2022-07-11 to 2026-07-11. The pack includes 6 preset records and 4 recorded modeled-cost and venue scenarios per preset. Reproducible extraction and arithmetic from stored fitted historical simulations; no fresh independent market replay

• All 6 source presets are included, with primary and secondary venue results at base and doubled modeled costs. This is a selected preset release, not the universe of strategies tried.
• Stored primary cumulative cash returns range from 31.09% to 125.63% over the declared four-year window; these are historical fitted simulation outputs, not annual or expected returns.
• Doubling modeled costs reduces stored primary cumulative returns by 5.74 to 12.36 percentage points. The largest stored primary peak-equity drawdown is 17.89%.
• The stored primary buy-and-hold return exceeds the preset return in 4 of 6 rows. These source benchmarks are context, not a risk-adjusted or independently reproduced comparison.
• The primary 15-minute BTC source reports 594 missing candles in 465 gaps, including a longest missing interval of 26 candles. Its completeness percentage alone does not establish strategy robustness.

How the pack was prepared

1. Read all typed records in src/app/data/dcaPresetBacktests.ts without selecting the best performers, changing preset parameters, downloading candles or submitting exchange orders.
2. Preserve the source period, venue names, selected release mode and 260 in-window warmup candles. The source simulator file is services/backtesting/dca_preset_backtest.py.
3. The base model declares 0.10% fees per side and 0.05% adverse slippage for market entries and modeled exits. Resting safety limits fill at their limit without an additional buy slippage charge.
4. The stored 2× scenario doubles fees to 0.20% and applicable slippage to 0.10%. Report stress minus base as a percentage-point change; do not call it a percent change.
5. The source return is cumulative net cash profit divided by starting quote cash, including terminal liquidation. Maximum adverse move below average entry and peak-equity drawdown are different measures and are separately named.
6. Export all four venue/cost combinations per preset and all available candle-quality rows. Unavailable measurements use null and a missing reason. Deduplicate shared candle datasets by source checksum.
7. Record SHA-256 hashes of the source table, simulator, stored-evidence generator, and this derivation. The JSON retains every original typed preset and candle-quality record as a source snapshot. Download the standalone reproduction script and JSON, then run node reproduce.mjs dca-cost-stress.json --out reproduced to regenerate and verify both CSV files without the repository, network, raw candles or an account.

What the evidence does not establish

• No historical simulation was independently rerun for this publication. Source verification flags are retained as provenance, not an independent audit, live execution proof, or future profitability claim.
• Presets were selected and fitted using historical data. The selection process and unknown rejected trials can bias all displayed returns upward; the positive release table does not show every losing strategy or market path.
• The adverse cost case is a stored sensitivity scenario, not a new losing-path experiment. Missing stress drawdown, stress stops, and other unavailable metrics remain null; base-case metrics must not be copied to a stress case.
• Candle gaps are described, but a controlled gap-removal or imputation sensitivity experiment has not been run. Their causal effect on returns is unmeasured.
• OHLC simulation omits some intrabar sequence uncertainty, live liquidity, queue position, outages, order rejection and venue-specific execution effects. Trade-only permissions and risk controls cannot remove market or operational losses.
• Base fees and slippage are model assumptions. Actual exchange fees, spreads, access, taxes, funding and subscription costs may differ; futures funding and liquidation are outside this spot study.
• Raw candle redistribution rights and complete independent replay inputs have not been cleared here. This pack publishes source-declared checksums, summaries and derivation code, without redistributing raw market candles.

Download and inspect the evidence

The JSON contains the structured evidence, the CSV files expose scenario and candle-quality rows, and the method note explains assumptions. Inspect negative outcomes, missing values and data gaps as well as positive results.

• All venue and cost scenarios (CSV)
• Dataset, provenance and field definitions (JSON)
• Candle quality and source checksums (CSV)
• Methodology and limitations (Markdown)
• Standalone extraction and CSV verifier (JavaScript)

Use the findings to question assumptions

Higher modeled costs, a different data venue or missing candles can change how a historical strategy is interpreted. Compare each scenario only within its stated period and assumptions. A result cannot establish native exchange or platform execution parity.

The next useful check is a reproducible replay with recorded data and configuration, including losing cases and any missing sensitivity tests. This pack makes the existing evidence inspectable; it does not silently replace those missing experiments.

• Read the full backtest methodology guide