One Bot Is Easy. Five Bots Need a Plan.
Five bots can still be one bet if they all lean the same direction. A core-satellite-reserve framework for allocating capital, setting ceilings, and rebalancing on triggers instead of feelings.
A single bot's risk is bounded by its own settings — range, allocation, stop-loss. A portfolio of several bots carries a second layer of risk that no individual bot's settings can catch: how much of your total capital any one bot type can command, whether your "diversified" bots are actually five different bets on the same direction, and what happens when more than one of them draws down in the same week. This is a framework for that second layer.
TL;DR
- Treat total bot capital as one allocation, not five separate budgets — give each bot type a ceiling it can't exceed regardless of how well it's performing
- A core–satellite–experimental–reserve structure (for example, 50/30/10/10) keeps any single bot type from becoming an accidental concentration
- Reserve capital that sits uncommitted is a risk-management tool, not idle money
- "Diversified" bot types can still be correlated — several long-biased bots on BTC/ETH pairs move together in a broad drawdown
- Rebalance on triggers — a sleeve drifting past its ceiling, a bot type's underlying market condition shifting — not on a fixed calendar or a feeling
- Portfolio mode groups bots launched through the AI Assistant under one shared Take Profit — useful, but it's not the same thing as managing capital across bot types you've set up manually
Why five bots isn't five times the risk of one bot — it's a different kind of risk
A single bot's risk is mostly contained by its own configuration. Set a realistic range, a sane allocation, a stop-loss, and the worst case is bounded by those numbers. The 10-point risk checklist covers that layer well: liquidity, pair selection, allocation, range validity, exit rules.
A portfolio of several bots adds a layer none of those per-bot checks can see. No single bot's settings tell you what share of your total capital is tied up in GRID versus COMBO. No stop-loss on one DCA bot tells you what happens to your account if three bots draw down in the same week because they were all exposed to the same move. The risk stops living inside any one bot's configuration and starts living in the relationship between them.
That's a portfolio question, not a settings question. It needs its own framework.
Give every bot a job — and a ceiling
The simplest failure mode in a multi-bot setup isn't a bad bot. It's a good bot that quietly grows past the share of capital it was ever supposed to hold, because it kept performing and nobody set a ceiling.
One way to structure this: split total bot capital into sleeves by role, not by bot type.
| Sleeve | Role | Starting share |
|---|---|---|
| Core | Your most-tested bot type — the one with the longest track record in your own account | 50% |
| Satellite | One or two additional bot types for a different market condition than your core | 30% |
| Experimental | A new bot type or pair you're trying with real (small) money, kept separate from demo | 10% |
| Reserve | Uncommitted capital, not assigned to any bot | 10% |
The exact percentages aren't the point — they're a starting shape to adjust to your own capital and conviction. The discipline is the point: each sleeve has a ceiling, and a bot that outgrows its sleeve gets trimmed back to it, not left alone because it's working. That's the opposite of most traders' actual instinct, which is exactly why it needs to be a rule instead of a judgment call made in the moment.
Reserve capital is a tool, not a rounding error
Money that isn't in any bot looks like waste until the moment it isn't. Reserve capital does three specific jobs: it covers the fees and slippage that quietly eat into a bot's stated numbers, it funds a genuinely good setup that shows up mid-cycle without forcing you to shut something else down for it, and it's the buffer that keeps a bad week from turning into a forced decision.
A portfolio with zero reserve doesn't have more capital working for it — it has less room to respond to anything unplanned. Ten percent sitting uncommitted looks like a smaller number than most traders are comfortable with, right up until they need it.
"Diversified" isn't the same as "uncorrelated"
Several different bot types can still be several bets on the same outcome. A GRID bot buying BTC dips inside a range, a DCA bot accumulating ETH, a LOOP bot cycling SOL, and a COMBO bot long-biased on a futures pair are four different mechanics — and, in a broad market-wide decline, four positions that can lose money at the same time. The bot types are different. The directional bet underneath most of them often isn't.
Real diversification at the portfolio level means checking two things a per-bot settings review won't catch on its own: whether more than one bot shares the same underlying direction on correlated assets, and whether more than one bot is exposed to the same exchange or venue risk. A portfolio can look varied on the bot-type list and still be one concentrated bet once you look at what each bot actually does when the market moves against it.
See how your setup behaves together, not just each bot alone. Run a proposed multi-bot allocation in Bitsgap's demo mode against live prices before any of it touches real funds.
Rebalance on triggers, not on a feeling
A sleeve that's grown well past its starting share isn't a sign to leave it alone because it's working — it's the exact condition the ceiling exists for. Concrete triggers hold up better than a vague intention to "check in sometime":
- A sleeve drifts more than roughly 50% past its target share. If core was set at 50% and growth has pushed it to 70%, that's a trim, not a compliment.
- The condition a bot type depends on changes. A bot built for a range-bound market doesn't stay well-suited to one that starts trending, no matter how well it performed before the shift.
- A bot graduates out of "experimental." Once something in that sleeve has a real track record in your own account, it earns a permanent allocation instead of living in the temporary slot indefinitely.
- A fixed review date, even a boring one. Monthly is enough for most setups. The point of a fixed date isn't precision — it's making sure a review happens even when nothing dramatic prompted it.
What Portfolio mode does — and what it doesn't
The AI Assistant's Portfolio mode groups several AI-launched bots under one shared Take Profit and status, which is useful for managing a set of bots as a single unit instead of clicking through each one individually. It's worth knowing one specific tradeoff before turning it on: a shared Take Profit means every bot inside closes together once the target hits, which can mean exiting a still-performing bot early alongside weaker ones in the same group. The full setup walkthrough covers the mechanics in detail.
That's a different thing from the sleeve framework above. Portfolio mode groups bots launched together through the AI Assistant; it doesn't manage the balance between a manually launched GRID bot, a manually launched DCA bot, and whatever else is running across your account. If your bots weren't all launched as one AI-configured group, the ceiling-and-rebalance discipline above is still something you're doing yourself.
Common mistakes
Sizing each bot in isolation. A bot that looks reasonably sized on its own can still be the fourth bot pointed in the same direction. Size against total portfolio exposure, not against what looks comfortable for that one bot alone.
Letting a winner's allocation grow by default. A bot that's compounding gains is quietly increasing its own share of the portfolio even without new capital added to it. Without a ceiling, success in one bot slowly turns into concentration risk.
Treating reserve capital as unallocated waste. Moving reserve into a fifth bot because "it's just sitting there" removes the one part of the setup built to absorb a surprise.
Skipping the review because nothing went wrong. The setups that drift the furthest are usually the ones nobody checked on, precisely because they weren't causing visible problems yet.
Ready to put a real allocation plan behind your bots? Connect an exchange API key to Bitsgap and structure your setup using the sleeves above — free for 7 days, no credit card required.
FAQ
How much of my capital should go into one bot type? There's no universal number — it depends on your conviction and track record with that bot type. A common structure is a "core" bot type at roughly half of total bot capital, one or two "satellite" types for different conditions, a small experimental slice, and an uncommitted reserve.
Is running several different bot types automatically diversification? No. Bot types can still share the same underlying direction — several bots that are all long-biased on correlated assets can lose money together in a broad decline, even though they're technically different strategies.
How often should I rebalance a multi-bot portfolio? On triggers rather than a fixed schedule alone: when a sleeve drifts well past its target share, when the market condition a bot type depends on shifts, or at minimum on a fixed monthly check-in even if nothing looks urgent.
Does Portfolio mode manage this for me? Only for bots launched together through the AI Assistant, where it shares one Take Profit and status across the group. It doesn't balance capital across bot types launched manually and separately.
Why keep capital in reserve instead of putting all of it to work? Uncommitted capital covers fees and slippage, funds a genuinely good setup that appears mid-cycle, and is the buffer that keeps a bad week from forcing a decision you didn't plan to make.
