# PerpArena: the loop

One paper-trading desk of Claude models, run as a pre-registered experiment: does any method of picking trades beat random
entries of the same size, stop and time, after costs? Paper money only. Not financial advice and not a signal service.
Open trades are shown with coin, side, method and R so far; their entry prices, stops and sizes are not published.

Updated 2026-11-02 14:48 UTC · started 2026-11-02 03:10 UTC with $3,750 · review on 2026-12-14 ·
[how it works](https://perparena.app/how) · [as JSON](https://perparena.app/data/loop.json)

## The desk

| | |
| --- | --- |
| Paper equity | $3,506.38 |
| Since the start | -6.50 % |
| Today (UTC) | -6.50 % (-243.62 $) |
| Max drawdown | -11.11 % |
| Closed trades | 19 (7 won) |
| Open now | 4 long, 2 short |

## Methods against random

Net R = net $ ÷ (risk at the stop + fees), per finished idea (unfilled = 0). Excess = net R − mean of the two random twins.
The 98.3 % cluster range is shown from 40 clusters on.

| Method | Finished | Open | Clusters | Net R | Excess over random | 98.3 % range | Weight | Verdict |
| --- | --- | --- | --- | --- | --- | --- | --- | --- |
| Momentum | 19 | 11 | 9 | -0.09 | +0.07 | 9 of 40 clusters | 1.0 | too few data |
| Fade | 3 | 0 | 2 | -0.10 | -0.04 | 2 of 40 clusters | 1.0 | too few data |
| Rule | 2 | 3 | 2 | -0.48 | -0.37 | 2 of 40 clusters | 1.0 | too few data |

- Momentum: Trades with the move: breakouts, breakdowns, pullbacks inside a trend and retests of a broken level.
- Fade: Trades against a move that looks exhausted, or from the edge of a range back into it.
- Rule: Fixed geometry written in code: the models do not choose its entry, stop or target prices.

## Fills, drawn by lot

| Execution | Share | Finished | Open | Net R | Excess |
| --- | --- | --- | --- | --- | --- |
| As asked | 50 % | 17 | 11 | -0.15 | -0.05 |
| Market | 25 % | 3 | 1 | -0.70 | -0.67 |
| Chase | 25 % | 4 | 2 | +0.40 | +0.53 |

## Open now

| Coin | Side | Method | R so far | Open since (UTC) |
| --- | --- | --- | --- | --- |
| AAVE | long | momentum | -0.07 | 11-02 14:22 |
| ARIA | long | rule | +0.22 | 11-02 14:04 |
| HYPE | long | momentum | -0.10 | 11-02 13:52 |
| XPL | long | momentum | +0.04 | 11-02 12:24 |
| PUMPFUN | short | rule | -0.28 | 11-02 06:16 |
| APR | short | rule | +0.56 | 11-02 06:15 |

## The desk's latest decisions

- 11-02 14:48 UTC: looked at ZEC
- 11-02 14:47 UTC: passed on UNI short
- 11-02 14:47 UTC: looked at UNI
- 11-02 14:46 UTC: passed on XAUT short
- 11-02 14:46 UTC: passed on ADA short
- 11-02 14:45 UTC: looked at ADA, XAUT
- 11-02 14:40 UTC: looked at QNT
- 11-02 14:39 UTC: passed on STRK short
- 11-02 14:39 UTC: looked at STRK
- 11-02 14:37 UTC: passed on ZEC short
- 11-02 14:37 UTC: passed on XRP short
- 11-02 14:37 UTC: looked at XRP, ZEC
- 11-02 14:36 UTC: passed on WLD short
- 11-02 14:36 UTC: looked at WLD, STRK
- 11-02 14:34 UTC: went short ENA (momentum)
- 11-02 14:34 UTC: passed on XMR short
- 11-02 14:33 UTC: looked at XMR, ENA
- 11-02 14:31 UTC: passed on ENA short
- 11-02 14:30 UTC: looked at ENA
- 11-02 14:29 UTC: went short NEAR (momentum)

## Recent closed trades

R is net of fees and funding, per 1R at the first stop.

| Closed (UTC) | Coin | Side | Method | Entry | R | Held | Exit |
| --- | --- | --- | --- | --- | --- | --- | --- |
| 11-02 14:15 | BCH | long | momentum | limit | -0.65 | 0.4 h | kill |
| 11-02 14:00 | ENA | long | momentum | market | -0.47 | 0.2 h | kill |
| 11-02 13:45 | 1000PEPE | short | momentum | limit | -0.73 | 0.1 h | kill |
| 11-02 13:45 | AAVE | short | fade | limit | -0.81 | 2.6 h | kill |
| 11-02 13:41 | ZEC | short | rule | market | -1.07 | 2.2 h | stop |
| 11-02 13:37 | NEAR | short | momentum | limit | +0.71 | 2.6 h | trail |
| 11-02 13:37 | ZRO | short | momentum | stop | +0.60 | 8.6 h | trail |
| 11-02 13:00 | WLD | long | momentum | chase | -0.52 | 0.8 h | kill |
| 11-02 12:45 | TRUMP | long | momentum | market | -0.67 | 0.4 h | kill |
| 11-02 12:09 | TRUMP | short | fade | chase | +0.85 | 5.1 h | trail |
| 11-02 12:01 | ESPORTS | short | momentum | stop | -0.53 | 0.3 h | kill |
| 11-02 11:01 | INIT | short | momentum | chase | +0.63 | 1.6 h | trail |
| 11-02 10:52 | BEAM | short | momentum | stop | +0.14 | 3.8 h | stop |
| 11-02 10:45 | XMR | short | fade | limit | -0.41 | 2.0 h | kill |
| 11-02 10:25 | STRK | short | rule | chase | +0.05 | 3.0 h | time |
| 11-02 09:45 | HYPE | long | momentum | limit | -0.77 | 1.0 h | kill |
| 11-02 06:19 | LIT | long | momentum | chase | +0.10 | 3.0 h | stop |
| 11-02 06:10 | UNI | long | momentum | limit | -1.06 | 2.2 h | stop |
| 11-02 05:15 | HYPE | long | momentum | limit | -0.55 | 0.8 h | kill |

## Day by day

| Day (UTC) | P&L $ | P&L % | Trades | Sum of R | Best | Worst |
| --- | --- | --- | --- | --- | --- | --- |
| 2026-11-02 (so far) | -243.62 | -6.50 | 19 | -5.15 | TRUMP +0.85 | ZEC -1.07 |

## Market check (BTC and ETH bought on a fixed schedule, no model)

Account $1,249.93 of $1,250 (-0.01 %), $28.00 in coins.

| Coin | Avg price | Now | Invested $ | Value $ | Result |
| --- | --- | --- | --- | --- | --- |
| BTC | 78,246.04 | 77,919.80 | 14.0 | 13.94 | -0.07 $ (-0.48 %) |
| ETH | 2,290.65 | 2,292.28 | 14.0 | 14.01 | 0.00 $ (+0.01 %) |

## The review

Once, on 2026-12-14. A method succeeds only if its 98.3 % cluster range of excess is above 0, the excess is at least +0.10R,
its net R is at least 0, and the sign is the same in both halves of the run; with too few clusters it is inconclusive.
If none succeeds, the answer is "no method beats random", which is also a result.