What are indicators of alpha decay of any particular algo.
Hi @nijinwilson
I would like to clarify that, in general, users often associate a period of underperformance with alpha decay. However, in practice, a drawdown phase does not necessarily indicate that an Algo’s alpha has decayed.
At Stratzy, we have multiple checks in place to help preserve the integrity of our Algos. One of these is controlling the capital capacity of each Algo. Once an Algo reaches its intended capacity, we restrict further allocations to help ensure that execution quality and Algo behaviour remain consistent. Based on our testing and live deployment, the current capacities of our Algos are well within acceptable limits, so users do not need to worry about capacity-related alpha erosion.
More importantly, every trading strategy goes through different market phases. There will be periods where the prevailing market conditions align well with the strategy, and there will also be periods where they do not. As a result, an Algo may go through a drawdown that can last for days, weeks, months, or, in some cases, even longer. This is a natural characteristic of systematic trading and should not be confused with alpha decay.
Profit and loss are both inherent to trading, and Algo trading is no exception. The objective of an Algo is not to avoid drawdowns altogether, but to execute its predefined logic consistently across varying market conditions.
I hope this helps provide a different perspective on how to interpret periods of underperformance.
4 Likes
Helpful. In short, period of underperformance need not be interpreted as Alpha decay.
But I hope u have method/process to identify an Algo which is not going to work in future.
Some algos were in drawdown from the start and not recovered till now. How to look those Algos ? Can we bet on them as they will recover in future because you released them after a lot of research and testing ?
1 Like
Hi @nijinwilson and @Avinash
I’m glad the earlier explanation was helpful.
I’d again like to highlight that drawdown, whether short-term or long-term, is a natural part of systematic trading and should not, by itself, be interpreted as alpha decay.
Before any Algo is made LIVE, it goes through a thorough research, development, and testing process across different market conditions. However, markets are dynamic, and no strategy can be expected to perform consistently across every market phase. As a result, there may be periods where an Algo underperforms depending on the prevailing market conditions.
It is also not possible to identify in advance when a drawdown will end or when a recovery phase will begin, and hence we do not make any assurances regarding future performance. The Algo simply continues to execute its predefined logic consistently throughout different market conditions.
This is one of the reasons we encourage users to evaluate an Algo over a longer horizon rather than drawing conclusions based on a single drawdown phase, and to build a diversified portfolio of uncorrelated Algos instead of depending on a single strategy.
Thanks for the reply. I have one question, I see that we can set the take profit percentage (mostly set at 50), but there is no stop loss criteria. I have observed this on credit spread algos so far. Can you deploy a loss threshold, rather than the underlying triggers criteria, thereby we can have more control in the drawdown?
I believe there’s a small misunderstanding regarding this feature.
The Profit Booking % setting does not define the overall target profit of the Algo. Instead, it determines how much of your open position is exited whenever a target level (T1, T2, T3, etc.) is reached.
For example, if an Algo enters with 10 lots and the Profit Booking % is set to 50%:
5 lots will be exited at T1.
50% of the remaining quantity will be exited at T2.
The process continues similarly for subsequent targets until the entire position is exited.
If you set the Profit Booking % to 100%, the entire position will be exited at the first target (T1) or at the Stop Loss, whichever is triggered first.
Similarly, if you set it to 0%, no partial profit booking will take place, and the complete position will remain open until the final target or the Stop Loss is hit.
Coming to your suggestion regarding a user-defined loss threshold, I’d like to clarify that the exit conditions of an Algo, including its Stop Loss logic, are an integral part of the strategy itself and are determined by the predefined trading logic. Allowing users to modify the Stop Loss independently would fundamentally alter the strategy’s behaviour and could lead to results that differ significantly from the tested performance.
That said, I appreciate your suggestion. We do have something in the pipeline that aligns closely with your requirement. Stay tuned and keep an eye on our announcements over the coming months; we’re looking forward to sharing more soon.
1 Like