FetteredChinos
Veteren member
- Messages
- 3,897
- Likes
- 40
maximum period between equity highs and max drawdown.
everything else is largely superfluous.
these two stats quantify the "tradeability" of a system.
you may well develop a super system that delivers x million points. but if it has a really low win % and high drawdown, then the time between equity highs is gonna be pretty long, and you are likely to abandon trading the strategy.
unless of course you can automate it and just leave it to run....
FC
everything else is largely superfluous.
these two stats quantify the "tradeability" of a system.
you may well develop a super system that delivers x million points. but if it has a really low win % and high drawdown, then the time between equity highs is gonna be pretty long, and you are likely to abandon trading the strategy.
unless of course you can automate it and just leave it to run....
FC
Thats interesting. So when you see these highly detailed stats with a million different facts and figures, most of it is pretty much superfluous.
I also think drawdown is very important too. I have a stat that I'm calculating which is avgWin (profit/no of winning trades) / avgLoss (loss/no of losing trades). Is that a useful stat and does it already have a name?
Annualised profit/loss, is that the average yearly profit loss?
Also, to accompany statistics would you like to be able to see individual stock equity curves as well as the overall equity curve. Having the trade entries displayed on the equity curve aswell?
How useful do you think open equity is? Do you think its more important than closed?
I also think drawdown is very important too. I have a stat that I'm calculating which is avgWin (profit/no of winning trades) / avgLoss (loss/no of losing trades). Is that a useful stat and does it already have a name?
Annualised profit/loss, is that the average yearly profit loss?
Also, to accompany statistics would you like to be able to see individual stock equity curves as well as the overall equity curve. Having the trade entries displayed on the equity curve aswell?
How useful do you think open equity is? Do you think its more important than closed?
The problem with the PLD ratio is that the longer the time frame you study, the higher the ratio tends to be. No problem if all systems are compared over the same period, but if not then yes, average the PnL and divide by the max drawdown seen over the whole period. It's just an easy way to assess the equity curve with one number.
rog1111
rog1111
pkfryer said:
FetteredChinos
Veteren member
- Messages
- 3,897
- Likes
- 40
pk, that is the profit factor that you are talking about. give or take a few fudges..
high profit factors (>5) may indicate an unsustainable period of performance.
anything about 2 is pretty darn good, IMHO.
high profit factors (>5) may indicate an unsustainable period of performance.
anything about 2 is pretty darn good, IMHO.
cheers.
I've now incorporated your suggestions into my statistics class. I've add Average Trade Value, max period between equity peaks and max drawdown (also got the run up which seems quite interesting). I can't quite do the annualised profit but I'm putting that on my list for future work.
I've got dozens of stats that I can get about the system performance for both the systems as a whole, and with individual stock in the portfolio. Just got to debug it make sure its behaving itself.
How useful do you think info about consecutive wins/loss is? I have max/min/average consecutive win/loss. For the system as a whole it calculates the consecutive winning days (and losing days)... for individual stocks it calculates the consecutive winning/losing trades.
What conclusions would you draw about consecutive win/loss? Is it useful or just plain guff?
I've now incorporated your suggestions into my statistics class. I've add Average Trade Value, max period between equity peaks and max drawdown (also got the run up which seems quite interesting). I can't quite do the annualised profit but I'm putting that on my list for future work.
I've got dozens of stats that I can get about the system performance for both the systems as a whole, and with individual stock in the portfolio. Just got to debug it make sure its behaving itself.
How useful do you think info about consecutive wins/loss is? I have max/min/average consecutive win/loss. For the system as a whole it calculates the consecutive winning days (and losing days)... for individual stocks it calculates the consecutive winning/losing trades.
What conclusions would you draw about consecutive win/loss? Is it useful or just plain guff?
FetteredChinos
Veteren member
- Messages
- 3,897
- Likes
- 40
making things more complicated than usual??
all that matters is that
a) the system is net positive
and
b) it is tradeable
confusing the issue with loads of stats can just make you hesitant to pull the trigger at times, even though the "system" says so.
if you are in doubt about a system, then just trade several at once to smooth out the old equity curve.
it may make things a bit complicated when running mutliple positions according to different criteria, but it is a good way to ride out the bad times.
FC
all that matters is that
a) the system is net positive
and
b) it is tradeable
confusing the issue with loads of stats can just make you hesitant to pull the trigger at times, even though the "system" says so.
if you are in doubt about a system, then just trade several at once to smooth out the old equity curve.
it may make things a bit complicated when running mutliple positions according to different criteria, but it is a good way to ride out the bad times.
FC
Actually I think thats a good point. The statistics isn't just about the system behaviour its about our behaviour and how we would psychologically/emotionally react whilst trading the system. I think as far as the system stats is concerned if it consistently wins in the long run thats all that matters. But for us trading it, the other stats plays more of a role.
Rog, you jogged my memory about why I was including consecutive wins/loss. It was actually for my own psychological needs to 'feel' like the system is working. I remember sitting down over a year ago and telling myself that I needed a system that produced twice as many wins than loses and would show reasonably long consecutive wins with short consecutive losses. Just so that on a day to day basis I felt confident, even though I know that its just as likely (or maybe even more likely) that a system with fewer wins but high win/loss ratio would produce better results in the long term.
The consecutive winning days is exactly the same as the run-up stat (opposite to drawdown), which shows how much profit was made without a single losing day. Useless as far as a system is concerned really but purely an emotional support.
Maybe if the system is going to be traded mechanically it doesn't matter as much, but if it is manual and discretionary then it probably has more impact. Although having really good stats including long consecutive wins shows signs of a possibly highly robust efficient system
Fettered, I'm planning on using multi-system trading, several systems running that specialise on range and trend trading. I'm also using reasonably large portfolio's to balance out risk. Also using recent history optimisation with a dynamic look back period (takes a lot of processing time for that though). One of the reasons I'm concentrating on stats and which stats are the best is also because I'm planning on changing the optimiser to possibly work off statistics for the look back period instead of just profit. So you can tell the optimiser your wish list and it will produce parameter set ups that will match it as best as possible. Although, whats a more important stat than making the most amount of money?
Is there a profit expectancy/probability stat per trade? I remember it being discussed before?
Rog, you jogged my memory about why I was including consecutive wins/loss. It was actually for my own psychological needs to 'feel' like the system is working. I remember sitting down over a year ago and telling myself that I needed a system that produced twice as many wins than loses and would show reasonably long consecutive wins with short consecutive losses. Just so that on a day to day basis I felt confident, even though I know that its just as likely (or maybe even more likely) that a system with fewer wins but high win/loss ratio would produce better results in the long term.
The consecutive winning days is exactly the same as the run-up stat (opposite to drawdown), which shows how much profit was made without a single losing day. Useless as far as a system is concerned really but purely an emotional support.
Maybe if the system is going to be traded mechanically it doesn't matter as much, but if it is manual and discretionary then it probably has more impact. Although having really good stats including long consecutive wins shows signs of a possibly highly robust efficient system
Fettered, I'm planning on using multi-system trading, several systems running that specialise on range and trend trading. I'm also using reasonably large portfolio's to balance out risk. Also using recent history optimisation with a dynamic look back period (takes a lot of processing time for that though). One of the reasons I'm concentrating on stats and which stats are the best is also because I'm planning on changing the optimiser to possibly work off statistics for the look back period instead of just profit. So you can tell the optimiser your wish list and it will produce parameter set ups that will match it as best as possible. Although, whats a more important stat than making the most amount of money?
Is there a profit expectancy/probability stat per trade? I remember it being discussed before?
