Loading skill
Install any skill in seconds. Free to start, no credit card required.
Get Started Free →Zipline Reloaded — event-driven backtesting engine. Minute and daily data, custom factors, pipeline API, risk and performance analytics. Forked from Quantopian's Zipline for continued development.
.claude/skills/mkurman-zipline-reloaded/SKILL.md| Test case | Without → With | Effect | Δ tokens | Δ turns |
|---|---|---|---|---|
| case-08 | ✗→✓ | ▲ Improved | -16% | 0% |
| case-02 | ✓→✓ | ＝ Same ✓ | -9% | 0% |
| case-03 | ✓→✓ | ＝ Same ✓ | -28% | 0% |
| case-04 | ✓→✓ | ＝ Same ✓ | -39% | 0% |
| case-06 | ✓→✓ | ＝ Same ✓ | -30% | 0% |
Zipline Reloaded is an event-driven backtesting engine (forked from Quantopian). Supports minute and daily data, custom factors, pipeline API, and built-in risk/performance analytics.
bashuv pip install zipline-reloaded
pythonfrom zipline.api import order_target, symbol from zipline import run_algorithm def initialize(context): context.asset = symbol("AAPL") def handle_data(context, data): price = data.current(context.asset, "price") sma20 = data.history(context.asset, "price", 20, "1d").mean() sma50 = data.history(context.asset, "price", 50, "1d").mean() order_target(context.asset, 100 if sma20 > sma50 else 0) results = run_algorithm(start=pd.Timestamp("2022-01-01"), end=pd.Timestamp("2023-01-01"), initialize=initialize, handle_data=handle_data, capital_base=10000)
| Case | Status | Duration (ms) | Turns | Tokens | Tool calls | ||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| Without | With | Δ | Without | With | Δ | Without | With | Δ | Without | With | Δ | ||
case-01 | fail→fail | 13,999 | 9,649 | -31% | 1 | 1 | 0% | 2,613 | 2,329 | -11% | 0 | 0 | — |
case-02 | pass→pass | 3,855 | 1,852 | -52% | 1 | 1 | 0% | 622 | 565 | -9% | 0 | 0 | — |
case-03 | pass→pass | 7,095 | 3,194 | -55% | 1 | 1 | 0% | 1,162 | 834 | -28% | 0 | 0 | — |
case-04 | pass→pass | 8,544 | 3,561 | -58% | 1 | 1 | 0% | 1,442 | 884 | -39% | 0 | 0 | — |
case-05 | fail→fail | 4,384 | 3,402 | -22% | 1 | 1 | 0% | 714 | 811 | +14% | 0 | 0 | — |
case-06 | pass→pass | 8,609 | 4,130 | -52% | 1 | 1 | 0% | 1,469 | 1,030 | -30% | 0 | 0 | — |
case-07 | fail→fail | 7,082 | 4,960 | -30% | 1 | 1 | 0% | 1,164 | 1,192 | +2% | 0 | 0 | — |
case-08 | fail→pass | 6,604 | 3,121 | -53% | 1 | 1 | 0% | 1,023 | 857 | -16% | 0 | 0 | — |
case-09 | fail→fail | 12,626 | 4,399 | -65% | 1 | 1 | 0% | 2,131 | 1,050 | -51% | 0 | 0 | — |
case-10 | pass→pass | 9,411 | 2,007 | -79% | 1 | 1 | 0% | 1,676 | 653 | -61% | 0 | 0 | — |
case-11 | pass→pass | 13,924 | 5,313 | -62% | 1 | 1 | 0% | 2,478 | 1,265 | -49% | 0 | 0 | — |
case-12 | pass→pass | 2,819 | 2,314 | -18% | 1 | 1 | 0% | 384 | 653 | +70% | 0 | 0 | — |
case-13 | pass→pass | 5,608 | 4,004 | -29% | 1 | 1 | 0% | 1,041 | 1,126 | +8% | 0 | 0 | — |
case-14 | pass→pass | 6,828 | 3,570 | -48% | 1 | 1 | 0% | 1,146 | 873 | -24% | 0 | 0 | — |
case-15 | pass→pass | 8,481 | 2,584 | -70% | 1 | 1 | 0% | 1,257 | 666 | -47% | 0 | 0 | — |
case-16 | pass→pass | 7,000 | 5,316 | -24% | 1 | 1 | 0% | 1,124 | 1,170 | +4% | 0 | 0 | — |
case-17 | pass→pass | 3,964 | 1,574 | -60% | 1 | 1 | 0% | 585 | 542 | -7% | 0 | 0 | — |
case-18 | pass→pass | 10,656 | 3,637 | -66% | 1 | 1 | 0% | 1,914 | 1,001 | -48% | 0 | 0 | — |
case-19 | fail→fail | 10,203 | 3,958 | -61% | 1 | 1 | 0% | 1,983 | 1,054 | -47% | 0 | 0 | — |
case-20 | pass→pass | 11,541 | 9,322 | -19% | 1 | 1 | 0% | 2,124 | 2,162 | +2% | 0 | 0 | — |
case-21 | pass→pass | 15,807 | 10,715 | -32% | 1 | 1 | 0% | 2,800 | 2,445 | -13% | 0 | 0 | — |
case-22 | pass→pass | 16,135 | 11,991 | -26% | 1 | 1 | 0% | 2,761 | 2,496 | -10% | 0 | 0 | — |
DecimalAI ran this skill against gemini-3.6-flash twice over the same eval suite — once with the skill loaded and once without — and compared the two runs case by case. 22 cases were attempted. The headline lift of +5 percentage points is the difference between those two pass rates over the 22 comparable cases.
Without the skill loaded, the model failed this case. With it loaded, the same prompt on the same model passed. This is one improved case from the latest verified run; every case, including any that regressed, is in the table above.
Other measured skills in the registry, with their headline benchmark lift.