My 5c. Take is with a grain of salt as anything you find online.
More often than not edge is there longer than many people think. That is for edges that are found on 10y+ data samples.
What typically happens is - market becomes more efficient and edge's expectancy goes down. Depending on the way how you extracted that edge (commissions structure, access to leverage, size of account, exact setup etc) - edge might become non tradable or profit potential not worth a risk anymore.
Quite often when someone says they've got an edge they mean they "feel" they have an edge but their data sample is not statistically significant. Edge discovered on a short time periods like 1-3 year or tested only during one sample of one market type will also not likely to last.
Edge also might more or less suddenly disappear due to
- A significant market change (decimalization, too strict margin requirement, wide spread of commissionless trading). Basically that means that something that has been not widely available now is not, or you are no longer able to trade this way
- Disappearance of the market traded
- Regulations
I try to stick to edges that can be
- Precisely measured
- Have 3000+ trades in test sample
- Tested across multiple market cycles and market types
- Work over at least 30 years, preferably 50
Good ones I've tested going back to 1950. Some would consistently get worse every decade but are still tradable.
Less obvious but quite popular, in my opinion, reason for edge to stop working - a setup is too specific.
the less rules a setup has > the more trades it will generate > the more significant findings will be > the better chance it will keep working in the future
All the edges I use are going back to at least early 80-ies, most all the way to 50-ies. Before - I don't have data to check, but that is sufficient proof for me.
My journal
Val
