@general-liquidity/sharpebench-mcp 0.20.0 → 0.22.0

This diff represents the content of publicly available package versions that have been released to one of the supported registries. The information contained in this diff is provided for informational purposes only and reflects changes between package versions as they appear in their respective public registries.
Files changed (3) hide show
  1. package/README.md +1 -1
  2. package/dist/server.js +1 -1
  3. package/package.json +2 -2
package/README.md CHANGED
@@ -13,7 +13,7 @@ An **MCP server** that exposes [SharpeBench](https://github.com/general-liquidit
13
13
  | `score_allocation` | Score a weight-vector trajectory (validity + turnover) |
14
14
  | `greeks` | One long European option: price, Greeks and local exposure flags |
15
15
  | `canary` | Derive a do-not-train contamination tripwire |
16
- | `is_my_sharpe_real` | Deflate one return series for its search footprint and render the honesty verdict (`trials_sr_std` annualized, `periods_per_year` default 252) |
16
+ | `is_my_sharpe_real` | Deflate one return series for its search footprint and render the honesty verdict (`trials_sr_std` and `sr_benchmark` annualized, `periods_per_year` default 252) |
17
17
  | `regime_compare` | Compare aligned returns inside caller-supplied regimes |
18
18
  | `percentile_selection` | Compare the observed point winner with a bootstrap-percentile winner |
19
19
  | `decompose_uncertainty` | Report aleatoric, epistemic, and distributional diagnostic legs |
package/dist/server.js CHANGED
@@ -44,7 +44,7 @@ export function createServer() {
44
44
  is_call: z.boolean(),
45
45
  }, async (params) => run(() => sb.greeks(params)));
46
46
  server.tool("canary", "Derive a deterministic do-not-train contamination tripwire token from seed material.", { seed: z.string() }, async ({ seed }) => run(() => sb.canary(seed)));
47
- server.tool("is_my_sharpe_real", "Answer 'is this Sharpe real, or an artifact of luck and multiple testing?' for a single return series. Deflates the observed Sharpe for n_trials (the search footprint), then returns a Pass/Borderline/Fail verdict with deflated Sharpe, PSR, haircut, MinTRL, and a plain-English explanation. n_trials = 1 is almost always a lie — pass the true number of strategies/configs you tried. trials_sr_std is annualized (default 0.5) and periods_per_year converts it to the per-period unit of the returns (default 252, daily bars; 365 daily crypto, 8760 hourly, 52 weekly). sr_benchmark is per period.", {
47
+ server.tool("is_my_sharpe_real", "Answer 'is this Sharpe real, or an artifact of luck and multiple testing?' for a single return series. Deflates the observed Sharpe for n_trials (the search footprint), then returns a Pass/Borderline/Fail verdict with deflated Sharpe, PSR, haircut, MinTRL, and a plain-English explanation. n_trials = 1 is almost always a lie — pass the true number of strategies/configs you tried. trials_sr_std is annualized (default 0.5) and periods_per_year converts it to the per-period unit of the returns (default 252, daily bars; 365 daily crypto, 8760 hourly, 52 weekly). sr_benchmark, the PSR and MinTRL benchmark, is annualized too (default 0.0) and converted the same way.", {
48
48
  returns: z.array(z.number()),
49
49
  n_trials: z.number().int().min(1).max(0xffffffff),
50
50
  trials_sr_std: z.number().optional(),
package/package.json CHANGED
@@ -1,6 +1,6 @@
1
1
  {
2
2
  "name": "@general-liquidity/sharpebench-mcp",
3
- "version": "0.20.0",
3
+ "version": "0.22.0",
4
4
  "description": "MCP server exposing SharpeBench's luck-robust quantitative evaluation tools for trading agents, including deflated Sharpe, pass^k, process discipline, and risk analysis.",
5
5
  "keywords": [
6
6
  "mcp",
@@ -30,7 +30,7 @@
30
30
  "prepublishOnly": "tsc"
31
31
  },
32
32
  "dependencies": {
33
- "@general-liquidity/sharpebench": "^0.20.0",
33
+ "@general-liquidity/sharpebench": "^0.22.0",
34
34
  "@modelcontextprotocol/sdk": "^1.0.0",
35
35
  "zod": "^4.0.0"
36
36
  },