Frosty's trading bot goes live part 2

frostengine · Jul 12, 2007

Its not that I find they don't produce enough returns. In almost every test they produce WORSE draw downs. Granted the reason could be my entries are not so great. But for my current system stops are a big mistake. Not only in terms of amount gained but also in the draw downs.

This goes against the prevailing belief about stops, but that is what my research has shown.

walterjennings · Jul 12, 2007

Quote from frostengine:

Its not that I find they don't produce enough returns. In almost every test they produce WORSE draw downs. Granted the reason could be my entries are not so great. But for my current system stops are a big mistake. Not only in terms of amount gained but also in the draw downs.

This goes against the prevailing belief about stops, but that is what my research has shown.

surprising. by drawdown you mean daily total right? a stop added to your current system should not result in any greater loss on a specific trade.

frostengine · Jul 12, 2007

nkhoi,

I have experimented with that type of thinking, but the way I implemented it was to make the bars range based instead of time. Also tried using volume. Never had any real good success using either method... however maybe there is another way of doing it.

originalskunk · Jul 12, 2007

Quote from frostengine:

This goes against the prevailing belief about stops, but that is what my research has shown.

Not really. I think most research comes to the same conclusion.

nkhoi · Jul 12, 2007

Quote from frostengine:

nkhoi,

I have experimented with that type of thinking, but the way I implemented it was to make the bars range based instead of time. Also tried using volume. Never had any real good success using either method... however maybe there is another way of doing it.

I am thinking more of if bar value > some ATR then increase bar count by some value, etc.. just to push the bot into faster react mode.

GTS · Jul 12, 2007

Quote from walterjennings:

surprising. by drawdown you mean daily total right? a stop added to your current system should not result in any greater loss on a specific trade.

Why do you say that?

Say you have a trade that goes against you for a loss of $1000 before it turns around and ends up positive +$300.

If you implement a stop of less then $1000 then you take the loss rather then ending up with a gain.

It is quite easy to end up with a (greater) loss on a specific trade by implementing a stop.

brownsfan019 · Jul 12, 2007

Quote from Businessman:

Frosties engine does not always catch these trend days.

It is possible to develop a profitable intra day system that misses these types of days.

Calling such a system a trend following system however is probably not appropriate.

You can develop systems to catch these types of trend day, but the trader may not like the drawdown profile or the win rate may be too low for the traders liking.

I agree 100%. I'm just saying that to call it trend following on an incredibly bullish day is not adding up. When I first read that frosty designed it to be trend following, then a day like today, should have made thousands and thousands. It *should* have gone long all day once it recognized the trend. It obviously did not.

That's not to say the bot won't work, I'm simply saying this is obviously not following the dominant trend. Even one as blatant as today.

trader225 · Jul 12, 2007

Quote from frostengine:

nkhoi,

You hit it on the head. The bot has a HUGE problem dealing with sudden moving markets. It needs several bars of data before it can realize that something is wrong..... sometimes it is deep in a hole by time it has enough data to make an informed decision.

So it does bar trading? Waits for some bar(s) to completely form before deciding? I don't think this can handle volatility very well. I don't think it's a very good method for anyone, except that it may be something that intro. daytrading books present because it's easy to explain and understand.

Trading the bars ... . People do realize that bar(s) are formed at arbitrary times, don't they? If they trade 5-min bars that are complete at 09:30, 09:35, etc., they may well get a different result than when trading bar(s) that are complete at 09:31, 09:36, etc.! Who came up with this magic time when bar(s) "complete?" I don't think the market respects anybody's "bar trading" methods.

GTS · Jul 12, 2007

Quote from brownsfan019:

I agree 100%. I'm just saying that to call it trend following on an incredibly bullish day is not adding up. When I first read that frosty designed it to be trend following, then a day like today, should have made thousands and thousands. It *should* have gone long all day once it recognized the trend. It obviously did not.

Thousands and thousands? Trading one ER2 contract?

The daily range on ER2 was not that much (nothing like YM) and it did have that morning downtrend and a fake reversal in the afternoon - given that I don't see how its reasonable to expect the bot to have been long the whole day.

Just keeping it real here.

nonlinear5 · Jul 12, 2007

Quote from trader225:

I don't think the market respects anybody's "bar trading" methods.

That's true, but at the same time, the market doesn't respect anybody's "tick trading" methods, either. My bot can evaluate the price action and execute literally hundreds of trades per second, but it wouldn't make any money. On the other hand, the same bot trading the end of the 5-minute bars produces a nice equity curve.

As many people pointed out here, some ideas (such as using stops) seem obvious, until you run them through a backtest. Sometimes it's amazing to see what works and what doesn't. For example, none of the simple moving average crossover systems (generating at least 1 trade per day on average from 1998 to 2006 trading ES) make any meaningful profit. Earlier in the thread, Frost also made that observation.