BUSINESS EXPERIMENTS · 6 JUNE–8 JULY 2026 · FAILED AND STOPPED
The bot made no profit. My retained trading record shows a small loss, and on 8 July 2026 I stopped the service and removed its authority to buy. No outside party validated a trading edge. The outcome recorded here is a shutdown after loss, not an automated-trading success.
The experiment began with a deliberately modest question. Instead of reacting to another price spike I had missed, I wanted to test whether a short-horizon rule could clear its costs. That required a bot that could be retired when the answer was no.
33 days
6 June–8 July 2026
Review date fixed
decision after 28 days
External validation: 0
no verified trading edge
Final state
service stopped, authority removed
Strategy changes were outrunning the evidence
I began with a short scalping approach, then moved through major-asset trend rules, accumulation and a basket of smaller assets. The final stage ran only as a shadow experiment. Each weak result made another rule look like the answer, leaving an endless route to “one more version.”
I interrupted that cycle by writing down the review date, minimum sample and after-cost performance threshold in advance. Parameters were meant to remain untouched until day 28. Alongside the chosen rules, I logged a random control exposed to the same market.
Risk controls and trading performance were separate tests
The system isolated its capital and included daily and weekly loss limits, a capital floor and a kill switch. I added persistent records around orders and a watchdog for stopped processes. Pre-deployment review led to confirmed defects being fixed, and another review exposed a collision between one signal and an exchange alert condition.
The weekly kill switch fired on 15 June. I did not resume buying. The bot remained in observation mode while the shadow record ran to the precommitted date. The controls limiting a loss and the strategy failing to make money were both true.
The random control pointed the same way
In the retained internal record, the selected rules did not clear the after-cost threshold. The random control also sat on the losing side. This does not establish that every form of short-horizon crypto trading is impossible. It says that, in this period and trading universe, another adjustment to the same family of rules had no demonstrated advantage.
The performance figures and trades have not been independently reproduced, and the account record remains private. The public conclusion is deliberately narrower: the experiment did not make money, the internal record shows a loss, and I closed it under the rules written beforehand.
A working kill switch does not turn a failed strategy into a success. It can stop that failure from becoming larger.
A shutdown had to reach the permissions
On 8 July, I recorded the failed decision and disabled the service. I also removed the bot’s buy authority rather than relying on a dry-run setting alone. The shadow and alert logs, order audit, operating charter and termination record were retained.
A future hypothesis would begin offline. It would need its own shadow registration, an after-cost control and a fixed retirement date before any live authority returned. This bot is not a dormant success waiting to be restarted. It is a record of why an operational stop must include revoking the ability to act.
Project record
- Period: 6 June–8 July 2026
- Activity: implementation of several trading rules, shadow logging and a random control, plus a fixed review date, loss limits and permission revocation
- Internal evidence: bot source, shadow, alert and order logs, operating charter, precommitment and termination record
- External validation: none; performance and trades are user-held records and have not been independently reproduced
- Final outcome: zero profit, a small loss in the internal record, service stopped and buy authority removed on 8 July 2026
- Withheld: exchange, account and keys; balances, assets, orders, trades and detailed strategy performance
This is not investment advice. It is a record of one personal experiment and its shutdown.