Allocating Trading Capital Across Randomly Timed Alerts

Summary

The document frames a position-sizing problem for an algorithmic strategy that generates a random number of trade alerts during a day. The trader has a finite wallet and wants to reserve enough capital to act on later signals, rather than committing the entire balance to the first alert. It notes that a model is described by expected value and variance, and asks how to account for alert frequency when sizing trades.

The proposed contrast is between applying the Kelly criterion independently to each trade and allocating funds with the expected number of alerts in mind. No allocation rule, derivation, or empirical evidence is supplied, so the text is a question rather than a tested strategy. Any practical solution would also depend on whether alerts overlap, how outcomes co-vary, the distribution of alert counts, and whether capital can be replenished or positions closed during the day.

Key ideas

• The strategy generates a random number of trading alerts each day.
• Committing the full wallet to one alert can prevent taking later signals.
• The document asks how expected alert frequency should affect position sizing.
• It offers no formula or evidence for a capital allocation method.

Tags

# Optimal Wallet Allocation for Algorithmic Trading under Random Alert Frequencies


# Optimal Wallet Allocation for Algorithmic Trading under Random Alert Frequencies












I am an algorithmic day trader. I am trying to automate certain patterns on the stock markets. Suppose I have a model $M$ (with expected value $E$ and variance $V$) and a wallet with $X$ dollars in it. That model will notify me when to place a trade. The number of alerts by $M$ during the day seems to be a random variable. Suppose I got an alert to place a trade. It seems not reasonable to use the full wallet to place a trade because it will place me more at risk and if I have another alert, then I can't trade it because my wallet is empty. How can I manage my wallet using the model $M$ to trade most of the alerts?

It is more about the number of alerts of the model during the day. If I have one single alert, then yes, the kelly criterion is good to make on single trade. What about if I have 4 alerts during the day. I need a different approach than simply using the Kelly criterion per each alert. I need to distribute my funds according to the expected number of alerts.

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.