Attack of the Robots - An Algo Journal

kevinkdog · June 26th, 2020, 08:02 AM

vmodus

Go Live
For the most part, slippage is almost a non-factor due to the use of stop orders (I have some positive slippage, too), though I always include it in my testing.

Nice progress, great to see!

I'm curious about your statement. Stop orders are really just market orders as far as the exchange goes. I have always seen slippage with them.

Maybe you are calculating slippage differently?

I calculate slippage as the difference between my actual fill and the strategy engine fill.

Some of my results (all market or stop orders):

For MES, last couple of weeks

I assume $1.25 per side slippage for MES

Average actual slippage based on 272 trades: $0.81 (or 65% of planned slippage - which is good!)

22% of orders had positive slippage
5% had $0 slippage
73% had negative slippage, as much as 8 ticks

Maybe I am doing something differently than you?

June 26th, 2020, 01:14 PM

kevinkdog

Nice progress, great to see!

I'm curious about your statement. Stop orders are really just market orders as far as the exchange goes. I have always seen slippage with them.

Maybe you are calculating slippage differently?

I calculate slippage as the difference between my actual fill and the strategy engine fill.

Some of my results (all market or stop orders):

For MES, last couple of weeks

I assume $1.25 per side slippage for MES

Average actual slippage based on 272 trades: $0.81 (or 65% of planned slippage - which is good!)

22% of orders had positive slippage
5% had $0 slippage
73% had negative slippage, as much as 8 ticks

Maybe I am doing something differently than you?

Thanks! It is nice to have my first strategy up and running, though I still consider this to be a live incubating period, thus the use of micros instead of e-minis.

I think we calculate the same, but for example, on a reversal entry I am seeing this (sorry no screenshot, Trade Manager is acting up and cannot get any data):

Buy to cover: 3005.50 (exit)
Buy: 3005.5 (entry)

At least, that is what I saw yesterday. I will try to get a screenshot of the trade list after we've exited our positions for the week. Regardless, we always add at least one point value in slippage, same as you. For this week, it has been slightly in our favor, but we haven't combed through all the data yet.

Considering I have just started using stop orders (in TradeStation): what is the purpose of setting a stop price if they just become market orders? Entering at stop versus market has a big impact (positive) on strategy performance. I'm guessing it is just when the strategy will send the order to TS.

~vmodus

kevinkdog · June 26th, 2020, 03:12 PM

vmodus

Thanks! It is nice to have my first strategy up and running, though I still consider this to be a live incubating period, thus the use of micros instead of e-minis.

I think we calculate the same, but for example, on a reversal entry I am seeing this (sorry no screenshot, Trade Manager is acting up and cannot get any data):

Buy to cover: 3005.50 (exit)
Buy: 3005.5 (entry)

At least, that is what I saw yesterday. I will try to get a screenshot of the trade list after we've exited our positions for the week. Regardless, we always add at least one point value in slippage, same as you. For this week, it has been slightly in our favor, but we haven't combed through all the data yet.

Considering I have just started using stop orders (in TradeStation): what is the purpose of setting a stop price if they just become market orders? Entering at stop versus market has a big impact (positive) on strategy performance. I'm guessing it is just when the strategy will send the order to TS.

~vmodus

So what happens with a stop order is that as soon as that price is hit, a market order is immediately sent, and is filled at the prevailing price. So it doesn't guarantee any kind of price like a limit order does. In fact, depending on the situation, your stop order may not be filled at all. That is rare though, thankfully.

I have seen stop orders executed as many as 10 points from the the actual stop price.

June 26th, 2020, 03:28 PM

kevinkdog

So what happens with a stop order is that as soon as that price is hit, a market order is immediately sent, and is filled at the prevailing price. So it doesn't guarantee any kind of price like a limit order does. In fact, depending on the situation, your stop order may not be filled at all. That is rare though, thankfully.

I have seen stop orders executed as many as 10 points from the the actual stop price.

Ah, okay that makes perfect sense, thank you so much! Have a great weekend, my friend!

~vmodus

June 26th, 2020, 05:21 PM

End of Week
This was a pretty wild week for us, in a good way I think. We were able to run our new strategy and it performed well on the micro indexes. We have a couple of variants that are currently in the testing phase, so that is exciting too. As I mentioned yesterday, I don't post actual P&L, but will indicate if I was up or down for the day, and overall for the week.

Today's Results - live

Positive P&L at close, trading MES and MNQ (1 contract each)
No open positions at close (I don't hold over the weekend)
26 round trip trades
38% profitable
Overall profit factor 1.16
System efficiency: -13% (worse than testing)
Execution errors: one

I'm not terribly happy about the execution error, which probably cost $50 or so. I was carrying two contracts of MNQ because I misinterpreted a message from TradeStation and probably caused a second order to be placed. I will review and correct.

Efficiency and profitability are not where I want them, but still profitable despite that. I will continue to monitor.

Two things I didn't like:

TradeStation flaked out and I could not get accurate data from the Trade Manager, nor do Trade Analysis. I couldn't do anything until market close when I was able to restart my PC. Restarting TradeStation did not resolve the issue.
TradeStation also would not allow me to cancel a limit order for my end of week exit. It told me the Trade Manager was locked. BS, I say.

~ + ~ + ~ + ~ + ~ + ~ + ~ + ~ + ~ + ~ + ~ + ~ + ~ + ~ + ~ + ~ + ~

Weekly Result

Positive P&L over 3 trading days; all days were +
53 round trip trades
44% profitable
Overall profit factor 1.85
System efficiency: 0% (better than testing)

So all-in-all, a good week. I see some potential system improvements, but it has proven to be tradable thus far.

Holiday Analysis
This weekend is a holiday weekend, for all intents and purposes, though unlike any holiday weekend we've seen in ages. Some of our past strategies have typically avoided trading holidays. I did some analysis on my latest strategy against historic holidays and it does well enough to run. I will leave this running through the Independence Day holiday (July 4th).

~ + ~ + ~ + ~ + ~ + ~ + ~ + ~ + ~ + ~ + ~ + ~ + ~ + ~ + ~ + ~ + ~

That's all for the week. More trading next week. Have a great weekend everyone and please be safe!

~vmodus

June 29th, 2020, 09:06 PM

I only have a few minutes, so I figured I would give a quick update:

Today's Results - live

Negative P&L at close, trading MES and MNQ (1 contract each)
2 open positions at close, both positive
26 round trip trades
19% profitable
Overall profit factor .34
System efficiency: -32%
Execution errors: several
System errors: several

This was my first day of losses with this system. Tracking the system, it would have been one of those losing days even without any errors.

I had a couple of execution (human) errors. I started my system exactly at open, to see what would happen. That cost me two bad trades immediately. I found I need for my signals to normalize, especially when we have a gap over the weekend. The other execution error was when I had to restart, I forgot to restart the automation on MES (missed two orders).

Then I had several bad (phantom) trades overnight, which were system errors. I have to determine how this happened and how I can prevent it. I also had a system lockup at about 945 ET (great timing, huh?), which required a reboot..... a VPS is looking better every day.

If something could have gone wrong, from my end, it did. Still, I was able to recoup some of the earlier losses. With my current open positions, I should be in good shape to end the month profitable, which is good considering I will only have had 6 trading days.

~ + ~ + ~ + ~ + ~ + ~ + ~ + ~ + ~ + ~ + ~ + ~ + ~ + ~ + ~ + ~ + ~

My overnight problem
So I had issues overnight, but everything ran fine while I was awake and monitoring. Now I'm thinking that maybe I just want to run this while I am able to monitor my positions. I would just confirm that my entries are occurring when and where they should, as well as reversals and exits. So I decided to check to see where the sweet spot of my strategy is, in other words: when does it make the most money? I dropped a sample of over 1700 trades into Excel, did a pivot table based on entry hour and summed the total for each hour, and then did a graph:

MNQ - 1700 Trades

So my sweet spot is 800 ET to 2100 ET, hours I am typically awake and active. So rather than run this 23/5, I can run this 13/5, which would account for 88% of my profits for this period, at least in this example.

~ + ~ + ~ + ~ + ~ + ~ + ~ + ~ + ~ + ~ + ~ + ~ + ~ + ~ + ~ + ~ + ~

Learning
I have learned some things about this system:

Wait 15 minutes after the Sunday open to start the system, to allow the data to normalize
Exit on Friday should right after 1615 ET for US equity indexes (ES, YM, NQ, RTY)
I don't need to trade this overnight
The alligator 'teeth' may be keeping me out of good trades

With this information, I can improve future versions of this. I am still determining if this strategy will be effective long-term, but the offspring show lots of potential.

~ + ~ + ~ + ~ + ~ + ~ + ~ + ~ + ~ + ~ + ~ + ~ + ~ + ~ + ~ + ~ + ~

That's all for tonight. See you all tomorrow!

~vmodus

June 30th, 2020, 10:38 AM

Preventing Duplicate Fills in TradeStation or MultiCharts
I forgot to mention one other thing I did yesterday, which was helping a fellow member here with a little problem. He was trying to prevent duplicate fills on a strategy, which occurs when a strategy condition is met more than once before the order has been filled. If you use PowerLanguage or EasyLanguage, check it out:

Getting duplicate fills in automated strategy

Okay, this was a little easier than I thought. So here is how it works, plain English:

pending_order: if this value is true, then we have placed an order, but it is not filled. This prevents a new order from being placed until the order is fulfilled. …

~vmodus

June 30th, 2020, 05:06 PM

Today's Results - live

Positive P&L at close, trading MES and MNQ (1 contract each)
2 open positions at close, both negative
12 round trip trades
42% profitable
Overall profit factor: 2.9
System efficiency: 51%
Execution errors: none
System errors: one, with a caveat

Today was okay. System efficiency was really good for a change and profit factor is was better. I'll take the positive P/L for the day. Yesterday was such a whack day that it has skewed this week. Things probably won't be back on level until Thursday, hopefully.

Execution was clean, no errors by me.

System Errors
There was one missed entry, which is okay, as I don't think the price ever hit. I solved a problem that I've been having when looking at backtested results versus live results. I will need to determine a method (programatically) for backtesting, but I know I cannot exactly use the built-in reports. In other words, I need to create my own strategy report to compare with live trading.

~ + ~ + ~ + ~ + ~ + ~ + ~ + ~ + ~ + ~ + ~ + ~ + ~ + ~ + ~ + ~ + ~

That's all for today. Let's do this again next month!

~vmodus

June 30th, 2020, 08:57 PM

Resolving Start of Week Entry Issue
I have run into an issue with some of my strategies, but in particular my latest, where my signals have not normalized (or settled down, if you will). So I have written a little bit of code to wait for 10 minutes before trading.

Here is the little snippet of EasyLanguage / PowerLanguage code that helps accomplish this little feat:

Code

// Do not trade the open
if Dayofweek(date) = 0 
   then
      begin 
         if 1810 < Time
            then 
               Condition52 = true 
            else 
               condition52 = false; 
      end 
   Else 
      Condition52 = true ;

I then add Condition52 to every buy and sellshort statement. If false, it prevents entry. Easy peasy lemon squeezy.

Start of Week Delay and End of Week Exit

If I had had this code in my strategy Sunday night, it would have saved me a lot of grief. It was a great lesson, though (tuition is steep).

~vmodus

July 6th, 2020, 11:21 AM

Sorry, no update late last week. I was busy with a lot of things.

I decided to put the alligator-fractal strategy on the shelf for now. Although sometimes profitable, the trading results did not match the entry signals. I strongly believe in the mantra, "what gets measured gets managed", and because I cannot accurately measure what the strategy is doing and get comparable results, I need to shelve it until I can dedicate more time to troubleshooting. I may need to rewrite from scratch and follow @Big Mike's recommendations on Renko charts (really old post here on FIO).

Anyhow, I am incubating another strategy with Renko bars (amongst other incubating strategies), so I will update the results of that strategy this week. It is a variant of one of my partner's strategies.

That's all for now. I will update later after market close.

~vmodus

Attack of the Robots - An Algo Journal

Discussion in Trading Journals

Attack of the Robots - An Algo Journal