Research and Development of Financial AI and AI Agents / Weekly review for the week of September 27 to October 3, 2026

After rushing from 0.02 to 0.10, design the stop-and-start conditions for lot expansion over the weekend

Right after funds increase, you want to increase trading volume. Recently, a user reported: “After the funds doubled, I increased from 0.02 lot to 0.04 lot, and then to 0.10 lot, and I made a huge mistake.” This is a valuable starting point from their experience, but it cannot be determined solely from the articles’ source documents which ticket had what loss, what the account balance was at which point, and what the planned stop-loss width was for each order. Therefore, we will not attach fictitious losses and will organize a mechanism to pause before increasing the lot size.

We are researching and developing financial AI and AI agents, using our own GOLD13 indicator and automated trading software (EA), and conducting live account verifications. We continue to provide 24/7 weekday trading live streams, but a temporary apparent profit on screen does not mean that the loss a trader can tolerate in the next trade has increased. The weekend theme is not “when can we win big?” but “will we break the risk ceiling right after a win?”

Lot size cannot be determined by balance alone

If you increase from 0.02 lot to 0.04 lot, with the same instrument, contract specification, and stop-loss distance, the profit/loss swing with price movement will be roughly double. 0.10 lot is five times 0.02 lot. However, merely knowing the balance doubled cannot explain whether the next trade’s stop-loss distance remains the same, whether the spread remains, or how many positions are held simultaneously. Even if you aim to keep the ratio to balance constant, if the stop-loss moves farther away, the expected loss per trade expands. A dashboard that merely lists numbers for lots will inevitably pair “how much I would lose if the price moves to this point.”

The skeleton of the calculation is: planned loss amount = number of lots × monetary value per price movement according to contract specifications × distance from entry to stop loss + estimated costs + buffer for slippage. The necessary contract size, profit at each price width, account currency, conversion rate, and margin requirements differ by broker and instrument. You cannot answer uniformly “0.04 lot if the balance is this amount” without confirming these. This time’s conclusion aligns with the reflection that you should not determine required balance just from the fact that it is GOLD.

Items to fill before promotion

What to enter first in the dashboard is not the current balance but the planned SL for the next order, the acceptable loss per trade, and the daily stop amount. Next, record the combined risk when multiple positions move in the same direction, the leeway during spread widening, and the possibility that another EA places orders while you hold. If these fields are empty, promotion to 0.04 is on hold. In particular, EAs may receive continuous signals, and a single discretionary decision may not conclude the process. The upper limit for individual orders does not determine the overall account limit.

The weekly report shows October 1: 0.02 lot with net +3293 across 14 orders, 0.04 lot with net +7736 across 15 orders, and October 2: 0.02 lot with net -4603 across 13 orders. From this ordering, we cannot claim “0.04 is correct, 0.02 is a failure.” The dates, market, counts, stop-loss widths, and currency units are not aligned or confirmed. Adopting the high net lot from one day promptly the next day could be tailoring rules to past results. The criteria for changing lots must be fixed before the win, not after.

Introduce a mechanical waiting period right after a win

If you set a rule not to raise the lot limit on the day of a large profit, you can reduce emotional influence. On the next day, verify whether the profit has been settled, whether it is net including commissions and swaps, and whether it matches the broker's statement. Then review several days of the same logic and same stop-loss design to observe the amplitude of losses. There is no magic number for the waiting period itself. What matters is to separate the timing for applying for promotion from the timing of auditing past results and to keep a record of the reasons for changes.

The EA side does not include a shortcut such as “profits rise and automatically double the lot.” Research on AI agents observing the market and selecting candidates is progressing, but designing the AI to raise safety limits such as maximum loss, maximum lot, number of open positions, or stop during communication failures should be avoided. The current method is that the EA or indicator proposes candidates, and AI approves them before the EA autonomously places trades. A different approach where AI leads from candidate discovery to position management is still in the research stage and not yet implemented in production. The write-up separates research ideas from current operational behavior.

Set stop conditions for failure first

If an unforeseen loss occurs after expanding the lot, simply returning to the original lot is not enough. Check whether the planned SL order was placed, the slippage of executions, whether the EA stacked multiple orders, and whether manual operations conflicted. Place broker statements, EA logs, AI approval records, tickets, and signal IDs in a single timeline. Until the cause is found, halt automatic lot promotions under the same conditions and avoid rushing recovery judgments for safety.

Stop conditions are not only “stop emotionally after a certain amount of loss.” They also require stopping in cases of orders without planned SL, failed readback of account balance, missing price updates, unclear EA version, mismatch between VPS and local settings, or concurrent holdings exceeding the intended limit. Even if P/L numbers look good, if controls are broken, you cannot correctly evaluate the next trade. For financial AI, the role is to detect losses, inconsistencies, and over-limit conditions and alert humans before profitability maximizes.

Restart conditions are stricter than promotion conditions

When restarting 0.02 lot operations after a failure, it is not enough to simply say “we set it back.” Read the actually loaded EA version, maximum lot settings, AutoTrading status, current positions, and order history to confirm alignment with the intended state. Files saved locally and those loaded on the production VPS are not guaranteed to be identical. Even if source code or EX4 update times are newer, that does not prove production operation. First, confirm practical settings by reading data, then trace whether the planned stop/settlement occurred correctly through automatic orders.

Whether a 0.04 lot promotion is considered after that depends on calculating a one-time risk from the planned SL, the maximum risk including concurrent holdings, the balance after a losing streak, and the downturn including costs and slippage, ensuring it stays within the cap. Even within the cap, if there is no material evidence that the past 0.02 lot records would yield the same decision, you cannot promote on mere expectations. The 0.10 lot is five times 0.02, and given past major failures, you should not promote to the next candidate based on appearances alone.

What we can promise readers, and what we cannot

An honest answer to “What balance would allow a move to 0.04?” cannot be determined with the current materials. This is not evasive; it is dangerous to quote a dollar amount without knowing contract size and stop-loss distance. What we can present are the necessary input items and evaluation formulas, and a practice of not promoting until verification is possible. The calculation skeleton described in the article is not investment advice tailored to each reader, but an auditing method for the research process.

We will shift this week’s reflections from mental discipline to recording and control. In the research and development of our own GOLD13 indicator, EA, and AI agents, we will verify stopping conditions before increasing orders rather than the conditions that trigger more orders. In the weekday 24-hour live trading, we will report how the balance, acceptable loss, and actual risk have changed, not just the increased balance. Details of the tools areThe sale page within GogoJunglefor FX and GOLD trading involves the risk of principal loss and does not guarantee profits or loss avoidance.