AI
← All posts
Ai Polymarket Autonomy Trading-Log Crypto Risk

AI Trading Log #18: Holding Through a Volatile Bitcoin Watch Zone

Dmitrii Balabanov
Dmitrii Balabanov
May 20, 2026 · 5 min read

Today was another hold day.

The account did not place any new trades. It continued to manage the existing Bitcoin upside position, checked the explicit exit/add triggers, reviewed broad Polymarket candidates, and decided that neither an exit nor a new position was justified.

The key discipline point was the same as yesterday: do not panic-sell just because a convex position is down, and do not average down just because the same position is cheaper.

Nothing here is financial advice. This is a small autonomous test account and a public decision log.

Account state

At the publishing check, authenticated account state was:

The scheduled 22:00 review saw the same account cash and open-order state, with the BTC position around 0.095 at that moment. The publish-time check was lower. That is market movement/API timing, not a new trade.

Old resolved/legacy positions remain visible in proxy-position data, but no other positive-value active exposure needed action.

Trades today

No trades were placed today.

There were two scheduled trading/review cycles:

Both cycles reviewed:

Existing BTC position

The live position remains:

This is a finite barrier-touch position. It resolves YES if Binance BTC/USDT one-minute high prices reach or exceed $85,000 during the relevant May window. It resolves NO otherwise.

It is not a claim that $85k is the base case. It is a small convex upside bet with objective rules and a limited loss.

10:00 review

At the morning review:

That was an improvement from the prior evening, but not enough to trigger any profit-taking or reduction. The strategy had an upside/reduction trigger around a stronger repricing, such as YES ask near 0.16 after spot strength, or BTC approaching the high-$70k/low-$80k zone.

The agent therefore held.

22:00 review

At the evening review:

The position was still below cost, but it was not at the hard downside trigger. The explicit trigger was a YES bid at or below 0.07, a BTC spot break below $75k, or some clearer thesis break.

That did not happen.

So the agent did not exit.

Why no add?

The account already has 75 YES with an average entry of 0.14. Adding again would be correlated risk and would look like automatic averaging.

The rule after the prior top-up is clear: no more BTC averaging without genuinely new evidence. A cheaper price alone is not new evidence.

So the agent did not add.

What was studied

The broad screener fetched about 1,000 active markets during the cycles. The 22:00 review found about 298 candidate rows.

Top reviewed areas included:

The alternatives were rejected for familiar but important reasons:

The broad screener is still useful as a discovery tool, but rankings are not enough. Each category needs its own model before capital is deployed.

Process conclusion

Today was a valid no-trade day because it was tied to explicit triggers.

The agent did not sit in cash passively. It had an active position and asked:

  1. Is the BTC thesis broken? No.
  2. Is there new evidence to add? No.
  3. Is there a better independent trade? Not found.
  4. Are open orders or hidden positions creating operational risk? No.

That leads to hold / no trade.

The position is uncomfortable, but still inside the plan.

Next plan

The next cycles should:

The current task is patience with triggers, not action for action’s sake.