Can you double your bot's returns just by doubling the risk per trade? Yes, actually. Almost exactly double, in my own numbers. Here's the part nobody wants to hear: your drawdown doubles too, at the same time, by roughly the same amount. There's no version of this where you get one without the other.
This is Part 6 of the build-in-public series documenting my Binance futures bot. Part 5 covered the ADX regime filter, the logic that keeps the bot flat during chop, and it ended with a promise to cover this exact topic next. So here it is. Risk sizing isn't a strategy decision. It's a dial. And the dial is linear, which is a less exciting sentence than it deserves to be given how many people get it wrong.
The dial, in my own numbers
My bot risks a fixed percentage of a per-symbol wallet slice on every trade, sized off a snapshot taken at the start of each cycle (that fix came out of the shared-margin bug in Part 4, a different lesson from the same project). I ran the corrected backtest (see Part 3 for why "corrected" matters) at two risk levels on the same strategy, same signals, same symbols, nothing changed except the size of each bet.
At 5% risk per slot, the backtest showed close to +17-18%/yr. At 7.5% risk per slot, roughly 1.5x the risk, the backtest showed close to +35%/yr, almost exactly 1.5x the return. Not approximately in the same ballpark. Almost proportional, to within noise.
Now the part that matters more. The drawdowns scaled the same way. Going from 5% to 7.5% risk didn't just lift the return line, it lifted the trough right along with it. Whatever multiplier you apply to your risk-per-trade, apply the same multiplier to your expected worst drawdown. That's not a rule of thumb. In my backtests it's close to exact.
I want to be honest about something here: I'm currently running the bot at 7.5%, the higher number. Not because I discovered some trick to break the relationship. I picked it because I did the drawdown math first and decided I could tolerate the downside, not because the upside number looked good. More on that below.
Why this is obvious in theory and still surprises people
Everyone nods along when you say "more risk, more reward." Fewer people actually sit with the fact that it's the same trade both directions, applied to every single bet the system makes, for the life of the account.
Position size is a multiplier on every outcome your strategy produces. If your edge generates a stream of wins and losses with some fixed distribution, and you scale every position by 1.5x, you scale every win by 1.5x and every loss by 1.5x. The wins compound into a bigger total return. The losses compound into a bigger total drawdown. There's no mechanism by which scaling touches one and not the other, because it's the same multiplication happening on both sides of every single trade.
A strategy's edge (win rate, payoff ratio, how good the signals actually are) determines the shape of the return curve. Risk-per-trade only determines how tall or short that same shape is drawn. You can't fix a bad edge by turning up the size. You can only make a bad edge lose faster.
Doubling risk per trade roughly doubles both the return and the drawdown. There is no sizing choice that gives you more of one without more of the other. Pick your drawdown tolerance first, then let the return follow from that, not the other way around.
Why 25-35% drawdowns are normal, not a red flag
Here's where most beginners misread their own equity curve. A 25% or 30% drawdown on a live account feels like something broke. Like the strategy stopped working, or a bug crept in, or the market shifted against you specifically. Sometimes that's true (three real bugs did hit this bot, none of them related to sizing logic itself). But at the risk levels I'm running, a 25-35% drawdown is just what the strategy does sometimes. It's baked into the math, not a symptom of anything going wrong.
The reason comes back to the win rate. This system wins roughly 30-40% of trades, and it's profitable because winners run about 2x the size of losers, not because losses are rare. A payoff ratio like that means streaks of 4, 5, sometimes 6 consecutive losses aren't a tail event. They're a normal, expected feature of a system with a sub-40% win rate. Part 1 of this series covered a 4-loss streak in the first week that felt catastrophic and was actually about 18% likely to happen just from variance on a system with these odds. Stack a couple of losing streaks into the same stretch, at 7.5% risk per slot, and a 25-35% drawdown is just arithmetic. It's not evidence the edge died.
Wikipedia's overview of risk of ruin covers the general version of this problem: repeated exposure to loss, even with a positive expectation overall, accumulates a real probability of severe account damage if bet sizes are too large relative to the edge. That's the formal version of what I'm describing informally with backtest numbers. Size too aggressively for your actual win rate and payoff ratio, and the "normal" drawdown stops being survivable.
Size for the bad case, not the good one
This is the actual lesson of Part 6, and it's the one I'd tell anyone building or buying into a bot before anything else. When you pick a risk-per-trade number, do it by asking "can I sit through the worst realistic drawdown at this size?" not "what return does this size produce?"
Most people do it backwards. They see a backtest showing +35%/yr at 7.5% risk, get excited about the return number, and only notice the corresponding drawdown after they've already lived through a chunk of it live. I nearly made that mistake myself. The return number is the one that sells. The drawdown number is the one that actually determines whether you stay in the seat long enough for the edge to play out.
Practically, that means working the sizing decision in this order:
Drawdown tolerance first: decide the maximum peak-to-trough decline you can genuinely sit through without panic-closing everything or abandoning the strategy. Not a number that sounds disciplined on paper, the number you'd actually survive emotionally and financially.
Backtest the drawdown at a few risk levels: run the same strategy at 2-3 different risk-per-trade values and look at the worst drawdown each one produced across multiple periods, not just the best-looking window.
Pick the risk level whose drawdown you already approved: the return you get is whatever comes attached to that size. You don't get to negotiate a bigger return at the same drawdown. It doesn't exist.
Discount even that number: live drawdowns tend to run worse than backtest drawdowns for the same reasons live returns run below backtest returns (funding drag, execution slippage, the odd rejected order). If a backtest says 30% drawdown at your chosen size, plan like it could be worse.
I picked 7.5% because I'd already stress-tested the drawdown side at that size across multiple periods and decided I could hold through it without doing anything stupid. If the drawdown number at 7.5% had scared me, the honest move would have been to drop back to 5% and accept the smaller return that comes with it, not to keep the size and hope the bad case doesn't happen.
What this means if you're not writing your own bot
If you're using a platform bot instead of building your own, this same dial exists, it's just hidden behind a slider or a "risk level" dropdown instead of a raw percentage in code. Whether that's Bitsgap's combo bot, 3Commas' DCA settings, or Bybit's built-in bots, somewhere in the settings there's a knob controlling how much of your capital each position risks. Turning that knob up will make the good months look better in the marketing screenshots. It will also make the bad months proportionally worse, and platform demos rarely show you the bad months.
If you're not sure which type of bot or risk profile actually fits your situation, the bot-match quiz is a five-minute way to get pointed at something that matches your actual risk tolerance instead of the return number that looked best on a landing page.
I still can't tell you whether 7.5% risk per slot is the "right" number in some universal sense. It's the number I decided I could survive the drawdown side of, on this strategy, with this win rate and payoff ratio. Your edge is different, your tolerance is different, and the honest answer is you have to do your own version of this math rather than copy mine.
Part 7 will get into the ongoing check-ins: live results as they come in, and a couple of dead ends I researched and dropped (funding-rate farming and pairs trading among them) that didn't survive contact with real data. Full series is on the journey page. Still live, still honest about the drawdowns along with the wins.
