Skip to content
All library documents

Sizing Capital and Managing Drawdowns Across a Strategy’s Lifecycle

Article Quant Q&A · Author: Jabran Zahid

Summary

The discussion considers how to allocate capital to an automated crypto strategy whose backtest suggests substantial drawdowns, and when to scale its exposure or stop trading as its edge may fade. The response emphasizes that changing position size is a material change to the strategy: the revised sizing rules should be backtested and forward-tested before live use. It also suggests setting initial exposure so account-level drawdowns fit investor tolerance, while retaining unused capital to withstand losses.

The example relates a strategy’s drawdown to a smaller account-level drawdown by allocating only part of the capital to it. The answer cautions that investors assess account-level performance and argues against cutting size during a drawdown. However, it offers no general rule for detecting alpha decay or deciding when to terminate a strategy, and its sizing example does not establish that drawdowns scale linearly in live trading. Its guidance is practitioner opinion, not a universal lifecycle framework.

Key ideas

  • Position-sizing changes alter a strategy and require fresh backtesting and forward testing.
  • Initial capital allocation can be set to keep account-level drawdowns within investor tolerance.
  • Unused capital can provide reserves during a strategy drawdown.
  • The response discourages reducing exposure during a drawdown but provides no evidence or universal rule for that view.
  • The discussion does not define a systematic test for alpha decay or strategy termination.

Tags

Full text
# Scaling in and out of a strategy


# Scaling in and out of a strategy












I have developed a fully automated crypto trading strategy for which has been showing promising results and I am now looking to raise money to expand and hopefully trade to its full capacity in the future. My analysis on the backtest results shows there is a small possibility of of 40-50% drawdown (probably once a year or so) if use the biggest position size my backtest allows but if I reduce my position sizes to half, then these DDs will drop to half as well. Now, the question I have is, lets say I have 100k, is there a best practice to scale up (and down) to the maximum amount of money we are going to invest considering the reality that all strategies eventually lose their alpha. Lets say we do hit a big drawdown, at what point we decide to pull the plug. I know we can aggressively reduce the position sizes as we lose but still I think there will be a big drawdown and we will have to decide whether to pull the plug or not. So in nutshell, how does a trader scale in and out of a strategy during the 'whole lifecycle' of a strategy which has worked fine for a while and eventually lost it edge ?

Thanks

## Answer by amdopt (score 1)

https://quant.stackexchange.com/a/73269

Drawing 40-50% once per year will make it very tough to raise any money. That's way too often for most investors unless your strategy makes more than 100% per annum. Even then, it will be too much volatility for most to handle.

If you desire to scale your strategy into or out of positions, you should adjust your algorithms accordingly, redo all your backtesting, redo all your forward tests, and transition to a live trading account once again. What you propose is a fundamental change to your strategy that you haven't tested.

Alternatively, as you suggest in your post, you could start by exposing your strategy to an amount of capital that results in a smaller drawdown for the account. Potential investors will want to see account-level returns, not necessarily strategy data (though some will ask), so keep the size appropriate.

For example, if I know potential investors won't tolerate more than a 10% drawdown and my strategy draws 50% annually, I will only expose 20% of my capital to that strategy. That will keep plenty of cash on hand to weather a drawdown and not reduce size.

FWIW, I think reducing size during a draw is never what you want to be doing, so I would throw the idea out.

Shown in full with attribution under the source's licence. Licence: CC BY-SA 4.0 (Stack Exchange)

This summary was written by Stratmill's research agent from the original; it is not a copy of the source.