Which measure is more important in estimating TS performance?
I think who C2 people is too much on equity performance. While isn’ t better analyze number like: adp ratio (average profit to drawdown), win% and win/loss ratio?
Example in ours TS we have: adp ratio 2.40, win% 86, average loss 0.083%, average win 0.338% and max loss 0.25%(on started capital+netprofit). With this number i know all i must for make from myself my risk/reward profile.
I prefer to control these values…
Domenico like like suggested to me 
Eugenio
Not sure about your english or sitgnal. However with a post like that you must be a genious at getting people to look at your signal.
Another example:
If someone suscribe my system he know: win% 86, average loss 0.083%, average win 0.338% and max loss 0.25%(on started capital+netprofit). Therefore if i put a order of buy 5.000,00 $ of xy the suscribe can decide how to risk: 5.000,00 / 0.25% * how to risk in % (ex.5.000,00 / 0.25% * 1% if i risk 1%) and he know who the odds of win are 86%, and if the trade win the attended gain is= how to risk in % / 0.25% * average win 0.338% while if the trade lose the attended lose is=how to risk in % / 0.25% * average loss 0.338%.
Threfore this are the more important number! %win, adp ecc…
In fact the ours TS that has elevated adp a ratio, %win but not stellar result is between first in classifies…
And after 8 months, a person would be able to save $150 a month of your system by simply buying and holding, since only in best-case without commissions or realism do you slightly outperform the S&P 500… And the APD, profit factor, and win% would be much higher as well…
In other words, self-congratulations should be left to the few people on C2 who actually are significantly outperforming the market.
In effects from your answer it seems that I write sanscrito!
The question isn’t “please subscribe my system”, but rather: beyond to perfomance there is nothing?
The measures you mentioned are all a sort of means. One reason to look at the equity curve is that variation and outliers are important too. A second reason is that simultaneous positions may be correlated. A third reason is that subsequent positions can be autocorrelated. A fourth reason is to compare it with buy & hold S&P. So I don’t think you can control your risk/ reward profile if you use only the statistics you mentioned.
Right observations, but i don’ t use only the statistics, I use the max loss for trade too.
well in comparing equity curves, it seems your system has a slightly bigger max DD than the S&P as well…
well in comparing equity curves, it seems your system has a slightly bigger max DD than the S&P as well…
Even so you will miss most of the points that I described.
The maximum loss is only one observation from the tail of the distribution. Basically you will still not have a clue of the probability that the next loss will be 4 times higher.
Similarly, you don’t know whether losses have the tendency to form conscecutive rows. A while ago TMG had 7 losses in row and someone computed that the probability of that “should” have been less than a billion or so.
You will also miss the effect of portfolio diversification.
I can understand that you like the beauty of a few simple statistics. But I’m afraid that this simplicity tells more about the human mind then about the data.
I think the greater concern is not so much the measure but the number of observations (either trading days, or trades). Many systems show great performance stats but they are based on too few observations to provide a reliable prediction for future performance.
Take for example the system MBN-1. It did great on almost any measure you could think of for a full 6 months after it started. It outperformed the S&P500 in a spectacular way, had low drawdowns, a high Sharpe ratio and an incredible return. Still, you would have lost money if you subscribed after the 6th month.
Ok, but drowdaw on a trade in profit!
"The maximum loss is only one observation from the tail of the distribution"
for me max loss don’t came from observation, but from money managment.
money to invest =money to risk (now 10%) / max loss (now 2,5%)* stop loss%. therefore we risk 0,25%
"Similarly, you don’t know whether losses have the tendency to form conscecutive rows"
ok, corrected observation! In fact the in ours money managment we put it:
money to invest = money to risk (now 10%) +/-netprofit / max loss (now 2,5%)* stop loss%
therefore if we have consecutive loss we never stopped out and never los more then 10%.
Not magic, but just a money managment system…
"I think the greater concern is not so much the measure but the number of observations (either trading days, or trades)."
Sure, I think the same
I think I don’t understand what you mean in the first paragraph. Where do you get that 2,5% from? Is that not an observation?
I don’t understand the second paragraph either. It seems like you are saying that you cannot loose more than you invest. This isn’t always true, but even if it is true, I don’t see how it helps to select a system.
I will be the first to admit that the number of observations is often too small, and more importantly, from a too small range of market conditions. But the question was why we need an equity curve besides the trade statistics. If the range of market conditions is too small to rely on the equity curve then it is also too small to rely on the trade statistics.
Well… but isn’t drawdown on a trade in progress a sign of profits that could have been locked in by closing the position? The vendor obviously still has the ability to open a new position later to complete this trade cycle.
lol
money to invest =money to risk (now 10%) / max loss (now 2,5%)* stop loss%. therefore we risk 0,25%
If I might to say.
Stop loss doesn’t guarantee a fill at the price in long term stocks system. Simply you’ll be filled at market on opening at your stop if bad news came for a stock. With average holding of ~70 days your risk is undefined even if you have stop loss. Your equity curve shows the same on Feb. drop. Particularly for AMEX, be my guest with market orders there lol
Eu
at the risk of digressing a little: I’m less worried about observing a limited range of market conditions, as you can hedge this (with some effort and skill) if you’re willing to give up some profits.
Going back to the original question: I think the equity curve can be considered an outcome measure. It is the ultimate measure that counts: How much do I see my account grow each day, week or month. Is that growth statistically meaningful. How much can I expect to lose on a bad day, etc.
Statistics based on trades can be considered a process measure. It gives some insight in the process of realizing the equity curve. For example, I like to see both a larger number of winning than losing trades, AND a larger average profit than loss per trade. I don’t like to see an extremely small profit per trade, as I know the system then might be vulnerable to slippage and execution problems. I don’t like to see occasional large drawdowns per trade, as I know one bad decision by the vendor could have a severe impact on my account, etc. etc.
2,5% where come from:
I choose max loss on each trade and the capital to risk, example:
I have 10% capital to risk and max loss on each trade 2,5%, therefore
capital to invest= 10% capital to risk / stop loss on trade * 2,5% max loss on trade.
Example my TS put a buy order on XY ETF with stop loss 5%
capital to invest= 10% capital to risk / 5% stop loss on trade * 2,5% max loss on trade therefore capital to invest is 5% of tatol capital.
