Dear JForex Contestants!

In accordance with the preliminary results of the JForex Contest Strategy in October, we could see that the most strategies became more efficient. However, the success of several strategies is still based on a few positive trades only.

Starting from November, the Trading Rule 6.3 is changed as follows:

The maximum profit on a single position must be below 25% of the total profit of all positive trades. If not, Dukascopy reserves the right to decrease the total balance of the account for an amount equal to the profit of the position and apply the rule recursively.

If we would apply the new requirement to October, only 2 strategies would fail to comply with it. Please see the image below:




The purpose of the toughening is to motivate participants to trade with as equal profits as possible, avoid random gambling and increase the quality of strategies which are used in the contest. Now you have to use your analytical thinking in order to plan the right trade amount, targets and exit points. It’s apparent that you need to make more profitable trades to win in the contest which becomes less a question of random luck.

Please also note that every contestant is allowed to re-register in the contest till November 15th.

If you’d like to change your strategy, please send your email request to [email protected] and we will remove your old registration. You’ll be able to apply for the new contest account again and start trading with a new strategy.

how do you count partial closes ? are those counted like separate trades (I think they should... as long as the close prices are not all the same of course). For example I have a strategy that closes positions progressively, to manage risk as the trade develops and hang on to profit as they are made.

If you just count one trade on one position that closed on several parts, that is penalizing even though the idea is to precisely mitigate risk and make profits more even.

You can make as many partial closes as you plan. But all closing and opening trades belong to one position and the total profit on this position must comply with the above mentioned rule.

hi
i Changed some (3 in 28) stop-loss orders manually Is this excludes me from the contest or
or these orders will not be counted.

Yes, but unless the amount of these orders is below 10%. If it's more, then we have to disqualify the strategy as it will already be a manual trading and not automatic.

hi
when i run my strategy can i put a new lot size and SL different
to the one i put it in the text of strategy

Ok seems good to avoid gambling but in some cases it could be painful even for good strategies...

For exemple :

1st
If after an amount of pips you trail the stop to breakeven + a few pips and the market turns
against you a few times there is a risk that your trades that close at take profit don't comply
to the 25% rule. In this case you better have to close at minus 1 pip in order that your
trades don't decrease the average of the "winning ones", it's a nonsense...

2nd
The strategy takes a trade but after a time there is not enough momentum in the
market and prices are just going sideways. What a "real" trader would do, he would
close his position even if the profit is only a few pips because there is a 50%
chance the market will reverse and he don't want to gamble at fifty-fifty.

Better making a few pips than a big loss...

You can see two practical examples where managing risk could kick you out of the contest.

The rule was intended for blocking gambling strategies from the contest. Sometimes it can happen that a good strategy can also have a big P&L on one trade. And we don't mind against normal strategies based on the market analysis and prediction.

We are keen on strategies which make one-two random trades and then switch to hibernate mode. When the strategy is made to evade the contest rules, we can see it clearly from the code. So, unless your strategy has a normal algorithm which aims to find the most effective trades according to its logic, we will not disqualify or cancel any trades. On the other hand, you'll always get an email from us with the information on trades that we don't like. So, you will have a chance to protect your strategy and explain why these trades were opened. We are not monsters at all. You can talk and write emails to us.

We will always be taking into account the behaviour of the strategy on the contest. Here are some signs and features we would really like to see in the strategy:

- you have a logical algorithm in your strategy. It can be based on indicators, patterns, statistical rules.
- your strategy does not have any equity targets (like to make 200 000 usd and stop trading till the end of the month)
- you keep you strategy running as long as possible without interruptions (like turning it on and off many times). Launch it once and leave it untouched. It's a robot contest .. not human after all. right?
- the back-testing result is the same or similar to your contest performance
- your script is clear and readable. You use some basic Java naming convention rules to make the strategy more readable for everyone. Lets speak Java!
- all comments are written not in a secret extraterrestrial language but in English, so that we could better understand it

I was afraid you was

Thank you for your post, very much appreciated. The contest goes in a good direction.

The last drawback I see right now is that we have to use to much leverage if we want
to be in the "10 winners". This causes some strategies to win one month and to blow out
the account the month after. For me, this is the only gambling left in the contest.

Maybe you should impose a maximum relative drawdown for the month to avoid strategies loosing
50% of the account before winning a 200%... no human would support that in a real
trading activity. The aim of such a contest is to show how automated trading could
simplify traders every days life and avoid him to die from a heart attack young

Greetings

I think that the with the gambling approach is that we can not decide if a strategy is "gambler's" or not with one month performance report only. This come from the definition of this type pf strategies - this month great, past months bad, next months bad too.

I have an idea for new rule: you can make your strategy as you want, but the average backtest results of this strategy for the past 3 (6,9 or 12) months have to be at least 50% (or 70%) of your current month result. This will mean that your strategy is covered.

Maybe this will lead smaller profits in the contest results, but the strategies will be more reliable.
For now if somebody uses profit targets, the others have to do it too in order to have same chances to win a prize.

Thanks!

Hi xxtazzz,
the idea of using the backtesting in the Contest sounds interesting and could open a new scenario. A good strategy performs almost in the same way in real-time and in the backtesting mode, so there could be a strategy Contest based exclusivly on backteting result. Every month there could be a Contest based on results of backtesting of last 3 (6 or 12) months including the current month, in this way there will be a monthly Contest but with a duration longer than one month, a gambling strategy doesn't survive on more months.

What do you, Contest Support, think about this ?

Best regards,
Frank

The idea to use the Historical Tester in order to verify the strategies makes sense. Actually, you can see this idea in the list of recommendations we've posted earlier in the topic.
However, it can be a tough rule for the strategies. Even too tough! In case if you were running the strategy the first 10 days of the month but the testing will be made for 30 days, the discrepancy can be huge in the performance.



That's correct. As you said - almost! Some trades might be rejected due to a connection latency while the tester will give you an execution. But to have a contest only in the back-testing mode sounds too boring. Isn't it? Because we don't see the racing and the competition!

Yes, comparing the same month real-time/backtest results won't bring good information.
But if a strategy pretends it's not gambler's, it won't be a problem to look at the previous months.
We take all the, for example, 11 past months and calculate an average profit.
This will be average result and would be decreased by some type of handicap.
No matter how much time your strategy was running this month, no matter rejected orders or so...
If you're pretending that your strategy made this result "not accidentally" - just prove it.

Thanks!

Well, thank you for a valuable idea! Indeed it can be a good rule. We can impose new restrictions on the strategies - the Maximum Daily Draw-Down and the Total Stop Loss level.

The Maximum Daily Draw-Down is expressed in the absolute value of the daily loss. Once the daily DD is reached, all positions are closed and the account is blocked till the settlement. At the settlement time, the account is reactivated.

The Stop Loss level is set as a level of equity when all positions are closed and the account is blocked for further trading.

What values we should apply for the both parameters? What is your opinion on this questions?

If we will use these 2 parameters in together, it will improve the quality of the strategies and participants will start paying more attention by trying to limit the exposure.

I'm a tad bit alarmed by some of these suggestions. Please don't complicate things to the point where the fun is lost. As long as it's simple and us programming newbies can have a chance to win I think the contest will be succesful and grow.

Here are some thoughts on the recent ideas:

The maximum daily draw-down: "Day" is an arbitrary time period. A trading set-up that you act on is a trading set-up that you act on, regardless if it comes right after a loss just minutes before, or if you have 100 good trades in a row. The quality of the strategy and the quality of the set-up has nothing to do with how much time passed since the last bad trade. We shouldn't force our shortcomings as humans on our automated strategies.

In addition, making it an absolute number would mean an account that has lost money initially can leverage the entire account, while a strategy that has made a lot of profit could be closed down for the day based on a loss that is not that big relative to the account.

If the goal is to limit leverage this rule would imo be a very complicated way to do it. It's likely people would just use it as an "outside stop-loss" and leverage away as usual - because a trade that goes the right way would never be affected by this rule. Gambling still pays.

The stop-loss level: There already is a maximum drawdown: 100%. To limit drawdown to any other percentage would mean the same thing, it would just be a new number. Sure, it would force leveraging numbers down towards levels that sound more realistic, but it would still encourage participants to max out their exposure relative to the new drawdown. I.e., the number of bad trades a contest account can stand wouldn't change, so it wouldn't do anything to discourage gambling.

To put it another way: Having a 30% stop-loss/drawdown in a 100k account is the same as a 100% stop-loss/drawdown in a 30k account. You can lose 30k, that's it, then you're done. It's 100% of 30k any way you twist it. And there already is a 100% limit used. This is always going to be the same no matter where the line is drawn.

With the above in mind I have one question: If the goal really is to more closely imitate live automated trading, is there some reason why we couldn't limit the use of leverage instead?

There might be some drawbacks, but at least it's simple and makes intuitive sense. Some strategies are always punished no matter how the rules are written.

I agree that decreasing the max leverage will soften the "gambling questions", but won't solve it. It's true that if you decrease the maximal loss to 30%, it will be the same as 100% with smaller account.
If we think in this direction, a gambling strategy needs more leverage than "normal" one, because the gambling strategy needs to close one or two big, fat positions with profit. In other hand the normal strategy doesn't need too high leverage in order to make the same month profit.
But the decreased leverage will be problem for the normal strategies when the strategy begins the with few loses, because initially it'll be near the max leverage and will have to decrease the exposure.

Maybe if there is "Maximum traded volume" it'll solve this problem. For example the max leverage is still same but you can not get exposure bigger than 1 000 000 (with initial 100 000). This way you could close few positions without need to reduce your initial volume. This will make the profits "not progressively growing", but it'll be like in real life.
And it'll make the rules for "one position only" and "max profit per trade" unnecessary.

What do you think?

Maximum dollar exposure at once = 1 000 000

Sounds good, I like it. From a competition perspective it makes sense to have the limit be a set number so that, as you say, you can have an initial loss and not immediately be forced to take smaller positions than those who start with a winner.

I somewhat suspect, or maybe I should say hope, that the "one position" rule is also there to keep the bar lower for newcomers, but yeah, with this kind of maximum exposure it wouldn't really matter. The need for the "profit limit" rule will definitely go away, which I think would be a good thing.

In general I think this would simplify the rules, and put more focus on entry/exit and less on taking on as much risk as possible.

We have doubts that the maximum profit and the single position rule will be removed. It will make the contest less effective. The maximum profit can be never equally replaced with any exposure limits. The strategy can still win on the basis of only one single trade, while the remaining 9 positions can be done with very small amounts.

The single position mode is required to make participants to have 10 different positions. If the requirement is removed, then you can simply open 10 positions at once and just wait when the market will turn in the right direction.

So, we will keep both rules unchanged but we'd like to make the rules more efficient in terms of money management.

I'll have to admit that I'm not following the logic here. The maximum exposure xxtazzz and me used as an example would be 1M dollars per trade. With the suggested exposure limit, could you explain how a strategy would win having only one profitable trade?