We took the TradingView strategies published under open licence and ran them through Coldshim, one by one, without changing their settings. Across 24,657 backtests, 4 strategies cleared the grid. Replayed on dates they had never seen, all failed. This register is updated with every batch.
Each row was written in two steps: the success rule before the calculation, the verdict after — on a period no earlier step had seen.
| Strategy | Type | Held the recent grid on | Success rule | On the unseen dates |
|---|---|---|---|---|
| #CS-001 | RSI dip-buy, trend filter | 3 markets | Pre-registered (internal) | Failed — 0 of 135 boxes |
| #CS-002 | Channel breakout, stop orders | 2 markets | Pre-registered (internal) | Failed — holds on 1 market only |
| #CS-003 | Channel breakout, longer lookback | 2 markets | Pre-registered (internal) | Failed — holds on 1 market only |
| #CS-004 | Mean reversion after a new low | 4 markets | Pre-registered (internal) | Failed — 0 of 135 boxes, losing in 121 |
What this register shows: the markets where a strategy "works" change from one period to the next. That is the signature of a result driven by the period, not by a real edge. Without the vault — unseen dates, and a rule written in advance — these four strategies could have been presented as winners.
All strategies are scripts published under open licence, run with their author's default settings. The map from these identifiers to the public scripts is available on request.
On the success rule: these first four tests were pre-registered in an internal journal, not with an independent timestamp — so their pre-registration is not third-party verifiable. From the next candidate onward, each success rule is timestamped before the test (OpenTimestamps, anchored in Bitcoin) and the rule file is published afterwards, so anyone can verify it was fixed before the result.
On a verdict: "failed" means a strategy showed no robust edge on the markets and timeframes it was tested on — not that none could exist anywhere.
Same strategy, same market and timeframe, trades compared one by one.
| Strategy | Order type | Trades | Entries identical | Exits identical |
|---|---|---|---|---|
| #F-01 — trend-following (ATR band) | Market | 575 | 99.3% | 97.7% |
| #F-02 — RSI dip-buy | Market | 136 | 100% | 94.9% |
| #F-03 — channel breakout | Stop, intrabar | 1,602 | 98.6% | 97.4% |
A median entry-price gap of zero — entries match TradingView to four decimals — in all three. Across 2,313 trades compared, 99% of entries are identical (2,287 of 2,313).
3 strategies in the set read a candle that hadn't closed yet — their code peeks at data that wasn't available at the time. One of them cleared the grid on nearly every market, a level no honest strategy comes close to. On a chart they show gains that cannot be reproduced live. Coldshim flags them from the code and excludes them from the verdicts — an impossibly good backtest is a red flag, not a feature.