How to keep in faith with a statistically profitable system? | Page 3 | Forex Factory

Options

Go to
Permalinks
Print Thread

Similar Threads

Statistically Infallible 11 replies

Statistically testing every indicator with my Genetic Algoritm 132 replies

A statistically rare pattern/anomaly - reversed? 9 replies

Are my 6-mth-old results statistically valid? 17 replies

What Constitutes a Statistically Significant Sample for Back Testing? 4 replies

Attachments: How to keep in faith with a statistically profitable system?
Exit Attachments

How to keep in faith with a statistically profitable system?

  • #41
  • Jan 4, 2011 5:23pm Jan 4, 2011 5:23pm
Quoting Mr J
We don't. Even after losing 30 trades in a row, we're still only 90% likely to win our next trade. The numbers will never "balance" (i.e. experience 30 losses, but don't expect 270 winners to make up for it), but the larger our sample grows, the more likely it will start to reflect our edge. Over time, those 30 consecutive losses will become a memory.

That said, losing 30 in a row is beyond horribly unlucky, and it's far easier for a 50% system to do it than a 90% system.
Oh I don't know, if it lost 30 times in a row I wouldn't bet on the 31st one being a winner!

There are two possibilities -

1. It really does have a 90% chance of success and the rarest event ever has happened - i.e. something with a 10% chance of happening (in this case a loss) has just happened 30 times in a row.

Or...

2. It doesn't really have a 90% chance of a successful trade and you just think it does due to curve fitting, not testing it properly, the market changing or whatever.


I'd say option 2 is millions of times more likely to be the actual case

If a roulette wheel came up with the number 27, twenty times in a row what would you think? What a coincidence or the wheel is broken or rigged in some way? If a coin came up heads 200 times in a row would you believe it was a fair coin?
  • #42
  • Jan 4, 2011 5:59pm Jan 4, 2011 5:59pm
Quoting ha-pattern
You can replicate much of live's psychology in demo, with practice. Why invoke taxes? There's a world of room in paper, simulator and demo to hone one's technical edge before testing with pennies.
I don't mean psychology. And no thats not the same either.

I mean no matter what a broker tells you the feeds from demo to live do NOT end up with the same results.

The charts may look the same in retrospect but the results will end up different. I've tested demo against live trades side by side, with different results.

The demo trade magically works out and even a stop being barely missed and the trade working, while the live account always suffers more from a badly placed entry or just doesn't have the same luck.

Demo is a game with no traders or big boys playing and moving price. Nobody is stop hunting demo accounts.

I'm throughly convinced that all brokers that offer free demo's rig them to make live trading seem easier than it is to sucker in as many new small accounts as they can.

Look over any forex trading forum and you'll find skads of new traders who do awesome in demo and post something to the effect of, "is it really this easy."

Forward and backward testing in Demo to establish a W/L ratio is a COMPLETE waste of time. At best its only good to work out the nuts and bolts of whatever your trading plan may be.
  • #43
  • Jan 4, 2011 6:16pm Jan 4, 2011 6:16pm
Interesing thread, comparing coin flips with no memory to trading isn't entirly fair, maybe in a 100% mechanical system but lose 30 trades in a row and pyschology wise your screwed and can expect to make more and more losses or cut your profits so short they might aswell be losses as there tiny compared to a losing trade.

A system which might work with 90% accuracy in some market conditions can easily flip to 90% losers in other market conditions, so as others have said take 2-3 losses in a row then reduce your size or back to demo and retest your system to make sure it's still performing in the current conditions.

The odds on 30losers in a row at a 90% win rate is also billions to 1 most likely 10^30 = 1E+30

Even at 50/50 to 1 the odds on 30 are 1,073,741,824

Don't get me wrong I've had 30losers in a row, it's cause I SUCKED and my system had a 99% failure rate, the poster with 34 losers in a row the conditions didn't suite his system for a while, but he kept going till it returned to normal I guess.
Nothing to it, but to do it!!! Stick to the plan FOOL!!!!
  • #44
  • Jan 4, 2011 11:34pm Jan 4, 2011 11:34pm
Quoting Davidee
Thanks for the table, what was the formula you used to calculate it?
Formula is in attached XLS.
Attached File(s)
File Type: xls Probability of X consecutive losing trades.xls   30 KB | 395 downloads
  • #45
  • Jan 4, 2011 11:38pm Jan 4, 2011 11:38pm
Quoting narafa
What you can control sometimes depends partially on the outcome

You can control your position size, but the next couple of bars (The outcome) can lead you to change your position size accordingly.

So in a sense, you have to care about both.

Regards,

Nader
yes, my position size may change according to what happens next. but that has no bearing how i come to make a desicion to put on the trade or not.
  • #46
  • Edited 12:11pm Jan 5, 2011 12:07am | Edited 12:11pm
grkfx hits the nail on the head, IMO.

You must have an edge that's somehow grounded in either a understanding of the hows and whys that drive price movement, and/or the price patterns that are created as a consequence; or some other method based on robust abstracts, e.g. trend, S/R, momentum, volatility, correlation, economic trends and drivers.

Anything else, over a large enough sample size, is nothing more than an exercise hit or miss, and the transaction costs will eventually defeat you. Hence I agree with tdion: it's no better than gambling. I haven't been able to devise a profitable method from (solely) data mining/optimization; I'm not saying that it can't be done, but I believe it's not a trivial task. So many quant/EA methods are based on this type of approach, and the vast majority of them, given enough time, fail. It's so easy to write an EA that enters based on a combo of indys, and scales out using X pip TPs and/or Y pip trailing stops, none of which bears any relation to how the heavyweight players are driving price.

Many other EAs employ MM methods that do nothing more than re-balance return and risk, and/or redistribute wins and losses.

And it's all too easy to be fooled by randomness: fluctuations can be such that a system that returned an 80% win rate over a 500 trade test, can easily return a 40% win rate over the next 500 trades.

Marv recently started an interesting thread about how randomly generated charts apparently exhibit the same kind of trends, channels, double tops, pin bars, the respecting of trendlines and fibo levels etc, etc, just as real price charts do. Given that a randomly generated chart can not be beaten by ANY system, this similarly demonstrates that to be profitable, a system must exploit inefficiences on price charts that are in some way qualitatively different from apparently similar patterns on randomly generated charts.


Here's a horse racing analogy. The savvy punter knows that studying factors like the track condition, weight carried, barrier draw, and jockey's experience can give him a real edge. That equates to what grkfx calls "skill". The uninformed punter might (for example) run a historical test on saddle blanket color, and over a large data sample discover that the horse that wears a yellow saddle blanket wins more often on Thursdays, while one wearing a green saddle blanket wins more often on Tuesdays. The results of the test, whatever the statistical error, are irrelevant, simply because the underlying premise has no real bearing on the outcome.

Much the same applies to forex. Either the variables that are being tested somehow reflect the underlying market drivers, or they don't. If they don't, then - no matter how large or representative the sample size is - the trader is effectively making random entries and exits, and the spread will eventually defeat him, just like the zeros on the roulette wheel ensure that the house always beats the punter if the latter plays for long enough.

And on top of all this, the trader still has to get his psychology right.
  • #47
  • Jan 5, 2011 7:30am Jan 5, 2011 7:30am
Quoting hanover
Formula is in attached XLS.
Thanks very much
  • #48
  • Jan 5, 2011 7:31am Jan 5, 2011 7:31am
Quoting Turveyd
Interesing thread, comparing coin flips with no memory to trading isn't entirly fair, maybe in a 100% mechanical system but lose 30 trades in a row and pyschology wise your screwed and can expect to make more and more losses or cut your profits so short they might aswell be losses as there tiny compared to a losing trade.

A system which might work with 90% accuracy in some market conditions can easily flip to 90% losers in other market conditions, so as others have said take 2-3 losses in a row then reduce your size or back to demo and retest...
You managed to find a system with a 99% failure rate? Awesome, please share it. Be great to take the opposite side of those trades
  • #49
  • Jan 5, 2011 12:13pm Jan 5, 2011 12:13pm
Quoting Davidee
Be great to take the opposite side of those trades
Maybe; maybe not (see myth #1 here).
  • #50
  • Jan 5, 2011 1:01pm Jan 5, 2011 1:01pm
Quoting Davidee
You managed to find a system with a 99% failure rate? Awesome, please share it. Be great to take the opposite side of those trades
I can give you a simple system with 99%+ failure rate: click buy, then click sell IMMEDIATELY

Replace sell with close in a non-FIFO account.

But you are right in principle: any 99% failure rate system which doesn't lose strictly because of the spread would be a gold mine, at least until the market changes again and it starts losing only 50% of the time
  • #51
  • Jan 5, 2011 3:01pm Jan 5, 2011 3:01pm
Quoting grkfx

So either I like to have a trading system that works in all markets, OR I can identify when the market is about to enter the market conditions that produce losses so I can stay out during that time. Then develop a way to determine when the market will enter the conditions that produce profits. It can be difficult, but doable.
I have to agree with grkfx. Tdion, you are saying you have a system that works in some market conditions and not others. Well, you need to incorporate rules into your system that make sure you are only trading when conditions are good for whatever type of system you are using.

I have a completely systematic way of trading; I take every single trade that meets all of my rules, and several of those rules keep me out during specific periods or when conditions (volatility, etc.) exist that have proven bad for my system in the past.

Of course market conditions can always change with time, but you need to think about what never changes, what always happens in the market. My system is based of concepts that will never and have never changed, unless of course the world adopts a single currency and trading altogether becomes nonexistent. Anyway, I'm not here to talk about my personal strategy, but I am just tired of people saying you can't successfully trade the spot fx market, or that it is just gambling, because info like that was discouraging to me when I started, and now I know it is possible to be successful.

Anyway, all I'm saying is you can adopt the mentality that it is all a crap-shoot and random, and no one can ever win, but I think that's a self-limiting mentality that will hurt your ability to be open-minded and find profitable ways to trade the markets. Of course a mentality like that makes you feel good that you "know" it's impossible.. but wouldn't you rather have a full bank account?

I respect your opinions and am not trying to bash, I just want to offer a different perspective.
  • #52
  • Jan 5, 2011 4:37pm Jan 5, 2011 4:37pm
Quoting Davidee
You managed to find a system with a 99% failure rate? Awesome, please share it. Be great to take the opposite side of those trades

It's a fairly easy system, but impossible to reverse.

Basically, get in, set a tight SL and greedily play for 200+ pips, 99% of the time your SL will hit or your money back.
Nothing to it, but to do it!!! Stick to the plan FOOL!!!!
  • #53
  • Jan 5, 2011 7:01pm Jan 5, 2011 7:01pm
Quoting hanover
grkfx hits the nail on the head, IMO.

You must have an edge that's somehow grounded in either a understanding of the hows and whys that drive price movement, and/or the price patterns that are created as a consequence; or some other method based on robust abstracts, e.g. trend, S/R, momentum, volatility, correlation, economic trends and drivers.

Anything else, over a large enough sample size, is nothing more than an exercise hit or miss, and the transaction costs will eventually defeat you. Hence I agree with tdion: it's no better than gambling....
I like your horse racing analogy. I have spent years of effort on historical test on saddle blanket color, which we were told that it is important and key to the profitability. I found nothing actually useful until I eventually notice that I was leaded to a wrong direction. Data mining/optimization is useful, provided that if you are doing it on track condition, weight carried, barrier draw, and jockey's experience.
  • #54
  • Sep 17, 2013 4:41pm Sep 17, 2013 4:41pm
I disagree with the horse racing analogy, it assumes that data-mining has to be dumb and can have no insights into causality. You can be successful with a pure data-mining approach (I have 15+ years of experience doing this)

The problem is that you need to find ways to distinguish correlation from causality. There are clear and systematic ways of doing this by analysing the statistics of the strategies you generate.
  • #55
  • Sep 25, 2013 3:10pm Sep 25, 2013 3:10pm
Quoting erik1
I disagree with the horse racing analogy, it assumes that data-mining has to be dumb and can have no insights into causality. You can be successful with a pure data-mining approach (I have 15+ years of experience doing this) The problem is that you need to find ways to distinguish correlation from causality. There are clear and systematic ways of doing this by analysing the statistics of the strategies you generate.
Oh yea. This will be something that I would like a lot to learn about! [The clear and systematic ways of doing this...]

Can you suggest some reading?

Interesting posts you're spreading around, BTW!
πάντα δοκιμάζετε τὸ καλὸν κατέχετε
  • #56
  • Sep 25, 2013 8:15pm Sep 25, 2013 8:15pm
Quote
Can you suggest some reading?

If you already have good knowledge about basic statistics I would suggest researching about data-mining on synthetic data (artificial data created with the statistical properties of a given financial time series), that should keep you busy for a few years. Good luck.
  • #57
  • Sep 26, 2013 12:11am Sep 26, 2013 12:11am
Quoting erik1
{quote} If you already have good knowledge about basic statistics I would suggest researching about data-mining on synthetic data (artificial data created with the statistical properties of a given financial time series), that should keep you busy for a few years. Good luck.
Thanks, mate, to have pointed to the right direction.
I think I perfectly understand what you mean, and, if your "best whishes" do not realize as crude as you spell them out, I also believe it could take a bit less than a few years.

Bye

BL
πάντα δοκιμάζετε τὸ καλὸν κατέχετε
  • #58
  • Feb 4, 2017 7:26pm Feb 4, 2017 7:26pm
Quoting in_drag88
Dear fellow traders, i have a question stuck long enough and still can't figure it out by myself, maybe i didn't learn enough about probability math. This question is appointed for any quants, please help me with this Here we go : Let's say we have test our system for once, say it 100 trades, and we got 90 winning trades and 10 losing trades (so we have simply 90% winning and 10% losing in percentage terms) Then, we run the system in the live account, and suddenly,we have like 30 consecutive losses. According to our system it was 10% of failures,...
How to keep the faith when there are large number of losing trades in a row with a mathematically profitable system?

When the losing trade's losses are just a small fraction of the winning trade's profits. For example your 100 straight losing trades lose 5 pips each time but your 1 winning trades win 1000 pips or more. That is how you keep the faith. When you can figure out how to make money with low win rates and high consecutive losses you are at a higher level of trading. Focus on making money not win rates.
You came here for the truth and let me unveil it for you
  • #59
  • Last Post: Feb 4, 2017 8:26pm Feb 4, 2017 8:26pm
Attached Image (click to enlarge)
Click to Enlarge Name: risk-management-strategy-concept-image-letter-blocks-forming-crossword-words-dice-white-backgrou.jpg Size: 79 KB
Attached Image (click to enlarge)
Click to Enlarge Name: sourcing strategy model.png Size: 51 KB
Attached Image (click to enlarge)
Click to Enlarge Name: Arrow-and-Maze.jpg Size: 765 KB
0 traders viewing now