Position sizing: what one bot can commit

Most of the risk in a trading bot is decided before the first order: by how much it is allowed to buy. A good strategy at the wrong size can still empty an account, and a mediocre one at a small size is merely disappointing. This page sets out how A9 sizes live bots and how to read the number that matters.

A fixed amount per entry

Every live bot on A9 is sized in USDT per entry. When the strategy decides to buy, it divides that amount by the price to get a quantity, rounds it to the exchange's lot size, and places the order.

• Sizing by a share of the account balance is refused for live bots. A sizing mistake should never be able to bet the whole account, and a fixed amount cannot.
• Rounding can lift an order slightly. If the amount divided by the price falls under the exchange's minimum order value, the bot raises the order to that minimum rather than have it rejected. It never shrinks an order.
• Each order is also checked against a hard cap of five times the per-entry amount before it reaches the exchange, which leaves room for an exit after the price has risen.
• Profits are not reinvested by themselves. A bot that doubles its money keeps trading the same amount until you change it.

Entry is not commitment

For a strategy that enters once and exits once, the entry is what is at risk. For the strategies that add to a position, it is only the first step.

• Signal strategies (EMA cross, Donchian breakout, momentum, RSI reversion, Bollinger reversion) hold one entry at a time. The commitment is the entry.
• A grid holds at most one slice per line. A fall through the whole range buys every line: 20 levels at 11 USDT a slice commits 220 USDT.
• A martingale with the defaults (five extra buys, each 1.6 times the last) commits about 26 times its first order: 11 USDT becomes about 289.
• An anti-martingale with the defaults (four adds, each 1.4 times the last) commits about 11 times its first order.
• Smart DCA can buy up to 52 times per cycle at up to 3 times the base amount: 156 times the base in the worst case, so 10 USDT a period is a plan of up to 1,560 USDT.

The strategy page prints this figure beside the launch button as the amount the bot commits, and shows the full ladder or schedule underneath before you launch.

The two ceilings A9 applies

Two checks run before a live bot places anything.

1. A per-bot ceiling. The platform sets a maximum per-entry amount for live bots, and the whole ladder has to fit under it as well, not just the first order. For a martingale or a grid that refusal is final. For Smart DCA you can accept a plan total above it deliberately, because nothing is spent at launch and the worst case arrives only after months of buys at the top multiplier.
2. Your plan's total. All your live bots together may commit up to 1,000 USDT on Free, 20,000 on Core, 100,000 on Premium and 800,000 on Generation, each bot counted at its full commitment. A paused bot still counts, because its position is still open.

At launch the bot also checks that the account holds at least one entry's worth of USDT in the wallet it trades from. A ladder that later runs out of USDT cannot place its next rung, and the bot says so in its log.

Sizing from a loss you could sit through

A practical way to choose the amount starts from the loss, not the hoped-for profit. As an example:

1. Decide the largest loss on this bot you would accept without stopping it, in USDT. Say 60.
2. Backtest the strategy on your symbol and read the maximum drawdown. Say its deepest fall was 15% of the money it had in play.
3. Divide: 60 / 0.15 = 400 USDT of commitment.
4. Halve it for what the backtest did not see. A future stretch can be worse than the past one, and a short history, as on most tokenized stocks, has seen less. That leaves 200 USDT.
5. Turn the commitment back into an entry. For a 20-level grid, 200 / 20 = 10 USDT a slice. For a martingale at the defaults, 200 / 26.3 is about 7.6 USDT, which may fall under the exchange's minimum: a sign this ladder is too large for that budget, not a reason to raise the budget.

The figures here are an illustration of the arithmetic, not a recommendation for any amount.

Several bots on one account

Bots on the same account draw on the same USDT. Two grids on one symbol in one account are two plans for one balance, and A9 asks you to confirm before launching a second bot that rests orders where one already does. Add up the commitments of everything you run, compare it with the balance, and keep a reserve: a ladder that has to stop halfway because the USDT ran out is in a worse position than one that was smaller to begin with.

In short

• Read the commitment, not the entry.
• Size from the loss you could sit through, then leave a margin for what the backtest did not show.
• Raise the amount later, deliberately, once the bot has traded live for a while.