There would not be a limit. Note that the Qualifying to be able to press the "Fund My Strategy" button (2 year backtest, Sharpe >1.25, Calmar >1.25, Trades/year >500, Total profit >$250k) does not automatically qualify you for funding. It is simply there to stop people randomly pressing it on every backtest. Once you hit the minimum requirements we would take a look at your strategy and attempt to help you refine it to a point where it could be funded. Knowledge of history makes it easy to make a backtest that can qualify for any requirements set.
Thanks for the response. I was wondering if you could tell me some of the details regarding refining my strategy to get funded? I am trying to figure out how much time to devote to this project if any at all and gauge if the conditions to get funded are realistically possible for me given the time I can devote to the project.
I am also wondering how can I get my strategy properly funded since it is already licensed and does great out of sample but there is still no significant allocation yet. Definitely disappointing.
Not knowing a lot of information about you (age, trading experience, programming experience, analytical experience) I can only give you an idea of what I have observed over 15 years of doing this.
Put very simply, there are two approaches to generating a model, the trader "event observation" model and the Quant "analysis of data" model. Most of my experience is with the former.
A trader traditionally notices an event in the market that looks like it generates a trading opportunity, it is profitable for them and they want to repeat it. So they make some kind of scanner which scans the market (whole or just a selection) which alerts then should this event occur again. This could be anything from a 5% move in 5 mins to President Trump tweeting about China. They then watch and trade some of these alerts, over time refining the alert until it is producing a solid trading opportunity more often than not. They may then choose to automate entry, manually choosing when to exit the position. If this continues to be profitable they may backtest the idea with a number of exit criteria based on their experience (exit criteria is absolutely as important as entry criteria). The backtest may suggest that the opportunity works and so they will allocate a little of their trading capital to this model. If this works out they will tweak and add more capital.
I have been involved at the "create an alert" stage and the "backtest a model" stage. I have seen thousands of trading ideas/concepts. I have written many hundreds of alerts for other people and written many hundreds of models for myself and others. I have worked with incredibly smart people doing the same work.
The most common misconception I have seen, particularly from people outside of the trading community but from inside as well, is the idea that "my concept is golden and will work first time". You are more likely to win the lottery.
Creating a working model takes a combination creative thinking and A LOT of hard work. From my position of observing thousands of ideas I would say the hit rate is just below 1% (and these are model ideas from people who TRADE... all day, every day). An idea can fail at the first hurdle or it can get all the way to trading live only for some unexpected cost or influence to cause it to fail.
Also, consider every trading idea you see out there on the internet as nothing more than training, if it worked it would not be out there in public for all to see. Bollinger Band Breakouts, Moving Average Crosses, they are all interesting but really just training. Nobody gives away a good idea, not even on a white paper from a prestigious college.
That all sounds a little deflating, right?
The thing about the lottery is that the odds never change, for you, for me, the odds are always the same.
But writing a trading algo is a creative endeavour. It is like writing music, I can teach you the concept, you can practice like crazy, you can get really good but will you ever be the Beatles? Do you have to be the Beatles? Is it enough to make a nice tune?
If you have coding skills, data analytic skills, you are prepared to learn all the ins and outs of trading and you have an interest then this can be a wonderfully creative outlet and one which CAN generate income, potentially massive income. Your view of the market is unique, and it may be unique enough to locate an interesting opportunity that no-one else has observed.
But even when you find a good idea you cannot rest on your laurels, most Quants have a set of models in play but are constantly researching new ideas and opportunities because they know that their current models will decay. The opportunity will be discovered by others and market efficiency will eventually decay the opportunity to nothing.
Approach it realistically, understand the amount of work involved, treat it as a hobby, as a learning opportunity that may result in profit. There are lots of things to learn, and lots of pitfalls. I could write hundreds of models right now that meet our "Fund My Strategy" criteria because I know what happened in the past, obviously that would be pointless and a waste of time for all of us, but people do it all the time unintentionally. People think they can short any symbol that they want at any time they want, or take any size they want without affecting the price at all. Understand that even if you make a model that meets our initial criteria, it does not mean that it will be taken live in the market. Even if it is taken live in the market, it does not mean it will automatically be sized up. You are not only competing with the market, you are competing with other models for capital.
Our purpose for making CloudQuant available was that there are lots of smart people who are not currently in the trading environment who could find new trading opportunities. We understood at the outset that it would be like finding needles in a haystack. We chose Python not only because it worked for us but understanding the success rate at building a model, our users should get something else out of this. Learning Python will be a lifelong career benefit for anyone. Understanding the markets also has lifelong benefits.
Thank you Paul for your wisdom. That was a very good write-up. If it helps anyone who is getting started, you'll realize what Paul is saying is true. I started heavy into CQ about 6 months ago. One of the reasons I wanted to use CQ was to learn Python. I've used FORTRAN, C, VB.NET and VBA in the past, but never dove into Python until now. It's been a great learning experience.
I thought I was coming into this with a reasonable arsenal of knowledge. While I don't trade for a living, I've developed my own multi-day trading models which I use in my retirement accounts and I trade those myself. Are they great? There are probably models out there that are better, but they work for me and since I created them myself I'm very fond of them. That means I follow the rules and I don't miss signals. Not missing signals or second guessing the model is more important to me than searching the earth for a model that is better. My retirement accounts are doing just fine after about 12 years of forward (unseen data) trading.
With that confidence I dove into CQ. It turns out all my ideas about day trading are pretty much false. I had several ideas, some of which were already back tested in limited scope, and all of them have failed miserably when I wrote scripts in CQ which provides a much more extensive way to back test day strategies than I've had access to in the past. So, as Paul said, this should be looked at as sort of a hobby. It's a creative endeavor. It's an opportunity to think about the market and try to find those statistical edges that exist but you might not see them on one single chart. You have to find it recurring over and over again. I've learned from CQ that you might have something that looks really good with 1000 trades, and then add a different symbol and it all falls apart.
CQ is FREE. Holy cow, it's FREE! 10 years ago, I was scraping data off of yahoo finance, building spreadsheets so I could adjust for splits, and spending tons of time putting together crude tests. Now I jump on CQ at lunch time, whip out another crazy idea script and hit the run button. Maybe nobody remembers the world 10 or 20 years ago, but the fact that CQ offers this platform completely free I think is amazing.
I'm happy to say that I at least have made it to the point that I could hit the "FMS" button. That was a very proud moment for me since I know that I didn't intentionally "stack the deck". My model has not yet made the "fundable" stage, and maybe it never will. I don't know. But, I do enjoy reading some of these posts, learning new things, tinkering around with code, and trying to figure out how to pull money out of the stock market.
No matter what happens, I'll have money to retire... even if CQ is always just a hobby.
I've gradually come to accept that I get a lot of enjoyment just from the process of building the strategies, but may never get to the level where I attract serious funding.
The CQ platform suits me fine, as I would rather investigate time series methods, rather than analyze accounting data.
The healthy attitude, as you say, is to treat it as a fascinating hobby.
Comments
There would not be a limit. Note that the Qualifying to be able to press the "Fund My Strategy" button (2 year backtest, Sharpe >1.25, Calmar >1.25, Trades/year >500, Total profit >$250k) does not automatically qualify you for funding. It is simply there to stop people randomly pressing it on every backtest. Once you hit the minimum requirements we would take a look at your strategy and attempt to help you refine it to a point where it could be funded. Knowledge of history makes it easy to make a backtest that can qualify for any requirements set.
Thanks for the response. I was wondering if you could tell me some of the details regarding refining my strategy to get funded? I am trying to figure out how much time to devote to this project if any at all and gauge if the conditions to get funded are realistically possible for me given the time I can devote to the project.
I am also wondering how can I get my strategy properly funded since it is already licensed and does great out of sample but there is still no significant allocation yet. Definitely disappointing.
Not knowing a lot of information about you (age, trading experience, programming experience, analytical experience) I can only give you an idea of what I have observed over 15 years of doing this.
Put very simply, there are two approaches to generating a model, the trader "event observation" model and the Quant "analysis of data" model. Most of my experience is with the former.
A trader traditionally notices an event in the market that looks like it generates a trading opportunity, it is profitable for them and they want to repeat it. So they make some kind of scanner which scans the market (whole or just a selection) which alerts then should this event occur again. This could be anything from a 5% move in 5 mins to President Trump tweeting about China. They then watch and trade some of these alerts, over time refining the alert until it is producing a solid trading opportunity more often than not. They may then choose to automate entry, manually choosing when to exit the position. If this continues to be profitable they may backtest the idea with a number of exit criteria based on their experience (exit criteria is absolutely as important as entry criteria). The backtest may suggest that the opportunity works and so they will allocate a little of their trading capital to this model. If this works out they will tweak and add more capital.
I have been involved at the "create an alert" stage and the "backtest a model" stage. I have seen thousands of trading ideas/concepts. I have written many hundreds of alerts for other people and written many hundreds of models for myself and others. I have worked with incredibly smart people doing the same work.
The most common misconception I have seen, particularly from people outside of the trading community but from inside as well, is the idea that "my concept is golden and will work first time". You are more likely to win the lottery.
Creating a working model takes a combination creative thinking and A LOT of hard work. From my position of observing thousands of ideas I would say the hit rate is just below 1% (and these are model ideas from people who TRADE... all day, every day). An idea can fail at the first hurdle or it can get all the way to trading live only for some unexpected cost or influence to cause it to fail.
Also, consider every trading idea you see out there on the internet as nothing more than training, if it worked it would not be out there in public for all to see. Bollinger Band Breakouts, Moving Average Crosses, they are all interesting but really just training. Nobody gives away a good idea, not even on a white paper from a prestigious college.
That all sounds a little deflating, right?
The thing about the lottery is that the odds never change, for you, for me, the odds are always the same.
But writing a trading algo is a creative endeavour. It is like writing music, I can teach you the concept, you can practice like crazy, you can get really good but will you ever be the Beatles? Do you have to be the Beatles? Is it enough to make a nice tune?
If you have coding skills, data analytic skills, you are prepared to learn all the ins and outs of trading and you have an interest then this can be a wonderfully creative outlet and one which CAN generate income, potentially massive income. Your view of the market is unique, and it may be unique enough to locate an interesting opportunity that no-one else has observed.
But even when you find a good idea you cannot rest on your laurels, most Quants have a set of models in play but are constantly researching new ideas and opportunities because they know that their current models will decay. The opportunity will be discovered by others and market efficiency will eventually decay the opportunity to nothing.
Approach it realistically, understand the amount of work involved, treat it as a hobby, as a learning opportunity that may result in profit. There are lots of things to learn, and lots of pitfalls. I could write hundreds of models right now that meet our "Fund My Strategy" criteria because I know what happened in the past, obviously that would be pointless and a waste of time for all of us, but people do it all the time unintentionally. People think they can short any symbol that they want at any time they want, or take any size they want without affecting the price at all. Understand that even if you make a model that meets our initial criteria, it does not mean that it will be taken live in the market. Even if it is taken live in the market, it does not mean it will automatically be sized up. You are not only competing with the market, you are competing with other models for capital.
Our purpose for making CloudQuant available was that there are lots of smart people who are not currently in the trading environment who could find new trading opportunities. We understood at the outset that it would be like finding needles in a haystack. We chose Python not only because it worked for us but understanding the success rate at building a model, our users should get something else out of this. Learning Python will be a lifelong career benefit for anyone. Understanding the markets also has lifelong benefits.
Ultimately, though, the decision is yours.
Thank you Paul for your wisdom. That was a very good write-up. If it helps anyone who is getting started, you'll realize what Paul is saying is true. I started heavy into CQ about 6 months ago. One of the reasons I wanted to use CQ was to learn Python. I've used FORTRAN, C, VB.NET and VBA in the past, but never dove into Python until now. It's been a great learning experience.
I thought I was coming into this with a reasonable arsenal of knowledge. While I don't trade for a living, I've developed my own multi-day trading models which I use in my retirement accounts and I trade those myself. Are they great? There are probably models out there that are better, but they work for me and since I created them myself I'm very fond of them. That means I follow the rules and I don't miss signals. Not missing signals or second guessing the model is more important to me than searching the earth for a model that is better. My retirement accounts are doing just fine after about 12 years of forward (unseen data) trading.
With that confidence I dove into CQ. It turns out all my ideas about day trading are pretty much false. I had several ideas, some of which were already back tested in limited scope, and all of them have failed miserably when I wrote scripts in CQ which provides a much more extensive way to back test day strategies than I've had access to in the past. So, as Paul said, this should be looked at as sort of a hobby. It's a creative endeavor. It's an opportunity to think about the market and try to find those statistical edges that exist but you might not see them on one single chart. You have to find it recurring over and over again. I've learned from CQ that you might have something that looks really good with 1000 trades, and then add a different symbol and it all falls apart.
CQ is FREE. Holy cow, it's FREE! 10 years ago, I was scraping data off of yahoo finance, building spreadsheets so I could adjust for splits, and spending tons of time putting together crude tests. Now I jump on CQ at lunch time, whip out another crazy idea script and hit the run button. Maybe nobody remembers the world 10 or 20 years ago, but the fact that CQ offers this platform completely free I think is amazing.
I'm happy to say that I at least have made it to the point that I could hit the "FMS" button. That was a very proud moment for me since I know that I didn't intentionally "stack the deck". My model has not yet made the "fundable" stage, and maybe it never will. I don't know. But, I do enjoy reading some of these posts, learning new things, tinkering around with code, and trying to figure out how to pull money out of the stock market.
No matter what happens, I'll have money to retire... even if CQ is always just a hobby.
I think that's the right attitude to have, JH.
I've gradually come to accept that I get a lot of enjoyment just from the process of building the strategies, but may never get to the level where I attract serious funding.
The CQ platform suits me fine, as I would rather investigate time series methods, rather than analyze accounting data.
The healthy attitude, as you say, is to treat it as a fascinating hobby.