> ## Documentation Index
> Fetch the complete documentation index at: https://shumi.mintlify.site/llms.txt
> Use this file to discover all available pages before exploring further.

# Shumi evidence: studies, backtests and decision record

> Shumi's published studies, dated readings, the walk-forward record and the decision record.

## Published studies

| Study | Published | Tested |
| - | - | - |
| [Spring 2026 Rally: 23 Trades](https://shumi.ai/studies/regime-engine-april-may-rally) | May 13, 2026 | Every closed trade from one live window in spring 2026, on Hyperliquid prices |
| [400 Experiments. 3 Survived.](https://shumi.ai/studies/400-experiments) | Mar 25, 2026 | Hundreds of strategy variants, 2021 to 2025, before and after real Binance funding costs |
| [Walk-Forward Performance, 2024-2025](https://shumi.ai/studies/walk-forward-performance-2024-2025) | Mar 10, 2026 | Trained on 2021 to 2023, tested on 2024 to 2025, Binance funding charged per entry |
| [Crypto AI Benchmark v2](https://shumi.ai/studies/crypto-ai-benchmark-v2) | Mar 11, 2026 | Several AI tools given the same trader questions, scored on a weighted rubric |

The 400-experiments and walk-forward studies used an older trend engine, since replaced by CoinRotator's current trend rules.

## Latest readings

Dated findings from [shumi.ai/studies](https://shumi.ai/studies):

* **Oct 17, 2026 · Market washout.** When 20% or more of Binance coins fall twice their normal 10-day move, the average coin rose 6.0% over the next 10 days, against 1.0% on any day. BTC rose 3.7%. That happened 18 times from 2021 to 2026. Risk: since July 2024 the gain averaged 3.0%, and the worst episode, June 2024, lost 6.6%.
* **Oct 11, 2026 · Options.** When BTC or ETH 7-day implied volatility is in the top 10% of its last 90 days, selling the 7-day at-the-money straddle and holding it to expiry made 2.12% of the coin's price per trade after costs, 2020 to 2026. Risk: the worst 1 in 20 trades lost 11% (BTC) to 16% (ETH) on average, and the worst single trade lost 28%.
* **Oct 7, 2026 · New coins.** On 196 Base launchpad coins from January to June 2026, the coins still trading a month later had more holders at 72 hours than the coins that stopped (median 636 against 431). Their top 10 wallets held less of the supply (median 35% against 41%). Risk: those coins still fell about 55% from day 3.
* **Sep 26, 2026 · Momentum.** Coins that rose most over the last 20 days kept beating the coins that fell most. Across the top 200 coins since December 2017, long the top fifth and short the bottom fifth made 1.00% a week after fees. Long only, the top fifth beat the top-200 average by about 0.35% a week. With BTC well above its 200-day average, the long-short made 2.19% a week. Risk: near or below it, it made about nothing. The 13 weeks to September 13, 2026 lost 0.56% a week.

## Decision record

Once a day Shumi gives one take on one coin, published unedited in the [decision record](https://shumi.mintlify.site/evidence/decision-record.md) and scored by rules fixed in advance in the [pre-registration](https://shumi.mintlify.site/evidence/pre-registration.md).

## Walk-forward record

`shumi walkforward outcomes` lists the tracked walk-forward outcomes: each entry, each exit and the reason it closed. `shumi walkforward signals` lists the entry and exit signals themselves (50 per call).

> [tab] Terminal
>
> ```bash theme={"dark"}
> shumi walkforward outcomes
> ```

> [tab] API
>
> ```bash theme={"dark"}
> curl "https://api.shumi.ai/api/cli/walkforward?action=outcomes" \
>   -H "Authorization: Bearer shumi_sk_your_key_here"
> ```
>
> `action` takes `signals`, `positions` or `outcomes`.

Each outcome has these fields:

| Field | Meaning |
| - | - |
| `symbol` | The coin |
| `direction` | `long` or `short` |
| `entryPrice`, `exitPrice` | Price at entry and at exit |
| `entryDate`, `exitDate` | When it opened and closed |
| `returnPct` | Return in percent |
| `holdDays` | Days held |
| `exitReason` | Why it closed |

A `summary` block counts the total, wins and losses.

## Signal quality

```bash theme={"dark"}
shumi signal-quality --asset BTC
```

Over the API, `GET /api/cli/signal-quality?asset=BTC`. From an AI agent, the `get_signal_quality` tool. `--signal-type` defaults to `mean_reversion`.

`reliability_tier` says how that signal held up when tested. `HIGH` or `MEDIUM` means it passed. `LOW` means it failed a check, drifted, or fell short of the bar. `INSUFFICIENT_DATA` means there is no record.

## Recheck the results

* **Outcomes.** Every closed outcome in a strategy study is listed with its dates and prices. Pull the same dates from any exchange's price history and recompute the return.
* **The walk-forward record.** `shumi walkforward outcomes` carries the same fields. Recompute any return from its entry and exit prices.
* **Benchmark scores.** Every tool's score on every question and rubric dimension is in [`data.js`](https://github.com/pxeodev/studies/blob/main/benchmark/data.js) in the studies repository.


This documentation is built and hosted on [Mintlify](https://mintlify.com), a developer documentation platform.