Forex trader success rates - some real data

[Rhody Trader]

As part of my PhD research (yes, I'm doing a doctorate focused on trading) I am looking into retail forex trader profitability. Since the whole question of whether 95% of traders actually fail or not is such a hot button topic, I figured I would take that on as a bit of a side project and share what I find with folks as I work through the numbers. I've posted some initial results on my blog in the post Starting to detail forex profitability data. As the title implies, it is just a start at this stage. There's a lot of data and a lot of ways to slice and dice it. Still, comments are very welcome.

I will post further results as I am able to do additional analysis.

 

[Shakone]

Thanks for the analysis John. Good to see some more realistic figures. So are the 8000 accounts the survivors? Presumably there were other traders, but they stopped or blew up or whatever.

So just to check I've understood, of the survivors, 3% (241) were profitable every month in the data, but only 0.4% (30) were profitable every quarter for more than a year, the other 211 may or may not have been.

I suppose it would be nice to know how profitable these accounts were, and how many trades they made, with the intent to find how many of them could be accounted for just by sheer chance. Also might be interesting to note how long the successful accounts have been trading and whether there is a strong correlation between success and experience.

 

[neil]

Thanks for the analysis John. Good to see some more realistic figures. So are the 8000 accounts the survivors? Presumably there were other traders, but they stopped or blew up or whatever.

So just to check I've understood, of the survivors, 3% (241) were profitable every month in the data, but only 0.4% (30) were profitable every quarter for more than a year, the other 211 may or may not have been.

I suppose it would be nice to know how profitable these accounts were, and how many trades they made, with the intent to find how many of them could be accounted for just by sheer chance. Also might be interesting to note how long the successful accounts have been trading and whether there is a strong correlation between success and experience.

Good point - Taleb reckons you need long periods of data to flatten out the "Luck" factor.

 

[BeginnerJoe]

The loosers will never admit to loose. So it seems like a fruitless project. Also, on the internet, everyone is a winner, and every tom, dick, and harry is turning £100 into £500k through spread betting in a matter of months. All for tax free too ! The winners usually work for BT, British Gas, etc, and maybe a taxi driver.

 

[Rhody Trader]

Thanks for the analysis John. Good to see some more realistic figures. So are the 8000 accounts the survivors? Presumably there were other traders, but they stopped or blew up or whatever.

The 8000+ accounts are not, actually, the survivors. It includes everyone who did at least one trade in a 3+ year period. There is survivorship involved in the % profitable each quarter (I think that will be the next post), but not in the quarterly figures.

So just to check I've understood, of the survivors, 3% (241) were profitable every month in the data, but only 0.4% (30) were profitable every quarter for more than a year, the other 211 may or may not have been.

There isn't any monthly data in those figures, just quarterly. I'll get to monthly stuff later. This time I just wanted to compare to the broker-reported stuff. You're right that only 30 were able to be profitable more than 4 straight quarters.

I suppose it would be nice to know how profitable these accounts were, and how many trades they made, with the intent to find how many of them could be accounted for just by sheer chance. Also might be interesting to note how long the successful accounts have been trading and whether there is a strong correlation between success and experience.

All things on my long list of topics for future analysis.

 

[Rhody Trader]

Good point - Taleb reckons you need long periods of data to flatten out the "Luck" factor.

It's more a question of quantity of observations rather than length of period. You're likely to get more information in a year about a trader who trades 20 times a month than you would in 10 years from a trader who trades only once a month.

 

[Rhody Trader]

The loosers will never admit to loose.

I'm not using self-reported data, except in terms of some demographic type information. These are actual broker-reported transactions.

 

[BeginnerJoe]

All things on my long list of topics for future analysis.

Glad to have an intelligent PhD on board. All the others I have seen here wanna play the market casino instead. They think the market is some kind of an ATM. I can foresee great things for you when the bookies will pay to have you doing nothing.

 

[the hare]

Your analysis shows a slightly higher percentage of traders are profitable each quarter than those reported by the US brokers. I assume you use the same criteria to define what constitutes a profitable trader.

I also assume that your sample population contains both US and non US brokers. If you compare the US brokers only do these differences still exist (I appreciate the differences might be statistically insignificant)

What assurances do you have that the broker data is an accurate reflection of reality, is it possible that losing trades have been removed from the data sets. Is it possible to correlate the data against published profit and loss data for the brokers ?

The percentage of profitable traders after X trades by random chance is simple enough to model, and I suspect that your data will verify that consistently profitable traders are a rare breed. However I am intrigued as to why any broker would happily supply data to be used to prove this point !

Surely It doesn't really make much commercial sense to reveal these figures ?

I would be interested in seeing equity curves plotted for the 8000 or so accounts. I suspect they'll look very similar to my random robots

 

[cablemonster]

Your analysis shows a slightly higher percentage of traders are profitable each quarter than those reported by the US brokers. I assume you use the same criteria to define what constitutes a profitable trader.

I also assume that your sample population contains both US and non US brokers. If you compare the US brokers only do these differences still exist (I appreciate the differences might be statistically insignificant)

What assurances do you have that the broker data is an accurate reflection of reality, is it possible that losing trades have been removed from the data sets. Is it possible to correlate the data against published profit and loss data for the brokers ?

The percentage of profitable traders after X trades by random chance is simple enough to model, and I suspect that your data will verify that consistently profitable traders are a rare breed. However I am intrigued as to why any broker would happily supply data to be used to prove this point !

Surely It doesn't really make much commercial sense to reveal these figures

when the data is from a vendor and the study is conducted by a vendor then this can cause a skew in the results.

 

[BeginnerJoe]

I would be interested in seeing equity curves plotted for the 8000 or so accounts. I suspect they'll look very similar to my random robots

Exactly. There's no advantage to random at all.

 

[Rhody Trader]

Your analysis shows a slightly higher percentage of traders are profitable each quarter than those reported by the US brokers. I assume you use the same criteria to define what constitutes a profitable trader.

To the best of my knowledge, the CFTC uses a very simple set of metrics for their required reporting - at least one trade and profitable, meaning P/L > 0. That's what I have used in my figures.

I also assume that your sample population contains both US and non US brokers. If you compare the US brokers only do these differences still exist (I appreciate the differences might be statistically insignificant)

You are correct. I do have a global sample, so to the extent that US traders may be worse than non-US traders there could be at least some explanation for the difference. That's one of the things on my list to look at.

What assurances do you have that the broker data is an accurate reflection of reality, is it possible that losing trades have been removed from the data sets. Is it possible to correlate the data against published profit and loss data for the brokers ?

The transaction data was (for the most part) not extracted after the fact, but rather in real time and were back-filled was extracted directly from trader accounts by a third party, not by the brokers themselves. As such, there could be no cherry picking of trades.

By published P/l data for the brokers, to what are you referring?

 

[Rhody Trader]

when the data is from a vendor and the study is conducted by a vendor then this can cause a skew in the results.

If you want to make some kind of accusation, just spit it out already and be done with.

 

[neil]

It's more a question of quantity of observations rather than length of period. You're likely to get more information in a year about a trader who trades 20 times a month than you would in 10 years from a trader who trades only once a month.

No-you have to include time as well -a long period of time to flatten out "luck" -please read "Taleb," it relates to some of what you allege to be attempting alongside your vending.

 

[snellys3lions]

is anyone on this site a paid up member of fat prophets ?

 

[Rhody Trader]

No-you have to include time as well -a long period of time to flatten out "luck" -please read "Taleb," it relates to some of what you allege to be attempting alongside your vending.

I've read Taleb. I know exactly what you're talking about. You're absolutely right that you want time as well. Ideally, you have in someone's history coverage of all potential future market conditions, though that's a bit unrealistic. Given an either/or choice, however, you favor quantity because it will help you establish if there are significant patterns of behavior/performance. Time is only useful if there are enough meaningful observations. A 50-year track record isn't any good if there are only 10 trades. Extreme example, obviously, but you get the point.

As for the vending crack, if my time as a member here, and my service to the community in various roles over the years, doesn't earn me any credibility then maybe I should pack it in and say my farewells.

 

[wackypete2]

As part of my PhD research (yes, I'm doing a doctorate focused on trading) I am looking into retail forex trader profitability. Since the whole question of whether 95% of traders actually fail or not is such a hot button topic, I figured I would take that on as a bit of a side project and share what I find with folks as I work through the numbers. I've posted some initial results on my blog in the post Starting to detail forex profitability data. As the title implies, it is just a start at this stage. There's a lot of data and a lot of ways to slice and dice it. Still, comments are very welcome.

I will post further results as I am able to do additional analysis.

Good luck with the research and PhD

Peter

 

[counter_violent]

I'd be interested to know if account size plays any part in all this.
Maybe this info is not available.

 

[Rhody Trader]

Good luck with the research and PhD

Peter

That's Peter. If nothing else, it will be an educational experience.

Yes, I really did make that joke. It's been a long day.

 

[Rhody Trader]

I'd be interested to know if account size plays any part in all this.
Maybe this info is not available.

I have seen stuff which suggests account size does play a factor in performance (I think maybe from FXCM/DailyFX). The data I have does include some account balances, but I hesitate to rely on them too much. Trade size, however, could be a dirty proxy for account size.