I define ‘edge’ as a positive expected value, after commissions. I think that different people define ‘edge’ in different ways, but we are (mostly) all roughly describing the same concept. More precisely, I define edge as an expected value, after commissions, that exceeds the risk-free rate, but since I only trade strategies that significantly exceed the risk-free rate, we can drop that complication. Moreover, it takes a lot of data to shrink the confidence interval on the point estimate of the expected value (depending on the strategy).
Having an edge is an insufficient criterion for trading a strategy because it leaves out the whole question of risk. To assess risk, I look at the distribution of profits and losses (mark-to-market, not just realized). I also conduct simulations, bootstrapping, etc. That is another topic.
Focusing just on edge, for me your question translates into:
“How do you find strategies that result in positive expected value?”
Simply put, I do things that the majority of market participants are unwilling or unable to do. I find these opportunities by thinking (sometimes literally sitting back in a recliner with a coffee) about the different types of market participants, their varying objectives, constraints, etc. I also read a lot.
The reason I crafted trading into a career is because I love to think, I love to read, I love math, statistics, programming and finance. I also love independence. This work allows me to be free to think as hard as I can and read as much as I want with complete freedom to decide when and what. For me, that's the perfect life.
Once I have a theory about the markets that makes sense, I test it. I try not to test too many ideas without considering them carefully first, because even with proper statistics, occasionally I’ll get something that looks good purely due to chance. This is just the nature of the world and everyone is subject to it. That said, if 9/10 strategies are based on a real signal, for example, and 1/10 are not, it’s not the end of the world… but I would like to minimize the number of strategies that appear good but are actually based on noise as much as possible.
Note that I am not satisfied that I’ve found an edge just because I’ve back-tested a strategy and found a positive expected value. The first step is to verify that the edge is statistically significant (simple), but even that is insufficient. I also want to see how well the distribution I have may reflect the potential future distribution of profits and losses… this takes much more work. I do this by performing simulations, looking at covariates, analyzing robustness, etc. In other words, I may have discovered an edge in my sample, but that in and of itself is not enough for me to conclude that I have an edge in the population.
I also work hard to always ask myself if the work I'm doing in the moment is the work that is most likely to be the most fruitful, or not. It’s very tempting (for me at least) to do work that has the mere potential to be fruitful, but that is really just the most interesting, exciting, intellectually stimulating, etc. I try, the best I can, to cut that out and focus on work that has the highest probability of being useful. Recently, that includes hiring out very important but mind-numbing tasks (eg data cleaning) that I can verify were correctly done after the fact with much less effort than it would have taken me to do them myself in the first place.
This is a rough description of my framework. I hope there's something useful there for you.
Having an edge is an insufficient criterion for trading a strategy because it leaves out the whole question of risk. To assess risk, I look at the distribution of profits and losses (mark-to-market, not just realized). I also conduct simulations, bootstrapping, etc. That is another topic.
Focusing just on edge, for me your question translates into:
“How do you find strategies that result in positive expected value?”
Simply put, I do things that the majority of market participants are unwilling or unable to do. I find these opportunities by thinking (sometimes literally sitting back in a recliner with a coffee) about the different types of market participants, their varying objectives, constraints, etc. I also read a lot.
The reason I crafted trading into a career is because I love to think, I love to read, I love math, statistics, programming and finance. I also love independence. This work allows me to be free to think as hard as I can and read as much as I want with complete freedom to decide when and what. For me, that's the perfect life.
Once I have a theory about the markets that makes sense, I test it. I try not to test too many ideas without considering them carefully first, because even with proper statistics, occasionally I’ll get something that looks good purely due to chance. This is just the nature of the world and everyone is subject to it. That said, if 9/10 strategies are based on a real signal, for example, and 1/10 are not, it’s not the end of the world… but I would like to minimize the number of strategies that appear good but are actually based on noise as much as possible.
Note that I am not satisfied that I’ve found an edge just because I’ve back-tested a strategy and found a positive expected value. The first step is to verify that the edge is statistically significant (simple), but even that is insufficient. I also want to see how well the distribution I have may reflect the potential future distribution of profits and losses… this takes much more work. I do this by performing simulations, looking at covariates, analyzing robustness, etc. In other words, I may have discovered an edge in my sample, but that in and of itself is not enough for me to conclude that I have an edge in the population.
I also work hard to always ask myself if the work I'm doing in the moment is the work that is most likely to be the most fruitful, or not. It’s very tempting (for me at least) to do work that has the mere potential to be fruitful, but that is really just the most interesting, exciting, intellectually stimulating, etc. I try, the best I can, to cut that out and focus on work that has the highest probability of being useful. Recently, that includes hiring out very important but mind-numbing tasks (eg data cleaning) that I can verify were correctly done after the fact with much less effort than it would have taken me to do them myself in the first place.
This is a rough description of my framework. I hope there's something useful there for you.
