{
  "id": 9406503,
  "title": "My crypto backtest had a sharpe of 13.4. Then I added trading costs",
  "url": "https://urgent.news/2026/09/23/my-crypto-backtest-had-a-sharpe-of-13-4-then-i-added-trading-costs",
  "topic": "finance",
  "section": "Finance & Markets",
  "published": "2026-09-23T19:53:44.000Z",
  "source": {
    "name": "Dev.to",
    "slug": "dev-to",
    "url": "https://dev.to/pavel_kkkkazantsev/my-crypto-backtest-had-a-sharpe-of-134-then-i-added-trading-costs-5718"
  },
  "original_language": "en",
  "account": "This reporter investigated a cryptocurrency trading strategy that initially showed promising results but subsequently collapsed when accounting for trading costs. The backtest utilized a cross-sectional reversal baseline, which measured the returns of various assets over the previous eight bars, standardized those returns, and then assigned weights based on recent relative performance. The strategy was tested on cached Bybit USDT perpetual data, spanning 15-minute bars from January 2021 to September 2025. The evaluation window consisted of 90-day formation periods followed by 21-day out-of-sample periods, with 76 non-overlapping out-of-sample windows rolled every 21 days. The strategy employed causal volatility targeting at 10% annualized volatility, with a 96-bar trailing estimate and a 3× leverage cap. The backtest was conducted without any model fitting or feature search, and weights computed at time t became live at t+1. When the transaction costs were ignored, the strategy achieved a Sharpe ratio of 13.44, indicating strong performance. However, once trading costs were factored in, the Sharpe ratio plummeted to -40.13, and none of the 76 out-of-sample windows remained positive. This dramatic shift in performance highlights the significant impact that trading costs can have on investment strategies, even when the initial results appear favorable.",
  "summary": "I was not trying to prove a grand theory about crypto. I was testing simple, falsifiable baselines and looking for one result worth a second experiment. Cross-sectional reversal looked almost too clean: Sharpe 13.44 before transaction costs, with positive returns in 67 / 76 reported out-of-sample windows. Then I charged the same signal for the trading it required. At 10 bps per side, net Sharpe…",
  "key_points": [],
  "editors_take": null,
  "illustration": null,
  "coverage": {
    "outlets": 1,
    "also_reported_by": []
  },
  "ai_generated": true,
  "disclaimer": "Summaries, key points and the editor’s take are written by software from other outlets’ reporting and may contain errors — always check the linked original."
}