Ramping Risk On A New Strategy

Almost no desk puts a new strategy on at the size it's ultimately expected to run. Even a strategy that's been thoroughly backtested and looks strong on paper starts small — often a fraction of its intended capacity — and gets scaled up gradually as it proves itself trading real capital in live markets. This is called ramping, and the gap between a backtest and live trading is exactly why it exists: a backtest can't fully capture execution costs, how the strategy behaves during periods it wasn't tested on, or how its signals interact with the rest of the market once real capital starts moving.

What ramping is actually testing for

The main thing a ramp period checks isn't whether the strategy makes money — a strategy can look profitable over a short live window purely by chance, in either direction. It's checking whether the strategy behaves the way its designer said it would: does its risk profile match what was expected, does it correlate with other strategies on the desk the way it was supposed to, does its execution cost the amount the backtest assumed, and does the trader or system running it operate it the way it was designed to be operated. A strategy that makes money in its first month but takes on much more risk than expected to do it has actually failed the ramp, even though the P&L looks fine.

This is also why ramp schedules are usually set in advance rather than decided reactively — a fixed schedule (say, 25% of target size after the first month if nothing looks wrong, 50% after the second) prevents both of the obvious failure modes: ramping too fast because early returns look great, which is really just performance-chasing on a very short sample, and stalling the ramp indefinitely out of excess caution even after the strategy has demonstrated it behaves as expected. Both failures come from letting recent, noisy results drive a decision that should be driven by whether the strategy matches its design.

A concrete example: a new pairs-trading strategy is designed to run $20m at target size, with a plan to start at $5m and add $5m increments every six weeks if realized volatility and turnover match the backtest. If the strategy is up strongly in its first two weeks, that alone is not a reason to skip ahead to $20m — the schedule exists precisely so a lucky first fortnight doesn't get treated as proof the strategy is ready for full size.