Frosty's trading bot goes live part 2

Hi Frosty,

I followed your first bot thread with interest, likewise this one. I wish you well with your trading bot. With that in mind a few thoughts and questions....

Based on the equity curve generated from the bots’ trading results it looks as though you have a favourable strategy for the duration of that simulation. I think that these results were generated by the bot based on data from August ’06 to mid May ’07. From your posting it looks like the equity curve is based on about 525 trades which is a good sample size in terms of numbers of trades - though maybe not long enough to face the likely range of market conditions that your bot will at some point encounter…..

If your bot hasn’t changed then the likely cause is a change in market behaviour / conditions. It’s pretty clear that you have been seriously spooked by the bot giving a set of results that has blown through your Max Loss ….maybe the live running period Ave Loss has also regularly been below the Ave Loss for the equity curve simulation period, likewise the Ave Win for the live running period was significantly lower as well?

Why? Again, if your bot has not changed then the only thing that can have (realistically) changed is the way your bot responds these ‘new’ market conditions. If this is the case the truth is that you really don’t know what your Max Drawdown and Max Loss (is/was) could be ….perhaps you have already had it perhaps not. Given this uncertainty I can understand why you are thinking of switching it off.

How long will these new conditions persist for? I don’t know what they are so I’ve absolutely no idea nor do I know your bot determines ‘trend’ so I can’t even give a guess.

As a general comment, for me, looking at a daily chart of ER2 for the simulation period versus the live running period there does seem to have been a change in the ER2 market (basically from upward to lateral). I may well be wrong of course and even if I’m right it may not be the real cause of the change in results.

I presume that your starting capital for the equity curve simulation is the same as your starting capital a few weeks back?

You can be sure of one thing…a similar set of market conditions to the simulation period will return and deliver profitable results…it’s a question of when.

It’s then a question of will you have sufficient trading capital left to take advantage of this. Well, in part, that really depends on the maximum the bot risks per trade…..

Best to your trading,
JR
 
Another pretty good day. Still deep in the hole at this point. But at least the 2 days I ran it live this week it made me some money. Should have ran it all 4 days this week, because all 4 would have made money. It has not had any big winning days lately,so I can not really say its starting to pick up, but its at least more in sync with the market this week than it has been the previous two. Earning season is normally the best time for my strategy, so hopefully as that starts to kick off the bot will get its groove back. Just happy to have had a GREEN WEEK. Todays results: +$245.60

Today +$245.60
Week to Date: +466.4
Since going live: -$2913.2
Total days: 12
Avg per day: -$242.76

Up Days: 6
Down Days: 6
 
Quote from frostengine:

Another pretty good day. Still deep in the hole at this point. But at least the 2 days I ran it live this week it made me some money. Should have ran it all 4 days this week, because all 4 would have made money. It has not had any big winning days lately,so I can not really say its starting to pick up, but its at least more in sync with the market this week than it has been the previous two. Earning season is normally the best time for my strategy, so hopefully as that starts to kick off the bot will get its groove back. Just happy to have had a GREEN WEEK. Todays results: +$245.60

Today +$245.60
Week to Date: +466.4
Since going live: -$2913.2
Total days: 12
Avg per day: -$242.76

Up Days: 6
Down Days: 6
More...

I know this is obvious, but something is still askew... 6up 6down days, yet cumulative since going live equates to 12 down days, even with the green week.

All the best,
Osorico :)
 
I have traded it live for 2 full weeks, and only 2 days out of this week. Therefore I have had 12 total trading days. 6 of those days up, 6 of those days down.
 
frost - I think osorico is pointing out how the end result in terms of CASH is adding up here. Your winning days are not offsetting, let alone taking you into profit, from your down days. That should be a concern in my opinion. Just winning is not enough if one losing day takes out 2-3 winning days.
 
Quote from brownsfan019:

That should be a concern in my opinion. Just winning is not enough if one losing day takes out 2-3 winning days.
More...

It's obviously not what you want to see when you turn your system on, but it's still only 12 days of data. Not enough to make decisions one way or the other IMO. I'm assuming the profit factor/expectancy observed in backtesting was good or the system would not be online now.
 
Part of my problem is I have had 2 days where I have lost over 1k each of those days. These events have happened in the past, but are generally VERY Rare.. but they do happen. I have had the misfortune of them happening TWICE since being live. Those outliers have skewed the results thus far. Also skewing the results is the day where I lost a good bit due to an internet outage. So thats 3 events out of 12 days that are severly messing up the numbers..... hopefully things return to normal and the bot climbs to profitability over the next few weeks..... if these events continue to happen then I'll have to reevaluate what i'm doing...
 
frost

i'd recommend you to place your code at the broker's servers for optimal performance and max reliability in terms execution. an outage like the one you had might ruin an otherwise profitable strategy and wipe you out in a blink of an eye. i have been there.

if your broker doesn't offer that option, you can always rent a dedicated server or lease server space at a leading data center.

my recommendation goes for testing systems as well. at least 50% of the difference between the backtesting and the real trading results is based on technical glitches and latency.

i have seen your other thread a couple of months ago and it looks like you're now revisiting the same problems. i am not sure whether you code really has an edge or not, probably you neither. that's ok. from what i have been reading here, it looks like you really have a good system but still lack proper management. i am pretty sure you cant evaluate your live performance without having the system either at broker premises or run in professional data center. i say this by experience. a disaster could happen anyhow but when we increase the risk factors, we're eroding our eventual edge and increasing exponentially our risk exposure.

considering that the markets are extremely competitive, it becomes evident that the technical infra structure put in place to run the system has to be at least professional. glitches like "isp went down", "power outage", "vista sucks" or added slippage due to high latency are system killers these days.

good luck
 
Quote from frostengine:

Part of my problem is I have had 2 days where I have lost over 1k each of those days. These events have happened in the past, but are generally VERY Rare.. but they do happen. I have had the misfortune of them happening TWICE since being live. Those outliers have skewed the results thus far. Also skewing the results is the day where I lost a good bit due to an internet outage. So thats 3 events out of 12 days that are severly messing up the numbers..... hopefully things return to normal and the bot climbs to profitability over the next few weeks..... if these events continue to happen then I'll have to reevaluate what i'm doing...
More...

you may also notice that looking at your new equity curve, the last 20% is not nearly as smooth as the rest and suffers from much higher draw downs. im not sure what time period the last 20% of the curve reflects but it might say something about the edge of your system in the near term market.
 
Hi Frosty,

Good to see you have had a couple of better days. Couple of thoughts on that Equity Curve….

Thing is we all keep referring to it. There’s no doubt as presented it is highly profitable and suggests a real edge for the simulation period. That’s probably the main reason many posters on this thread suggest you keep going despite the tough time you’ve had since you started a couple of weeks back.

I don’t have a view on whether you should turn the bot off as I just don’t have enough info on the bots’ methods, strategy and risk management. However like most I’d say that equity curve does look profitable. Having read the last few posts though I do have a few questions about that curve ….

For example: 1) Does the equity curve take into account estimated losses owing to technical problems? 2) Does it take into account slippage? 3) Does it take into effect commission?

These are real life trading costs. If you have already estimated and factored in these ‘costs’ then don’t waste any more time reading this post because you already have a realistic equity curve.

If not, particularly 1 & 2, then the equity curve may not be realistic and is probably best case. You might find it helpful to check the impact on your existing equity curve by estimating ‘costs’ in each of these three categories and re-drawing a ‘Realistic’ Equity Curve

For example:
- Technical Glitches…..if you have e.g. one ISP outage every 15 days and you loose money when this happens factor that in. For a ‘realistic’ equity curve deduct your Ave Loss from your $ result every 15th day. Choose whatever rules you wish to use based on experience but generate ‘costs’ due to technical glitches.

- Slippage……you will not get the expected entry price every time if you use MKT or STP orders on ER2 - slippage will apply. Factor it in. Make a few assumptions based on experience. Realistic Case Curve say, e.g. 3 Tics lost on every 3 round turns. Choose whatever rules you wish, but generate slippage ‘costs’. Whatever they are deduct them from your daily $ result.

- Commission …. Simpler cause you know what your commission is and it will be the same for each trade, again, deduct it from your $ result for the day.

Choose whatever rules you want - the point is that you make some estimate of these costs and deduct them to give you a more realistic equity curve. This would help you decide on how strong an edge you realistically have based on the simulation period.

As always, best to your trading,
JR
 
You must log in or register to reply here.
Back
Top